Thanks for linking to this video. If this is not a reaffirmation to not abandon this strategy due to inflation concerns I don't know what is. You heard it right from the Fed Chairs mouth. This is not the 1970's, policy has changed since then, and we will fulfill our mandate of keeping inflation around 2% on average. I for one am not selling . Who do I trust more the WSJ or the Fed? Don't fight the FedOohLaLa wrote: ↑Tue Jun 01, 2021 11:12 pm It's alright guys! JPow to the rescue!
Very interesting comments that Robot Monster linked in another thread:
https://www.youtube.com/watch?app=desktop&v=zG4BmzD_T8A
Comments about "tools" to keep inflation in check (I imagine some good ol' quantitative tightening). Also, his comments about the nature of the inflation right now is in line with my gut feeling, whatever that's worth. lol
Now the question is: who do you trust more, the government or WSJ?
HEDGEFUNDIE's excellent adventure Part II: The next journey
Re: HEDGEFUNDIE's excellent adventure Part II: The next journey
VT & HFEA
Re: HEDGEFUNDIE's excellent adventure Part II: The next journey
No reason to panic: it's the Wall Street Journal quoting folks running inflation protection ETFs (PFIX, IVOl and the like) so basically umbrella merchants telling us it's going to rain...Gui0507 wrote: ↑Tue Jun 01, 2021 10:49 pmCouldn’t read it since I’m not subscribed to WSJ. So what you are saying or rather what its saying is that TMF for example will not provide any insurance as a hedge for UPRO when and if a market crash with rising rates since bonds notoriously perform poorly in that scenario?taojaxx wrote: ↑Tue Jun 01, 2021 10:35 pm Not sure there's a pay wall but today's WSJ has an interesting article on the belief that all underpinnings of HFEA as regards the Fed mandate and policy action are now a thing of the past.
https://www.wsj.com/articles/thewalls ... lead_pos13
Basically inflation returning and positive stocks bonds correlation with no more rate cuts possible when stocks crash.
OP's nightmare and ours too
Damn, I just started this less than a month ago
They have another article today saying inflation bets are losing steam...
Better lