Renaissance Technologies - Inside a Moneymaking Machine Like No Other

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
User avatar
matjen
Posts: 1925
Joined: Sun Nov 20, 2011 11:30 pm

Renaissance Technologies - Inside a Moneymaking Machine Like No Other

Post by matjen » Mon Nov 21, 2016 7:30 pm

So I generally describe myself as a Quant & Hold Investor. I believe the markets are generally efficient and that the Vanguard approach to inexpensive and diversified beta or the DFA/AQR/Swedroe approach of factor investing at a reasonable price are the best methods for 99.9% of investors. HOWEVER, what can you say about Renaissance? I mean these returns (even if a little inflated) are spectacular over a lengthy period. This isn't like Bill Miller where the returns are slightly better than the S&P500, and you would have had to be invested day 1 etc. This is a different magnitude. It does seem odd that their public funds don't even come close to matching them though. I wonder what they tell their investors about that.

https://www.bloomberg.com/news/articles ... ackest-box

Inside a Moneymaking Machine Like No Other
The Medallion Fund, an employees-only offering for the quants at Renaissance Technologies, is the blackest box in all of finance.

The fabled fund, known for its intense secrecy, has produced about $55 billion in profit over the last 28 years, according to data compiled by Bloomberg, making it about $10 billion more profitable than funds run by billionaires Ray Dalio and George Soros. What’s more, it did so in a shorter time and with fewer assets under management. The fund almost never loses money. Its biggest drawdown in one five-year period was half a percent.
A man is rich in proportion to the number of things he can afford to let alone.

livesoft
Posts: 62367
Joined: Thu Mar 01, 2007 8:00 pm

Re: Renaissance Technologies - Inside a Moneymaking Machine Like No Other

Post by livesoft » Mon Nov 21, 2016 7:32 pm

The book "The Quants" by Scott Patterson is something that folks who use AQR should read. It has write-ups of a handful of quant organizations including RT.
Wiki This signature message sponsored by sscritic: Learn to fish.

User avatar
David Jay
Posts: 5485
Joined: Mon Mar 30, 2015 5:54 am
Location: Michigan

Re: Renaissance Technologies - Inside a Moneymaking Machine Like No Other

Post by David Jay » Mon Nov 21, 2016 9:52 pm

And professional athletes get 9 digit contracts. Not actionable. Not even discuss-able, really, due to the black box nature of the fund.
Prediction is very difficult, especially about the future - Niels Bohr | To get the "risk premium", you really do have to take the risk - nisiprius

GuitarXM
Posts: 403
Joined: Sat Jun 18, 2016 12:09 am

RENAISSANCE TECHNOLOGIES Quant Trading

Post by GuitarXM » Mon Nov 21, 2016 10:54 pm

Not sure if any of you have heard of this fund but the performance is pretty amazing since 1988.
See for yourself,

https://www.bloomberg.com/news/articles ... ackest-box

Looks like they use quantitative trading. What I don't understand is, how is it that trading costs don't kill their performance?
How can trading be profitable since 1988? Their way of investing goes against all advice of the buy and hold investor?

I'm guessing they somehow use leverage to get that kind of return but whats crazy is that they pretty much had no down year since 1988. How can that be?

Not to mention that quant trading looks at past data similar to technical analysis to predict the future?
Hasn't it been proven time and time again that stock price can't be predicted based on past price data?
It seems that this fund has some secret up its sleeve.
Last edited by GuitarXM on Mon Nov 21, 2016 11:24 pm, edited 1 time in total.

User avatar
LadyGeek
Site Admin
Posts: 47520
Joined: Sat Dec 20, 2008 5:34 pm
Location: Philadelphia
Contact:

Re: Renaissance Technologies - Inside a Moneymaking Machine Like No Other

Post by LadyGeek » Mon Nov 21, 2016 11:13 pm

^^^ I merged GuitarXM's post into here.
Wiki To some, the glass is half full. To others, the glass is half empty. To an engineer, it's twice the size it needs to be.

APB
Posts: 113
Joined: Mon Apr 25, 2016 8:24 pm
Location: NYC

Re: Renaissance Technologies - Inside a Moneymaking Machine Like No Other

Post by APB » Mon Nov 21, 2016 11:49 pm

Interesting to see the outsized returns. Perhaps it indicates that those with PHDs running math departments and spending years codebreaking should consider quantitative trading? :)

Credit where it's due, Simons is worth $16.5 Billion and most of his life points to him being a genius. In hindsight, he had an incredible winning strategy, especially because he got the fees associated with the fund as well!

Of note, another quant firm that killed it for most of the 90s was Long Term Capital Management (LTCM). Here are their historic returns: https://upload.wikimedia.org/wikipedia/ ... c/LTCM.png

In order to have enough knowledge to pick the Medallion fund over LTCM, you'd likely need to know enough to reproduce the strategy in the first place. For what it's worth, all of the nobel laureates at LTCM had most of their own money tied up in LTCM!

Also of note, if somebody truly has a proven winning strategy, they aren't going to be beating down your door to give you a piece of the action. Medallion has not accepted any outside investment since 2005!
My posts represent my own opinion and do not constitute financial advice. I am simply a hobbyist. :)

GuitarXM
Posts: 403
Joined: Sat Jun 18, 2016 12:09 am

Re: Renaissance Technologies - Inside a Moneymaking Machine Like No Other

Post by GuitarXM » Tue Nov 22, 2016 12:35 am

LTCM lasted only 4 years, his fund is going strong since 1988.

What I don't understand is how trading costs don't kill the entire return.

It is most likely leveraged and maybe not even just stocks.
Its just interesting that it has outperformed for so long...

The stock market is a combination of emotions. This means that those models have to somehow predict and understand something as complex as human emotion. This would require unimaginable computing power that might not even exist today. I can't see how a computer can be smarter than a person. The code fails to see the bigger picture that a human brain can.

But since they have been outperforming since 1988, the computing power wasn't nearly as strong as it is today.
Something else must be behind this. Or it can simply be due to extraordinary luck. After all it is possible. His other funds haven't performed nearly as good.

lack_ey
Posts: 6605
Joined: Wed Nov 19, 2014 11:55 pm

Re: RENAISSANCE TECHNOLOGIES Quant Trading

Post by lack_ey » Tue Nov 22, 2016 1:01 am

GuitarXM wrote:Looks like they use quantitative trading. What I don't understand is, how is it that trading costs don't kill their performance?
How can trading be profitable since 1988?
Trading costs do eat into their performance. It's just high enough that they're profitable after those costs. Like all not-terrible quant shops, they model trading frictions and impact, and spend time trying to figure out more efficient ways to transact.
GuitarXM wrote:Their way of investing goes against all advice of the buy and hold investor?
Buy-and-hold is the way to get the underlying asset performance (minus fees/costs/taxes), nothing more. It's better than alternatives if you don't have a way to transact to add performance that survives the additional costs. In some sense random actions can't be expected to contribute any positive alpha on average before costs, and after costs it's negative. So you need some kind of edge, which they've quite apparently had.
GuitarXM wrote:I'm guessing they somehow use leverage to get that kind of return but whats crazy is that they pretty much had no down year since 1988. How can that be?
Yes, leverage. As for the performance, I'd look at it from two different angles. First of all, simply, if your mean return is really high, then a bad year (let's say, 10 percentile) may still be positive.

Secondly, you have to consider what kind of trades they're making and what they're investing in, at least at a high level. Generally, many asset classes make money in most (over half) of years but not in a lot of them. All the academic factors (size, value, momentum, quality, you name it) are losers in a lot of years too, similarly. My point is that these things are not consistent on a year-to-year basis. But they're not doing that, instead doing a range of trades from extremely short term (split second) to maybe some seasons, as reported. They do a lot of different strategies and hold all kinds of positions in a given year.

Also keep in mind they're trading in a lot of things other than just stocks. Their more public, higher capacity strategies using equities have performed a lot worse than Medallion.

Having more positions, if they all aren't correlated, can help with consistency. Imagine that you get to flip a coin a hundred times a year, where you make $1 on heads and lose $0.80 on tails. You're going to make money most years, but there will be some losing years where you end up with 56+ tails. Now imagine instead if you get a thousand flips a year of a payout of $0.10 on heads and lose $0.08 on tails. You're now virtually guaranteed to win in a year in net.

You just have to find the right trades that are actually favorable, perhaps some virtually guaranteed to work and maybe others likely to but still worth a shot. They aren't exactly forthcoming on what they do, but presumably they've found a number of things.

The best high-frequency traders make so many moves, many likely to be profitable, that they're pretty much always making money.
GuitarXM wrote:Not to mention that quant trading looks at past data similar to technical analysis to predict the future?
Hasn't it been proven time and time again that stock price can't be predicted based on past price data?
It seems that this fund has some secret up its sleeve.
First of all, a lot of quant trading looks at plenty more than just price (or performance), and of course so it goes for them as well. The Paris weather example clearly is not purely about price information with nothing more.

But it's not clear how widely you meant "stock price can't be predicted based on past price data." It is actually unquestionably true that past price movements tell you something about future price movements. For example, past recent volatility predicting future volatility far from perfectly but at a level way better than chance, assuming the average at all times. So you should ask if price movements tell you anything about the direction of future price movements. Historically, that's also been the case, very obviously so before costs. After costs, there's still some decent evidence across many different time periods, asset classes, and countries that price movements do contain some information about the future direction—not enough to be even remotely consistent, but right more often than wrong in a way that was enough to survive costs and make money in most years, even with pretty simple signals.

There's just obviously not some price signals that result in short-term (on the order of hours, days, weeks, etc.) profitability after trading costs in any way that's remotely consistent. Of course lots of people are looking for these kinds of things. So you're personally probably not going to find them: easy pickings have been picked clean, if they were ever there to begin with. Same for not-so-easy pickings.

qwertyjazz
Posts: 1077
Joined: Tue Feb 23, 2016 4:24 am

Re: Renaissance Technologies - Inside a Moneymaking Machine Like No Other

Post by qwertyjazz » Tue Nov 22, 2016 11:20 am

matjen wrote:So I generally describe myself as a Quant & Hold Investor. I believe the markets are generally efficient and that the Vanguard approach to inexpensive and diversified beta or the DFA/AQR/Swedroe approach of factor investing at a reasonable price are the best methods for 99.9% of investors. HOWEVER, what can you say about Renaissance? I mean these returns (even if a little inflated) are spectacular over a lengthy period. This isn't like Bill Miller where the returns are slightly better than the S&P500, and you would have had to be invested day 1 etc. This is a different magnitude. It does seem odd that their public funds don't even come close to matching them though. I wonder what they tell their investors about that.

https://www.bloomberg.com/news/articles ... ackest-box

Inside a Moneymaking Machine Like No Other
The Medallion Fund, an employees-only offering for the quants at Renaissance Technologies, is the blackest box in all of finance.

The fabled fund, known for its intense secrecy, has produced about $55 billion in profit over the last 28 years, according to data compiled by Bloomberg, making it about $10 billion more profitable than funds run by billionaires Ray Dalio and George Soros. What’s more, it did so in a shorter time and with fewer assets under management. The fund almost never loses money. Its biggest drawdown in one five-year period was half a percent.
Matjen
What is a Quant and Hold investor? I usually think of them as opposite unless it is something like pick a factor for long run
Curious
QJ
G.E. Box "All models are wrong, but some are useful."

GuitarXM
Posts: 403
Joined: Sat Jun 18, 2016 12:09 am

Re: Renaissance Technologies - Inside a Moneymaking Machine Like No Other

Post by GuitarXM » Tue Nov 22, 2016 12:48 pm

It is actually unquestionably true that past price movements tell you something about future price movements. For example, past recent volatility predicting future volatility far from perfectly but at a level way better than chance, assuming the average at all times
From what I have studied, past price movements tell you nothing about future price movements.
Past volatility does not predict future volatility either.
Beta studies historical price movements which have nothing to do with the future. That is one of the problems with CAPM, MPT, and efficient frontier. Standard deviation will change as new information comes to the market so it has no bearing on the future.

I have studied and used technical analysis to the fullest and after a decade of trial and error, I assure you there is no edge other than luck in using technical analysis. I tested every possible indicator known to man and even indicators outside investing such as Astrological indicators (Delta phenomenon). These indicators can provide correlation not causation by studying the past. The problem is that the future makes the correlation weaker as new information is unpredictable.

Humans are pattern seeking animals and will find patterns even in random data. Countless psychological studies have shown that and I was very impressed with a study that was shown in the book "Your money and your brain" that showed that pigeons are better traders than humans because they simply don't look for patterns. They don't have a cerebrum the size of ours and its amazing that in the experiment they beat humans.

http://money.cnn.com/2000/11/01/zweig_o ... /index.htm

What they are doing with that fund is most likely something that only High frequency trading firms can do. They place their super computers close to the stock exchange and make money off the spreads which in turn also make the market more liquid. I doubt they have any other advantage as over time any indicator becomes useless. If they can get in and out of positions within seconds and those positions are leveraged I can see how they can never lose money. Interesting enough, if he had a real money making machine, you would think his net worth wouldn't be 16 billion but more like 1 Trillion. I think that he wouldn't be able to do that with a large amount of money because it would move the price so much that he wouldn't have any advantage. This is probably the reason why the fund is not open to the public and is controlled by having a max of 10 billion invested. Any more than that, and their edge would disappear.

lack_ey
Posts: 6605
Joined: Wed Nov 19, 2014 11:55 pm

Re: Renaissance Technologies - Inside a Moneymaking Machine Like No Other

Post by lack_ey » Tue Nov 22, 2016 1:15 pm

GuitarXM wrote:
It is actually unquestionably true that past price movements tell you something about future price movements. For example, past recent volatility predicting future volatility far from perfectly but at a level way better than chance, assuming the average at all times
From what I have studied, past price movements tell you nothing about future price movements.
Past volatility does not predict future volatility either.
Beta studies historical price movements which have nothing to do with the future. That is one of the problems with CAPM, MPT, and efficient frontier. Standard deviation will change as new information comes to the market so it has no bearing on the future.
No, that's very provably false, or there's some kind of miscommunication here. Just check VIX and options pricing, if nothing else. If there's no predicitability then VIX not always being at the long-term average is a mispricing in options and there's a huge moneymaking opportunity for you.

There's something of an open question of which measure might be better than another, different autoregressive models etc. but the premise is not in doubt. Something like this doesn't happen by chance:
Image

That is almost certainly not just people finding "patterns" in noise.

GuitarXM wrote:What they are doing with that fund is most likely something that only High frequency trading firms can do. They place their super computers close to the stock exchange and make money off the spreads which in turn also make the market more liquid. I doubt they have any other advantage as over time any indicator becomes useless. If they can get in and out of positions within seconds and those positions are leveraged I can see how they can never lose money. Interesting enough, if he had a real money making machine, you would think his net worth wouldn't be 16 billion but more like 1 Trillion. I think that he wouldn't be able to do that with a large amount of money because it would move the price so much that he wouldn't have any advantage. This is probably the reason why the fund is not open to the public and is controlled by having a max of 10 billion invested. Any more than that, and their edge would disappear.
Supposedly they're not exactly HFT, not quite at that timescale. Maybe some overlap but that's not it completely, and they do take at least some positions much longer than HFT shops do.

GuitarXM
Posts: 403
Joined: Sat Jun 18, 2016 12:09 am

Re: Renaissance Technologies - Inside a Moneymaking Machine Like No Other

Post by GuitarXM » Tue Nov 22, 2016 4:15 pm

30 days isn't enough time to predict volatility.
Also standard deviation is a very general concept.
I'm not a math wiz, but I think 1st standard deviation contains almost 70% of all the prices and 2nd standard deviation includes almost 90% of the prices.
All that chart you posted shows is that the price is contained within that first band. It also doesn't show whether price was up or down, just a range of possible prices. Also note that as the average moves forward so does the band. So it basically adjusts to new information that hasn't been known before.

I agree with value investors on the subject of beta. If something goes down and its still worth the same amount it becomes less risky than more risky. The concept of riskiness based on past price action is nonsense. Imagine prior to 2008-09 crash you look at not 1 month of volatility by say 1 year and then try to predict future volatility of vix, the model will crash and burn. As the market crashes which can't be predicted, volatility increases and no vix model or any beta concept can predict that. The end result is that Beta concept has flaws. The only concept I agree with MPT is that without any analysis whether technical or fundamental, the least riskiest portfolio is the market portfolio not because of beta but because you are holding every single company. Though picking a portfolio of say 30 stocks with beta less than 1, wont guarantee you that in 5 years from now that beta will still be less than 1. That's the problem with looking backwards. A new day will bring new information and therefore update the Beta.

qwertyjazz
Posts: 1077
Joined: Tue Feb 23, 2016 4:24 am

Re: Renaissance Technologies - Inside a Moneymaking Machine Like No Other

Post by qwertyjazz » Tue Nov 22, 2016 4:31 pm

Best comment on this comparing utility of making money vs academic minutiae

http://marginalrevolution.com/marginalr ... ogies.html

Is finance really the best use for two string theorists?

I cannot understand physics or finance at this level - so index investing for me and a non-tenured job
G.E. Box "All models are wrong, but some are useful."

GuitarXM
Posts: 403
Joined: Sat Jun 18, 2016 12:09 am

Re: Renaissance Technologies - Inside a Moneymaking Machine Like No Other

Post by GuitarXM » Tue Nov 22, 2016 4:58 pm

Throughout history, usually, if something doesn't make sense it wont work.

Trader/Investor
Posts: 246
Joined: Wed Dec 02, 2015 9:35 pm

Re: Renaissance Technologies - Inside a Moneymaking Machine Like No Other

Post by Trader/Investor » Tue Nov 22, 2016 5:02 pm

matjen wrote:So I generally describe myself as a Quant & Hold Investor. I believe the markets are generally efficient and that the Vanguard approach to inexpensive and diversified beta or the DFA/AQR/Swedroe approach of factor investing at a reasonable price are the best methods for 99.9% of investors. HOWEVER, what can you say about Renaissance? I mean these returns (even if a little inflated) are spectacular over a lengthy period. This isn't like Bill Miller where the returns are slightly better than the S&P500, and you would have had to be invested day 1 etc. This is a different magnitude. It does seem odd that their public funds don't even come close to matching them though. I wonder what they tell their investors about that.

https://www.bloomberg.com/news/articles ... ackest-box

Inside a Moneymaking Machine Like No Other
The Medallion Fund, an employees-only offering for the quants at Renaissance Technologies, is the blackest box in all of finance.

The fabled fund, known for its intense secrecy, has produced about $55 billion in profit over the last 28 years, according to data compiled by Bloomberg, making it about $10 billion more profitable than funds run by billionaires Ray Dalio and George Soros. What’s more, it did so in a shorter time and with fewer assets under management. The fund almost never loses money. Its biggest drawdown in one five-year period was half a percent.
AQR has over 30 funds and their performance (with but one or two exceptions) has been average to dismal. Some have negative 5 year annualized returns. And the forum favorite here - QMHIX - is down 14.45% over the past year and has made only 3.04% annualized the past three years. These are faddish funds and only a kool-aid drinker would invest in them.

lack_ey
Posts: 6605
Joined: Wed Nov 19, 2014 11:55 pm

Re: Renaissance Technologies - Inside a Moneymaking Machine Like No Other

Post by lack_ey » Tue Nov 22, 2016 5:09 pm

GuitarXM wrote:30 days isn't enough time to predict volatility.
Also standard deviation is a very general concept.
I'm not a math wiz, but I think 1st standard deviation contains almost 70% of all the prices and 2nd standard deviation includes almost 90% of the prices.
All that chart you posted shows is that the price is contained within that first band. It also doesn't show whether price was up or down, just a range of possible prices. Also note that as the average moves forward so does the band. So it basically adjusts to new information that hasn't been known before.
This response doesn't make sense, or I am bad at reading. Maybe some combination of the two.

One thing you seem to be pointing out is that outcomes fall within 1 standard deviation of the mean about 70% of the time and within 2 standard deviations 90% of the time. The figures are approximately 68.3% and 95.5%, as commonly cited, but that's for a normal distribution. More generally that's not necessarily the case.

But that's not what is even being discussed. The right plot shows realized stock market (S&P 500) next-month daily volatility against the daily volatility of the previous month. They're shown to be significantly correlated with r=0.769 over the period. Among other things, this means that if the stock market is more volatile than usual in one month, it's more likely to be more volatile than average in the next month as well (than in a randomly selected month).

This in of itself doesn't imply any profitable trading strategy or tell you about direction. It's just an example that's a direct refutation of what you said earlier:
GuitarXM wrote:From what I have studied, past price movements tell you nothing about future price movements.
Past volatility does not predict future volatility either.
If correlation were more like 0, then sure. But it's not, and it's high enough with enough data points (there's plenty of other data not shown in the graph too) that it seems statistically extremely unlikely to all be a fluke with your hypothesis validated.

GuitarXM wrote:I agree with value investors on the subject of beta. If something goes down and its still worth the same amount it becomes less risky than more risky. The concept of riskiness based on past price action is nonsense. Imagine prior to 2008-09 crash you look at not 1 month of volatility by say 1 year and then try to predict future volatility of vix, the model will crash and burn. As the market crashes which can't be predicted, volatility increases and no vix model or any beta concept can predict that. The end result is that Beta concept has flaws.
There's no problem with beta, which is just a statistic that can be computed on available data. Any problem is with misinterpretation of statistics, drawing the wrong conclusions, etc. You can compute beta and know something about what happened in the past. You need some kind of model, informal or not, if you want to take that understanding to apply that to predictions about the future.

Just for reference:
Image

There's a big difference between knowing that the distribution has potentially shifted and knowing exactly when a crash is coming. Nobody here mentioned anything about being able to predict crashes so I wonder why you're off on this tangent.
GuitarXM wrote:The only concept I agree with MPT is that without any analysis whether technical or fundamental, the least riskiest portfolio is the market portfolio not because of beta but because you are holding every single company.
That's almost certainly wrong, at least in most understandings of "riskiest" from measures of volatility to drawdowns to anything along those lines.

There are a lot of ways to reweight the total market to make it less risky (and probably lower return as well). You can screen just based on price data, looking at volatility. Or look at fundamentals like operating leverage. Underweight the relatively highly leveraged companies in each sector and that would do the trick too. This isn't always going to result in less realized risk in any way you might define, but it should improve the odds. We're talking about what we can know ex-ante, probabilities. Not certainties, which nobody suggests there are.
GuitarXM wrote:Though picking a portfolio of say 30 stocks with beta less than 1, wont guarantee you that in 5 years from now that beta will still be less than 1. That's the problem with looking backwards. A new day will bring new information and therefore update the Beta.
Obviously no guarantee, but is your bet really that this portfolio of 30 stocks is just as likely to have a beta over 1 as below 1? I would take the under. If there's no information about the future in these statistics, as you seem to be saying elsewhere, then it shouldn't tell you anything. I doubt the informational content is zero. Probably less than a lot of people think, but not zero.

User avatar
matjen
Posts: 1925
Joined: Sun Nov 20, 2011 11:30 pm

Re: Renaissance Technologies - Inside a Moneymaking Machine Like No Other

Post by matjen » Tue Nov 22, 2016 5:32 pm

qwertyjazz wrote: Matjen
What is a Quant and Hold investor? I usually think of them as opposite unless it is something like pick a factor for long run
Curious
QJ
It just means that I subscribe to all the traditional Boglehead principles and also believe in factors. So my portfolio isn't cap weighted but I do rebalance to keep myself at the risk level I feel comfortable with. So for instance, rather than the traditional Three-Fund Portfolio you could have this instead and have a Small Cap Value Tilt and Momentum Screens if one has DFA access.

DFA US Vector Equity I DFVEX (instead of VTI Total Market)
DFA World ex US Targeted Val Instl DWUSX (instead of VXUS Total International)
Vanguard Intermediate-Term Govt Bd ETF VGIT or very high quality Muni Fund like BMBIX

If I was starting from scratch today the above would be my portfolio. Instead (along with the above) I have plenty of VTI and VXUS so I also have IJS/VSS for instance to give me the tilts before I had DFA access. Next bear I will switch them out if possible but can't now because of tax considerations.

Or you could add a bit of spice with a 5% allocation to an alternative fund which I also have.

AQR Style Premia Alternative I QSPIX
A man is rich in proportion to the number of things he can afford to let alone.

User avatar
matjen
Posts: 1925
Joined: Sun Nov 20, 2011 11:30 pm

Re: Renaissance Technologies - Inside a Moneymaking Machine Like No Other

Post by matjen » Tue Nov 22, 2016 5:44 pm

Trader/Investor wrote: AQR has over 30 funds and their performance (with but one or two exceptions) has been average to dismal. Some have negative 5 year annualized returns. And the forum favorite here - QMHIX - is down 14.45% over the past year and has made only 3.04% annualized the past three years. These are faddish funds and only a kool-aid drinker would invest in them.
I do believe Larry Swedroe is fond of QMHIX as of late but the AQR forum favorite is most definitely AQR Style Premia Alternative I QSPIX!
A man is rich in proportion to the number of things he can afford to let alone.

bigred77
Posts: 1997
Joined: Sat Jun 11, 2011 4:53 pm

Re: Renaissance Technologies - Inside a Moneymaking Machine Like No Other

Post by bigred77 » Tue Nov 22, 2016 5:45 pm

They seem to have been outlandishly successful, but none of us are able to invest with them, so I guess it doesn't matter much.

I will also note that it strikes me as odd that they have a super successful fund that is closed to outside investment, but there funds that do accept outside money are quite disappointing. One could see how the optics of this look bad right? A little "if it's too good to be true, it's probably not"?

Then there's the option that they have just been extremely lucky and are at risk of an extreme blow up one day. That has happened before too.

*3!4!/5!
Posts: 1256
Joined: Fri Sep 02, 2016 1:47 pm

Re: Renaissance Technologies - Inside a Moneymaking Machine Like No Other

Post by *3!4!/5! » Tue Nov 22, 2016 5:49 pm

GuitarXM wrote:From what I have studied, past price movements tell you nothing about future price movements.
Past volatility does not predict future volatility either.
Beta studies historical price movements which have nothing to do with the future. That is one of the problems with CAPM, MPT, and efficient frontier. Standard deviation will change as new information comes to the market so it has no bearing on the future.

I have studied and used technical analysis to the fullest and after a decade of trial and error, I assure you there is no edge other than luck in using technical analysis. I tested every possible indicator known to man and even indicators outside investing such as Astrological indicators (Delta phenomenon). These indicators can provide correlation not causation by studying the past. The problem is that the future makes the correlation weaker as new information is unpredictable.

Humans are pattern seeking animals and will find patterns even in random data. Countless psychological studies have shown that and I was very impressed with a study that was shown in the book "Your money and your brain" that showed that pigeons are better traders than humans because they simply don't look for patterns. They don't have a cerebrum the size of ours and its amazing that in the experiment they beat humans.
You've been making a lot of posts about the "technical" aspects of investing. I really haven't been able to make any sense out of what you're saying, but now I think I see where you're coming from.

randomguy
Posts: 6193
Joined: Wed Sep 17, 2014 9:00 am

Re: Renaissance Technologies - Inside a Moneymaking Machine Like No Other

Post by randomguy » Tue Nov 22, 2016 6:05 pm

bigred77 wrote:They seem to have been outlandishly successful, but none of us are able to invest with them, so I guess it doesn't matter much.

I will also note that it strikes me as odd that they have a super successful fund that is closed to outside investment, but there funds that do accept outside money are quite disappointing. One could see how the optics of this look bad right? A little "if it's too good to be true, it's probably not"?

Then there's the option that they have just been extremely lucky and are at risk of an extreme blow up one day. That has happened before too.
One thing that was mentioned in the article is that they limit the fund size to like 10 billion. Now that is a lot of money but it is a lot less than you would expect with growth levels like that. I am guessing the other funds they run don't have access to as many limited liquidity opportunities.

For fun read the Wiki page on the fund. Sounds like they have it set up so employees can hold this in ROTH IRAs.:)

GuitarXM
Posts: 403
Joined: Sat Jun 18, 2016 12:09 am

Re: Renaissance Technologies - Inside a Moneymaking Machine Like No Other

Post by GuitarXM » Tue Nov 22, 2016 6:08 pm

You've been making a lot of posts about the "technical" aspects of investing. I really haven't been able to make any sense out of what you're saying, but now I think I see where you're coming from.
That was to show you how extensive of a Technical Analysis approach that I tried.

My posts have not been about Technical Analysis but about fundamental analysis such as P/E ratios, Earning Yield, etc..
That has nothing to do with Technical Analysis.

When I talk about Technical aspects I am referring to the study of historic stock price data which includes price action, support/resistance, momentum indicators like MACD, Moving Averages, Elliot Wave Theory, Candlestick analysis, Stochastics, Fibbonaci Retracements, Bollinger Bands....and even something called Delta Phenomenon which originates from Astrological trading indicators. It's similar to Elliot Wave theory.

I have literally researched all technical indicators over the past 10 years, backtesting, and demo trading. The ones I named are just a few of the many that I tested. I tested so many I lost count. Hell you can invest back test all these indicators on MT4 platform free.
I have concluded that the Fundamental Analysts are correct when they say that past price movement can't be used to invest profitably. No matter how hard I tried, I could not get even a measurable statistical advantage by using past price data.
Its amazing how many people fall for the Technical Analysis trap, myself included.
Its hard to avoid it because its everywhere. Not to mention all the trading forums and all broker platforms have unlimited number of indicators.

I can tell you that its all nonsense and who ever claims that it works, I would like to see statistical evidence.

One thing is more or less certain, and that is the fact that as long as America is alive and well, the economy and its companies will keep on growing and so will their stock price. You don't need any analysis to invest into an index fund which is guaranteed to grow over the long term.

User avatar
snowshoes
Posts: 348
Joined: Tue Feb 03, 2015 12:33 pm

Re: Renaissance Technologies - Inside a Moneymaking Machine Like No Other

Post by snowshoes » Tue Nov 22, 2016 6:17 pm

-deleted- :happy
Last edited by snowshoes on Tue Nov 22, 2016 8:35 pm, edited 1 time in total.

bigred77
Posts: 1997
Joined: Sat Jun 11, 2011 4:53 pm

Re: Renaissance Technologies - Inside a Moneymaking Machine Like No Other

Post by bigred77 » Tue Nov 22, 2016 6:19 pm

GuitarXM wrote:
You've been making a lot of posts about the "technical" aspects of investing. I really haven't been able to make any sense out of what you're saying, but now I think I see where you're coming from.
That was to show you how extensive of a Technical Analysis approach that I tried.

My posts have not been about Technical Analysis but about fundamental analysis such as P/E ratios, Earning Yield, etc..
That has nothing to do with Technical Analysis.

When I talk about Technical aspects I am referring to the study of historic stock price data which includes price action, support/resistance, momentum indicators like MACD, Moving Averages, Elliot Wave Theory, Candlestick analysis, Stochastics, Fibbonaci Retracements, Bollinger Bands....and even something called Delta Phenomenon which originates from Astrological trading indicators. It's similar to Elliot Wave theory.

I have literally researched all technical indicators over the past 10 years, backtesting, and demo trading. The ones I named are just a few of the many that I tested. I tested so many I lost count. Hell you can invest back test all these indicators on MT4 platform free.
I have concluded that the Fundamental Analysts are correct when they say that past price movement can't be used to invest profitably. No matter how hard I tried, I could not get even a measurable statistical advantage by using past price data.
Its amazing how many people fall for the Technical Analysis trap, myself included.
Its hard to avoid it because its everywhere. Not to mention all the trading forums and all broker platforms have unlimited number of indicators.

I can tell you that its all nonsense and who ever claims that it works, I would like to see statistical evidence.

One thing is more or less certain, and that is the fact that as long as America is alive and well, the economy and its companies will keep on growing and so will their stock price. You don't need any analysis to invest into an index fund which is guaranteed to grow over the long term.
I think the reason technical analysis is so prevalent (and this is not an original theory mind you) is because financial markets spew out an unfathomable amount of numbers and data. People like to try to find patterns in data. That, plus greed, is why people keep on trying to find "the signal within the noise".

Trader/Investor
Posts: 246
Joined: Wed Dec 02, 2015 9:35 pm

Re: Renaissance Technologies - Inside a Moneymaking Machine Like No Other

Post by Trader/Investor » Tue Nov 22, 2016 6:29 pm

GuitarXM wrote:
You've been making a lot of posts about the "technical" aspects of investing. I really haven't been able to make any sense out of what you're saying, but now I think I see where you're coming from.
That was to show you how extensive of a Technical Analysis approach that I tried.

My posts have not been about Technical Analysis but about fundamental analysis such as P/E ratios, Earning Yield, etc..
That has nothing to do with Technical Analysis.

When I talk about Technical aspects I am referring to the study of historic stock price data which includes price action, support/resistance, momentum indicators like MACD, Moving Averages, Elliot Wave Theory, Candlestick analysis, Stochastics, Fibbonaci Retracements, Bollinger Bands....and even something called Delta Phenomenon which originates from Astrological trading indicators. It's similar to Elliot Wave theory.

I have literally researched all technical indicators over the past 10 years, backtesting, and demo trading. The ones I named are just a few of the many that I tested. I tested so many I lost count. Hell you can invest back test all these indicators on MT4 platform free.
I have concluded that the Fundamental Analysts are correct when they say that past price movement can't be used to invest profitably. No matter how hard I tried, I could not get even a measurable statistical advantage by using past price data.
Its amazing how many people fall for the Technical Analysis trap, myself included.
Its hard to avoid it because its everywhere. Not to mention all the trading forums and all broker platforms have unlimited number of indicators.

I can tell you that its all nonsense and who ever claims that it works, I would like to see statistical evidence.

One thing is more or less certain, and that is the fact that as long as America is alive and well, the economy and its companies will keep on growing and so will their stock price. You don't need any analysis to invest into an index fund which is guaranteed to grow over the long term.


I would not know where to begin to answer this. I've known literally scores of traders/researchers like yourself (and written about some of them) who immersed themselves 24/7 in all the indicators you list and more. Most are superfluous indicators pandered by the Dream Merchants to the gullible. None of these vendors ever made a penny trading the indicators they were pandering. The best quote I ever read about this was from Jack Schwager's The New Market Wizards in his interview Charles Faulkner regarding all these worthless indicators "The bottom line is that these methods seem to work only because the people who use them (successfully) have developed some sort of intuitive experience about price" As for me what has worked since 1985 is pure price action stripped of all the indicators that the left brained revel in slicing and dicing.

randomguy
Posts: 6193
Joined: Wed Sep 17, 2014 9:00 am

Re: Renaissance Technologies - Inside a Moneymaking Machine Like No Other

Post by randomguy » Tue Nov 22, 2016 6:55 pm

bigred77 wrote:
I think the reason technical analysis is so prevalent (and this is not an original theory mind you) is because financial markets spew out an unfathomable amount of numbers and data. People like to try to find patterns in data. That, plus greed, is why people keep on trying to find "the signal within the noise".
Medallion suggests there are signals. Now if those signals are in the financial data (i.e. all that crap about tops/bottoms/support levels/...) or outside (i.e. effects of weather patterns on ethanol costs), I am going to vote outstide. And obviously none of them are obvious (i.e. those have been price out). What a team of mathematicians doing signal analysis on a zillion parameters looking for find a dozen good ones does is bit different than the guy at home buying a stock because we hit the 90 day average.

Or it will turn out these guys are just great hackers and have been trading on insider info for 20 years now.:)

GuitarXM
Posts: 403
Joined: Sat Jun 18, 2016 12:09 am

Re: Renaissance Technologies - Inside a Moneymaking Machine Like No Other

Post by GuitarXM » Tue Nov 22, 2016 7:20 pm

That's true. You wouldn't know until they get caught haha

"The bottom line is that these methods seem to work only because the people who use them (successfully) have developed some sort of intuitive experience about price" As for me what has worked since 1985 is pure price action stripped of all the indicators that the left brained revel in slicing and dicing.
That is absolutely correct. Pure price action is the fastest indicator. All other indicators lag behind price action.
Thankfully I haven't lost any money to trading indicators because I was using a demo account.
But I now argue that price action itself doesn't indicate future direction and is useless.
Trading through price action will produce bad results. I'm sticking to long term. Buy and hold is the way to go.

The money I did lose was based on a so called tip that I received from a friend at work who invested in precious metal stocks successfully.
After reading a few books I thought it made sense. I didn't know any fundamental data at that point an paid a good price for my mistake. Never again.
It was a Uranium and Gold miner.

The gold miner was a disaster that went bankrupt.
The Uranium explorer is still in business and the fundamentals of Uranium as a whole looks promising due to increasing population and the need for clean energy around the world. However the stock as I have learned now is still a speculative trade due to the nature of its profits which obviously depend on the price of Uranium itself. I paid dearly and learned a good lesson from that. Another lesson was foreign currency risk. Canadian dollars are now worth 30% less than they were when I bought the Canadian Stock. The market is an expensive place to learn which is why I am educating myself with Fundamental Analysis, Accounting, etc. No more speculation for me. I learned my lesson the hard way.

lack_ey
Posts: 6605
Joined: Wed Nov 19, 2014 11:55 pm

Re: Renaissance Technologies - Inside a Moneymaking Machine Like No Other

Post by lack_ey » Tue Nov 22, 2016 7:30 pm

Are you still sticking by the claim of volatility not predicting future volatility?

GuitarXM
Posts: 403
Joined: Sat Jun 18, 2016 12:09 am

Re: Renaissance Technologies - Inside a Moneymaking Machine Like No Other

Post by GuitarXM » Tue Nov 22, 2016 7:38 pm

Until proven otherwise, Yes.

It logically doesn't even make sense.
I don't understand how a current price range can predict a future price action/range

Standard deviation follows follows the mean. It basically tracks the deviation of price around a moving average.
A moving average shows the past price data and changes as new information comes in and moves the price.
The standard deviation will also move with the moving average but it is no way predicting where the moving average will go.
It doesn't magically know if the moving average will go up or down. All it says is that with 1 standard deviation, the price will be within the boundaries of the standard deviation range about 65% of the time. If you take 2 standard deviation then it captures something like 95% of the price action.

But again, its HISTORIC price action around the moving average line. Where the price will go and how large or small the range might become nobody knows. Look up Bollinger Bands from John Bollinger I believe his name is. He made it popular to use it in Range Trading because the range is defined by 1 or 2 standard deviations. The idea behind it, is that if price action trades in a range, you can pick tops and bottoms with 2 standard deviations because 95% of the price is contained within the given range and odds are that price will not break out of that range. Sounds real good on paper. But guess what? Price can break out any time it wants to, because a trend can start at any moment. It would be beautiful to know in advance whether the price action will be range bound or trending. You would be able to make millions. Reality though is that standard deviation is a nicely mathematically drawn lines around the price which shows you nothing more or nothing less than PAST price action.

I love Jason Zweig's book in how he looks at charts and says that the right side of the chart can never be predicted. A chart shows what a stock HAS been doing, not what it WILL do. You can retrofit a chart with so many lines and indicators that it would seem that the system is unbreakable and give you signals that are 100% accurate without a single loss over many many years. Reality is quiet different though.

Can you please explain your point of view? How does standard deviation of past price action predict the future price action?

lack_ey
Posts: 6605
Joined: Wed Nov 19, 2014 11:55 pm

Re: Renaissance Technologies - Inside a Moneymaking Machine Like No Other

Post by lack_ey » Tue Nov 22, 2016 7:54 pm

I'm not talking about technical analysis, Bollinger bands, or any of that.

As brought up before, just look at the correlation between standard deviation of daily returns in the past month vs. standard deviation of daily returns in the subsequent month. You're effectively saying there are no relationships to be found, r=0.

I can run some more data myself if you want...?

GuitarXM
Posts: 403
Joined: Sat Jun 18, 2016 12:09 am

Re: Renaissance Technologies - Inside a Moneymaking Machine Like No Other

Post by GuitarXM » Tue Nov 22, 2016 8:03 pm

Bollinger bands are standard deviation bands.

From month to month, you will see correlation because its relatively short term.
I guess it also depends on the month. Because I bet the month before the 2008 crash, VIX wasn't predicting huge volatility.
Can you seriously imagine the amount of money that can be made in options if you can predict volatility?

Make sure you also read the criticism section on wikipedia as I am not the only one who thinks VIX is nonsense and so is BETA


"The VIX is quoted in percentage points and represents the expected range of movement in the S&P 500 index over the next year, at a 68% confidence level (i.e. one standard deviation of the normal probability curve). For example, if the VIX is 15, this represents an expected annualized change, with a 68% probability, of less than 15% up or down. One can calculate the expected volatility range for a single month from this figure by dividing the VIX figure of 15 not by 12, but by √12 which would imply a range of +/- 4.33% over the next 30-day period. [7] Similarly, expected volatility for a week would be 15 divided by √52, or +/- 2.08%."

There is a correlation between VIX and SP500.
Matter of fact I don't understand why its not 100% correlated as it should be.

But this correlation is on PAST price data not on FUTURE price data.
Just think logically... How can you predict future volatility? lol Its impossible...

Its like saying ok, the VIX shows the past volatility as 15 and it expects annualized change of 15% up or down with a certainty of 68%. That means the price will be within 1 standard deviation. Its true and SP500 will confirm that the price action has been within 1 standard deviation. But how will the future be even with 1% or 2% or the crazy 68% certainty that the price will be within that range. What if there is a terrorist attack? What if the markets crash like they did in 2008. The volatility will be much higher than 15. How can you say that the past will repeat. Its impossible

TomCat96
Posts: 510
Joined: Sun Oct 18, 2015 12:18 pm

Re: Renaissance Technologies - Inside a Moneymaking Machine Like No Other

Post by TomCat96 » Tue Nov 22, 2016 8:16 pm

matjen wrote:So I generally describe myself as a Quant & Hold Investor. I believe the markets are generally efficient and that the Vanguard approach to inexpensive and diversified beta or the DFA/AQR/Swedroe approach of factor investing at a reasonable price are the best methods for 99.9% of investors. HOWEVER, what can you say about Renaissance? I mean these returns (even if a little inflated) are spectacular over a lengthy period. This isn't like Bill Miller where the returns are slightly better than the S&P500, and you would have had to be invested day 1 etc. This is a different magnitude. It does seem odd that their public funds don't even come close to matching them though. I wonder what they tell their investors about that.

https://www.bloomberg.com/news/articles ... ackest-box

Inside a Moneymaking Machine Like No Other
The Medallion Fund, an employees-only offering for the quants at Renaissance Technologies, is the blackest box in all of finance.

The fabled fund, known for its intense secrecy, has produced about $55 billion in profit over the last 28 years, according to data compiled by Bloomberg, making it about $10 billion more profitable than funds run by billionaires Ray Dalio and George Soros. What’s more, it did so in a shorter time and with fewer assets under management. The fund almost never loses money. Its biggest drawdown in one five-year period was half a percent.

If someone or a group of people were to ever figure out a way to consistently beat the market, they would be smart to do everything these guys have done.

1. Shut up about it.
2. Quietly make your money.
3. Don't overplay your hand. Large capital flows attract analysis and scrutiny.
4. Minimize the number of people involved.

I believe it was speculated that the Black-Scholes model for options pricing was actually known prior to its discovery by other individuals who made money off of it.

lack_ey
Posts: 6605
Joined: Wed Nov 19, 2014 11:55 pm

Re: Renaissance Technologies - Inside a Moneymaking Machine Like No Other

Post by lack_ey » Tue Nov 22, 2016 8:28 pm

GuitarXM wrote:Bollinger bands are standard deviation bands.

From month to month, you will see correlation because its relatively short term.
I guess it also depends on the month. Because I bet the month before the 2008 crash, VIX wasn't predicting huge volatility.
Can you seriously imagine the amount of money that can be made in options if you can predict volatility?

Make sure you also read the criticism section on wikipedia as I am not the only one who thinks VIX is nonsense and so is BETA


"The VIX is quoted in percentage points and represents the expected range of movement in the S&P 500 index over the next year, at a 68% confidence level (i.e. one standard deviation of the normal probability curve). For example, if the VIX is 15, this represents an expected annualized change, with a 68% probability, of less than 15% up or down. One can calculate the expected volatility range for a single month from this figure by dividing the VIX figure of 15 not by 12, but by √12 which would imply a range of +/- 4.33% over the next 30-day period. [7] Similarly, expected volatility for a week would be 15 divided by √52, or +/- 2.08%."

There is a correlation between VIX and SP500.
Matter of fact I don't understand why its not 100% correlated as it should be.

But this correlation is on PAST price data not on FUTURE price data.
Just think logically... How can you predict future volatility? lol Its impossible...

Its like saying ok, the VIX shows the past volatility as 15 and it expects annualized change of 15% up or down with a certainty of 68%. That means the price will be within 1 standard deviation. Its true and SP500 will confirm that the price action has been within 1 standard deviation. But how will the future be even with 1% or 2% or the crazy 68% certainty that the price will be within that range. What if there is a terrorist attack? What if the markets crash like they did in 2008. The volatility will be much higher than 15. How can you say that the past will repeat. Its impossible
Okay, if you see non-negligible correlation in the short term then that contradicts your earlier assertion that "past volatility does not predict future volatility." It's true that one coin flip does not predict the outcome of future coin flips. It seems very untrue that past volatility does not predict future volatility.

Obviously there is still a lot of unpredictability remaining, everything from the terrorist attack to the much more mundane that might help cause one outcome as opposed to another.

Maybe it's a miscommunication resulting from an understanding of the word "predict" in this context. I usually mean it in a statistical sense, of predictive power for a model, allowing an estimate to be improved (could be from terrible/no information at all to just plain bad/low information). Not of a sure thing.

And nobody was suggesting VIX is useful and highly successful for market timing or fairly accurately predicting crashes or anything else. I just contend that it predicts future short-term volatility to some degree. If it didn't, that would imply that options pricing is way wrong and there is plenty of money to be made on options.

But for what it's worth:
Image

Again, VIX, beta, etc. are all statistics that can be computed. Statistics in of themselves don't imply anything about the future. It's bad interpretations, poor models, etc. that are the problem. You have to say which interpretations you have an issue with, what it's bad for.

GuitarXM
Posts: 403
Joined: Sat Jun 18, 2016 12:09 am

Re: Renaissance Technologies - Inside a Moneymaking Machine Like No Other

Post by GuitarXM » Tue Nov 22, 2016 9:16 pm

The point I'm trying to make is that beta, or vix is unable to provide any edge.
It will not increase your profitability and like you said yourself you can't use it to time the market and gain any competitive advantage.

If it truly was able to predict the future, people would use it with very good results and we won't need funds like Renaissance

Then what is the purpose of beta? People argue that by selecting a portfolio of stocks with beta less than 1 will be less risky.

The entire CAPM models rests on the theory of Beta and risk.
This concept has many flaws and I argue that past Beta can't predict future price action. It can't say that this stock is less risky because it has been less risky in the past. Past and future are two different time and will also have 2 different betas. It doesn't mean that going forward the stock will be less risky. You can't determine that from past price action. You can however determine risk by studying and understanding the company and its sources of profits. And even that isnt full proof.

From a fundamental approach the entire beta concept makes no sense.

TomCat96
Posts: 510
Joined: Sun Oct 18, 2015 12:18 pm

Re: Renaissance Technologies - Inside a Moneymaking Machine Like No Other

Post by TomCat96 » Tue Nov 22, 2016 9:45 pm

lack_ey wrote:
GuitarXM wrote:Bollinger bands are standard deviation bands.

From month to month, you will see correlation because its relatively short term.
I guess it also depends on the month. Because I bet the month before the 2008 crash, VIX wasn't predicting huge volatility.
Can you seriously imagine the amount of money that can be made in options if you can predict volatility?

Make sure you also read the criticism section on wikipedia as I am not the only one who thinks VIX is nonsense and so is BETA


"The VIX is quoted in percentage points and represents the expected range of movement in the S&P 500 index over the next year, at a 68% confidence level (i.e. one standard deviation of the normal probability curve). For example, if the VIX is 15, this represents an expected annualized change, with a 68% probability, of less than 15% up or down. One can calculate the expected volatility range for a single month from this figure by dividing the VIX figure of 15 not by 12, but by √12 which would imply a range of +/- 4.33% over the next 30-day period. [7] Similarly, expected volatility for a week would be 15 divided by √52, or +/- 2.08%."

There is a correlation between VIX and SP500.
Matter of fact I don't understand why its not 100% correlated as it should be.

But this correlation is on PAST price data not on FUTURE price data.
Just think logically... How can you predict future volatility? lol Its impossible...

Its like saying ok, the VIX shows the past volatility as 15 and it expects annualized change of 15% up or down with a certainty of 68%. That means the price will be within 1 standard deviation. Its true and SP500 will confirm that the price action has been within 1 standard deviation. But how will the future be even with 1% or 2% or the crazy 68% certainty that the price will be within that range. What if there is a terrorist attack? What if the markets crash like they did in 2008. The volatility will be much higher than 15. How can you say that the past will repeat. Its impossible
Okay, if you see non-negligible correlation in the short term then that contradicts your earlier assertion that "past volatility does not predict future volatility." It's true that one coin flip does not predict the outcome of future coin flips. It seems very untrue that past volatility does not predict future volatility.

Obviously there is still a lot of unpredictability remaining, everything from the terrorist attack to the much more mundane that might help cause one outcome as opposed to another.

Maybe it's a miscommunication resulting from an understanding of the word "predict" in this context. I usually mean it in a statistical sense, of predictive power for a model, allowing an estimate to be improved (could be from terrible/no information at all to just plain bad/low information). Not of a sure thing.

And nobody was suggesting VIX is useful and highly successful for market timing or fairly accurately predicting crashes or anything else. I just contend that it predicts future short-term volatility to some degree. If it didn't, that would imply that options pricing is way wrong and there is plenty of money to be made on options.

But for what it's worth:
Image

Again, VIX, beta, etc. are all statistics that can be computed. Statistics in of themselves don't imply anything about the future. It's bad interpretations, poor models, etc. that are the problem. You have to say which interpretations you have an issue with, what it's bad for.

FWIW, i greatly enjoyed the discourse between you and GuitarXM.

I can see both sides of the argument.

In support GuitarXM's objection. That image of the performance of the VIX you quoted from wikipedia appears to stand for the proposition that "VIX has virtually the same predictive power as past volatility, insofar as the shown correlation coefficients are nearly identical."

Your contention was this: Past volatility is correlated with Future volatility, that there's some predictive power there.

But my objection is how strongly does your graph really show that? The second diagram graphs the S&P current month stddev with the next month's stddev. The relationship seems to hold around the .01 stddev mark, and even the .02stddev mark. But it quickly breaks down around the .03std mark and onward.

In other words, what is the relationship? If the S&P STDDEV is .03 this month. From reading the graph what can we expect from the STDDEV of the SP 500 next month? some of the time, next month, the STDDEV the following month was .02, sometimes it was .03. Curiously at no point was it .04, and sometimes it was .05.

The only "correlation" I can see the graph showing is that if the STDDEV of the S&P was .02 or below this month, more often than not, it was .02 or below the following month. The problem with that correlation though is that it fails to account for another explanation, namely that by and large the STDDEV of the market just isn't that high.

This is what I mean. Let's say I engage you in conversation. A measure of how loud each syllable I utter is made. Some syllables are emphasized and some or softer. Unless I'm going nuts, I'm not about to yell out every other syllable in conversation. I might yell out an expletive every now and then but by and large most of the things I say are going to fall within a certain range. In other words, if one syllable has one level of volume, it's very likely that the next syllable will have a similar level of volume. I wouldn't call that predictive power so much as the raw data of how normal conversation goes.

In fact, let's consider the Standard deviation of the standard deviation of the S&P point returns in the graph you showed comparing the VIX and the S&P prior month standard deviation. What the right graph seems to show me is that most of the time, the standard deviation of the S&P usually hovers around .01 or .02. In fact it hovers around .01 or .02 so much that I might assert, 68% of the time, it hangs out around .01, and 95% it hangs out around .02. The standard deviation of the standard deviation of the S&P points seems to follow a bell-curve.

The strongest evidence I see in your favor is this, for extremely low std devs one month (.01), there were zero std devs the next month significantly higher (.05). That IS indeed interesting!

If i were to use that information you posted, this is the extent of the predictive power I would make. If the STDDEV of the S&P was extremely low this month. It is extremely unlikely it will be extremely high next month. In other words, if the volatility of the market is at record lows, it has never been the case that next month volatility will be at record highs.

randomguy
Posts: 6193
Joined: Wed Sep 17, 2014 9:00 am

Re: Renaissance Technologies - Inside a Moneymaking Machine Like No Other

Post by randomguy » Tue Nov 22, 2016 10:21 pm

TomCat96 wrote:

If someone or a group of people were to ever figure out a way to consistently beat the market, they would be smart to do everything these guys have done.

1. Shut up about it.
2. Quietly make your money.
3. Don't overplay your hand. Large capital flows attract analysis and scrutiny.
4. Minimize the number of people involved.

I believe it was speculated that the Black-Scholes model for options pricing was actually known prior to its discovery by other individuals who made money off of it.

The other way is to go big, suck up say 200 billion in AUM, have good returns for a couple of years and skim 5%&40% off the top and retire when you lose your edge.

The length of outperformance is what is impressive. I have heard of other companies put up returns like this over short periods of time (say 5-10 years) by being a couple of years ahead in things like high speed trading and the like or VC fund that does abnomrally well for 10 years but having almost 30 years of returns with 1 losing year is pretty outrageous.

GuitarXM
Posts: 403
Joined: Sat Jun 18, 2016 12:09 am

Re: Renaissance Technologies - Inside a Moneymaking Machine Like No Other

Post by GuitarXM » Tue Nov 22, 2016 11:36 pm

the VIX represents one measure of the market's expectation of stock market volatility over the next 30-day period.
True. But the expectation changes on the following day, and the day after, and also the day after the day after.
The strongest evidence I see in your favor is this, for extremely low std devs one month (.01), there were zero std devs the next month significantly higher (.05). That IS indeed interesting!
You will most likely not see 0.05 deviation often because that would mean the price is way way outside the price range. It makes perfect sense that both charts show most price action captured by 2 standard deviations which suppose to capture 95% of the prices in its range. VIX and SP500 are very much correlated with each other when looking backwards.
The VIX is quoted in percentage points and represents the expected range of movement in the S&P 500 index over the next year, at a 68% confidence level (i.e. one standard deviation of the normal probability curve). For example, if the VIX is 15, this represents an expected annualized change, with a 68% probability
So with 1 standard deviation, you can say that 68% of the time the sp500 can be predicted to move up or down 15%
Sounds about right since average stock returns are about 10%.
But note thats average. What happened in 2008? The prediction was wrong. Simple as that.
Performance of VIX (left) compared to past volatility (right) as 30-day volatility predictors, for the period of Jan 1990-Sep 2009. Volatility is measured as the standard deviation of S&P500 one-day returns over a month's period. The blue lines indicate linear regressions, resulting in the correlation coefficients r shown. Note that VIX has virtually the same predictive power as past volatility, insofar as the shown correlation coefficients are nearly identical.
What does that mean? That means VIX is no different at predicting the future as the past standard deviation of SP500 price volatility.
That whole statement doesn't really make much sense because obviously you can't predict future volatility by looking at the past volatility.
If nothing changes, yes, volatility will not change. But we all know that the market doesn't always stay the same due to new information constantly hitting the market. So is VIX correlated with SP500 price action. Yes. Does it predict the future?

Does the SP500 price action and volatility predict the future? I think you understand where I am coming from.

Now I am not saying that I'm right. I may be wrong. But someone has to prove to me how VIX or BETA or Standard Deviation can predict future price movement/volatility. If it can't be logically proven, it doesn't make sense. I honestly don't know much about stock and know nothing about options. I am just trying to understand this through common sense and common sense says that nothing we know of has any predictive power that can give you a trading edge. Surely there are correlations when back testing the data. The question though is how long does that correlation last? The Renaissance fund is impressive because it seems as if it beat all odds. We may never know.

I see why Warren Buffet is right when he says the stuff they teach at Wharton Business school is nonsense.

Here are a few articles I found:

http://fspinvest.co.za/articles/strateg ... -6475.html
http://dare.uva.nl/cgi/arno/show.cgi?fid=215486

lack_ey
Posts: 6605
Joined: Wed Nov 19, 2014 11:55 pm

Re: Renaissance Technologies - Inside a Moneymaking Machine Like No Other

Post by lack_ey » Wed Nov 23, 2016 1:45 am

TomCat96 wrote:FWIW, i greatly enjoyed the discourse between you and GuitarXM.

I can see both sides of the argument.

In support GuitarXM's objection. That image of the performance of the VIX you quoted from wikipedia appears to stand for the proposition that "VIX has virtually the same predictive power as past volatility, insofar as the shown correlation coefficients are nearly identical."

Your contention was this: Past volatility is correlated with Future volatility, that there's some predictive power there.

But my objection is how strongly does your graph really show that? The second diagram graphs the S&P current month stddev with the next month's stddev. The relationship seems to hold around the .01 stddev mark, and even the .02stddev mark. But it quickly breaks down around the .03std mark and onward.

In other words, what is the relationship? If the S&P STDDEV is .03 this month. From reading the graph what can we expect from the STDDEV of the SP 500 next month? some of the time, next month, the STDDEV the following month was .02, sometimes it was .03. Curiously at no point was it .04, and sometimes it was .05.

The only "correlation" I can see the graph showing is that if the STDDEV of the S&P was .02 or below this month, more often than not, it was .02 or below the following month. The problem with that correlation though is that it fails to account for another explanation, namely that by and large the STDDEV of the market just isn't that high.

This is what I mean. Let's say I engage you in conversation. A measure of how loud each syllable I utter is made. Some syllables are emphasized and some or softer. Unless I'm going nuts, I'm not about to yell out every other syllable in conversation. I might yell out an expletive every now and then but by and large most of the things I say are going to fall within a certain range. In other words, if one syllable has one level of volume, it's very likely that the next syllable will have a similar level of volume. I wouldn't call that predictive power so much as the raw data of how normal conversation goes.

In fact, let's consider the Standard deviation of the standard deviation of the S&P point returns in the graph you showed comparing the VIX and the S&P prior month standard deviation. What the right graph seems to show me is that most of the time, the standard deviation of the S&P usually hovers around .01 or .02. In fact it hovers around .01 or .02 so much that I might assert, 68% of the time, it hangs out around .01, and 95% it hangs out around .02. The standard deviation of the standard deviation of the S&P points seems to follow a bell-curve.

The strongest evidence I see in your favor is this, for extremely low std devs one month (.01), there were zero std devs the next month significantly higher (.05). That IS indeed interesting!

If i were to use that information you posted, this is the extent of the predictive power I would make. If the STDDEV of the S&P was extremely low this month. It is extremely unlikely it will be extremely high next month. In other words, if the volatility of the market is at record lows, it has never been the case that next month volatility will be at record highs.
IMHO you should just cite correlation (Pearson's product-moment correlation, r) in cases like this rather than arbitrarily segmenting and interpreting results more qualitatively, graphically, unless something seems off in the data. The latter approach is prone to overfitting peculiarities of that data set, overinterpreting noise. It's not like you can't see points outside of the ones shown in the scatterplot; that's just what happened in that period.

I wouldn't read incredibly much into it. My main point there is just that r is safely not 0 over the period.


By the way, I just did a quick sanity check on a different data set. I took Nikkei 225 (Japanese stock market) price returns going back to 1984—total returns would be better but over short periods, dividends aren't going to change all that much so probably doesn't make much of a difference—and subdivided into 20-trading-day intervals, which roughly corresponds to a month. Then calculated standard deviation of the daily returns for each of those 20-day sets.

Running a correlation test of the standard deviation vs. next-period standard deviation, I get r = 0.575 with a 95% confidence interval of 0.506 to 0.637 (t-stat of 14.1).

As briefly mentioned earlier, there are different autoregressive models that are likely a bit better than just taking past recent volatility, but that's a bit outside of the point and scope of this post.

Anyway, there's still a lot of uncertainty just in what plain volatility is going to look like, never mind which direction the market is going to go. But this is clearly predictive in the statistical sense.

GuitarXM wrote:So with 1 standard deviation, you can say that 68% of the time the sp500 can be predicted to move up or down 15%
Sounds about right since average stock returns are about 10%.
But note thats average. What happened in 2008? The prediction was wrong. Simple as that.
This is a very strange and I think might be said to be straight-up incorrect interpretation of statistics and probabilistic models. You can't just binarize and evaluate on a single outcome. That's like the weather forecast saying a 40% chance of rain and it rains and you say the forecast is wrong. No, if the forecast is accurate, the next 100 times it says there's a 40% chance of rain, it should rain in somewhere around 40 of them, of course the exact number being a bit up to chance. The weather forecast has predictive ability, even if it doesn't actually know any outcome for sure.
GuitarXM wrote:What does that mean? That means VIX is no different at predicting the future as the past standard deviation of SP500 price volatility.
That whole statement doesn't really make much sense because obviously you can't predict future volatility by looking at the past volatility.
If nothing changes, yes, volatility will not change. But we all know that the market doesn't always stay the same due to new information constantly hitting the market. So is VIX correlated with SP500 price action. Yes. Does it predict the future?
All these potential changes and uncertainties are already captured in the figures cited before. When over the period in the graph, last month vol had a correlation of 0.77 with next-month vol, that includes all the times when the next month's volatility is way different from the previous month's, including all those cases in 2008. After all, the correlation wasn't 1.
GuitarXM wrote:Does the SP500 price action and volatility predict the future? I think you understand where I am coming from.
Again, yes it does in the sense of providing information that can improve some predictions about the future (make the magnitude of error smaller on average). But not by any incredible amount, not in this particular way in a sense that can generate profitable trades any more likely.


Though if you look at all the data on time-series momentum across many different decades/asset classes/countries, it seems highly likely that actually there's some pure price over time predicting future price (extremely far from consistently, but at a rate high enough to be profitably tradable long-term unless things change pretty drastically in the future). Perhaps not for everything, and seemingly not really for individual stocks, but I think the data should at least cast some serious doubts. For reference, I'd say the data on vol predicting vol is more or less indisputable, at least that it was a phenomenon in the past. Markets do fundamentally change and evolve over time, so it's always a little of a stretch to assume that relationships will persist in the future, so the certainty is smaller going forward, but it would be a pretty drastic shift and I don't see a reason why this would go away.

Tanelorn
Posts: 1521
Joined: Thu May 01, 2014 9:35 pm

Re: Renaissance Technologies - Inside a Moneymaking Machine Like No Other

Post by Tanelorn » Wed Nov 23, 2016 3:56 am

Good article, thanks. As for some of the discussion, volatility is clearly auto correlated. This isn't obviously profitable since you can't bet on volatility directly, so it's not like knowing the future price of a stock. You can buy VIX futures, but those are priced based on the market expectations, which include this correlation, so there's no free money that way.

paper200
Posts: 260
Joined: Sat Feb 02, 2008 11:40 am

Re: Renaissance Technologies - Inside a Moneymaking Machine Like No Other

Post by paper200 » Wed Nov 23, 2016 11:34 am

The fund growth for $10K invested in 1988 Jan 1st would have reached $148Million @39.3% CAGR in 2016. The poor customer compounded at 18.3% to $1.3Million. Investing the 5% fees and 20%/44% gain every year (DCA) from the $10K invested for the customer the fund manager grew the zero value to $147million. The Managers were willing to :sharebeer but not :moneybag !! By 1994 the manager starting from zero in 1988 had more to invest than the client. The managers were brilliant for getting the client 18.3% CAGR over 29 years but understood compounding better and predicted the market trends and, in addition, their own success. S&P 500 total returns compounded from 1988 to 2015 at about 10.2% :( .

From a Boglehead perspective:
a) This is a rare event signal from outer space - most of us won't get a shot at this even though we understand "compounding".
b) John Bogle saved a lot of customers from being siphoned of hard earned wealth by teaching us how much a 1% differential can impact your investment return.
c) And then he says do not look for SETI signal for your investment growth since that ain't coming, i.e. back to point (a)
Having freedom, food and roof is being 90% lucky in life and so is index investing. So, don't let the remaining 10% bother you.

GuitarXM
Posts: 403
Joined: Sat Jun 18, 2016 12:09 am

Re: Renaissance Technologies - Inside a Moneymaking Machine Like No Other

Post by GuitarXM » Wed Nov 23, 2016 1:06 pm

Try running a correlation on a yearly basis instead of monthly or daily basis and see what the correlation is.
Let's see how well the model predicts volatility of previous 1 year compared to the future volatility

Tanelorn
Posts: 1521
Joined: Thu May 01, 2014 9:35 pm

Re: Renaissance Technologies - Inside a Moneymaking Machine Like No Other

Post by Tanelorn » Wed Nov 23, 2016 2:20 pm

paper200 wrote:The fund growth for $10K invested in 1988 Jan 1st would have reached $148Million @39.3% CAGR in 2016. The poor customer compounded at 18.3% to $1.3Million.
No, the 39% average is after the fees. It's like 70-80% or something before fees.

garlandwhizzer
Posts: 1943
Joined: Fri Aug 06, 2010 3:42 pm

Re: Renaissance Technologies - Inside a Moneymaking Machine Like No Other

Post by garlandwhizzer » Wed Nov 23, 2016 3:31 pm

matjen wrote:
Inside a Moneymaking Machine Like No Other
The Medallion Fund, an employees-only offering for the quants at Renaissance Technologies, is the blackest box in all of finance.
I believe that a smallish group of very sophisticated investors can substantially outperform the market as long as their total asset base remains small enough to be nimble and take advantage of promising low asset capacity investments. Warren Buffett and Charlie Munger had incredibly great results early along until their asset base became too large to be accommodated in the limited space of poorly researched and overlooked small cap value stocks. Now their asset base is so large that they must of necessity deploy most assets to only large/mega cap companies which are intensely scrutinized by professional investors and likely possess less potential for outsized gains than the overlooked and under-researched small cap value diamonds in the rough which Buffett and Munger were so good at finding.

The fact that Medallion is an employee only quant fund is applicable. They keep their trades secret and do not suffer from huge public capital inflows that can quickly exceed the available supply of alpha. What goes unsaid is that the existence of the Medallion Fund in itself represents an obvious conflict of interest. I suspect that the fund's managers keep the best trades to themselves rather than sharing it with their publicly available etfs so as to avoid overcrowded trades. Medallion keeps its best insights reserved for employees only in a well insulated black box capable of taking advantage of strategies that too much capital inflow would dilute. It is likely that excellent investment opportunities with limited capacity for capital inflows wind up with Medallion which directly benefits the employees of Renaissance Tech and that such insights are not shared by their commercial products for public consumption from which they also derive fees which benefit them. It is a brilliant example of how the financial industry does a fabulous job of lining its own pockets and a rather less successful job of lining the pockets of its clients/customers, the investing public. As the saying goes, "Where are the customer's yachts?"

Garland Whizzer

lack_ey
Posts: 6605
Joined: Wed Nov 19, 2014 11:55 pm

Re: Renaissance Technologies - Inside a Moneymaking Machine Like No Other

Post by lack_ey » Wed Nov 23, 2016 4:21 pm

GuitarXM wrote:Try running a correlation on a yearly basis instead of monthly or daily basis and see what the correlation is.
Let's see how well the model predicts volatility of previous 1 year compared to the future volatility
Okay, I haven't seen the results and am going to check the French data library for US stock returns. He has daily factor returns going back to 1926 for the US market. If you add Mkt-Rf and Rf, that's Mkt, the market return.

There are roughly 251 trading days a year these days, but it wasn't always that way. I could segment by year, but let's just do it by 251-day groupings. Hope you don't mind. So that comes out to be 95 non-overlapping periods of 251 trading days each (so I guess a month is more like 21 trading days on average, not 20), with a couple left over we'll ignore. Now, we can compute the standard deviation of the daily returns for each period.

Doing a test for correlation for those figures for one period to the next's, I see r = 0.673 with t = 8.74, a 95% confidence interval of 0.545 to 0.771. Wow, that was a lot higher than I was expecting.

Running the same test on just the last 30 years, since 1986: r = 0.580 with t = 3.71, a 95% confidence interval of 0.272 to 0.781. Less data, obviously, so lower confidence. May not be everything, though, but there doesn't seem to be some obvious shift over time in this relationship, either from this or graphically or some other tests.

So yeah, volatility we see in one year seems to predict the kind of volatility we'll see in the next. Not too strongly but to some degree. By the way, part of the volatility calculation for the 251 days obviously includes recent results, say the last 30 or so. And part of the figure for the next 251-day period includes the next month ahead, so you might ask if it just reflects the short term and might reverse later. The answer is seemingly no. There are a number of different things you could look at, but just checking the correlation of one period with two periods over (so a whole 251 trading days skipped in between), I see r = 0.387, t = 4.00.

Again, you could look at any number of other things, and this isn't much of a model or any serious analysis. Just checking correlations.

GuitarXM
Posts: 403
Joined: Sat Jun 18, 2016 12:09 am

Re: Renaissance Technologies - Inside a Moneymaking Machine Like No Other

Post by GuitarXM » Wed Nov 23, 2016 6:20 pm

Interesting. I am sure there is an explanation for this correlation.
But I still can't understand how a back test can predict the future.
It doesn't make any sense at all.

GuitarXM
Posts: 403
Joined: Sat Jun 18, 2016 12:09 am

Re: Renaissance Technologies - Inside a Moneymaking Machine Like No Other

Post by GuitarXM » Wed Nov 23, 2016 6:23 pm

Warren Buffett and Charlie Munger had incredibly great results early along until their asset base became too large to be accommodated in the limited space of poorly researched and overlooked small cap value stocks.
I don't know how safe it is to buy the small cap stocks. There is always a chance that since these stocks are overlooked that even if they have more value than what they trading at, it is possible that these stocks wont appreciate to their true value as these stocks are overlooked. Graham states this concept in "The Intelligent Investor" and he says that it is much safer to buy stocks of top grade companies with a large market share and large market cap.

lack_ey
Posts: 6605
Joined: Wed Nov 19, 2014 11:55 pm

Re: Renaissance Technologies - Inside a Moneymaking Machine Like No Other

Post by lack_ey » Wed Nov 23, 2016 7:18 pm

GuitarXM wrote:Interesting. I am sure there is an explanation for this correlation.
But I still can't understand how a back test can predict the future.
It doesn't make any sense at all.
Prices are usually more volatile when uncertainty is higher, when news of the day make a greater impact in terms of projected long-term prospects and profitability, when sentiments are less consistent, and so on. None of these to me sound like conditions that happen randomly with no pattern. The underlying economics are somewhat cyclical too—if there's been a recession the last 6 months, the next month is more likely than the average month to also still be in a recession.

I mean, if people are freaking out about stocks the last couple months and the prices are bouncing around (perhaps some economic distress or some other shock too), do you really expect average behavior going forward? Say, suddenly all the things people were freaking out over were resolved, with the next month being no more likely to be crazy than a given month say ten years from now? Seriously? Of course anything can happen but my bet would be on elevated levels of crazy (happening with a chance higher than the average). Most times aren't crazy but these tend to cluster together.

Doesn't mean you can predict anything accurately or forecast black swans or any other kind of animal. But there are at least some patterns in a statistical sense.

To me it would be much weirder if there were no correlation.

GuitarXM
Posts: 403
Joined: Sat Jun 18, 2016 12:09 am

Re: Renaissance Technologies - Inside a Moneymaking Machine Like No Other

Post by GuitarXM » Wed Nov 23, 2016 9:54 pm

Maybe. We can only speculate on what the truth really is. None of us are academic professionals.
I need to do some more research on this topic but I see many flaws with past price action analysis.

I wonder what Larry would say about this. He comes on here a lot.

vencat
Posts: 211
Joined: Thu Sep 10, 2009 6:30 pm

Re: Renaissance Technologies - Inside a Moneymaking Machine Like No Other

Post by vencat » Wed Nov 23, 2016 10:29 pm

'I wonder what Larry would say about this. He comes on here a lot.'

This was my first thought....hope he takes the bait.

lack_ey
Posts: 6605
Joined: Wed Nov 19, 2014 11:55 pm

Re: Renaissance Technologies - Inside a Moneymaking Machine Like No Other

Post by lack_ey » Wed Nov 23, 2016 10:45 pm

You can try PMing him after Thanksgiving.

I will note that the idea of vol predicting vol (or similar for beta) is very non-controversial, also said by others coming through the thread. In different contexts he's posted about this before:
viewtopic.php?f=10&t=179351&p=2716107&h ... t#p2716107
viewtopic.php?f=10&t=177763&p=2690504&h ... t#p2690504
viewtopic.php?f=10&t=176134&p=2682783&h ... t#p2682783
viewtopic.php?f=10&t=197807&p=3024479&h ... t#p3024479

Here, again, to be clear, "predict" is meant in the statistical sense (improving your information), not of certainty.


For the original subject of RenTech, nobody really knows what they're doing specifically, as has been established. This side discussion was just an example of the markets having some patterns, some predictability in some ways that we can define and discuss. This particular pattern is not an exploitable (profitable) one and is not at all in conflict with EMH or anything else.

RenTech clearly is on to something with a range of profitable trades with some nontrivial but apparently bounded capacity. Well, somebody's got to be out there making the markets more efficient.

Post Reply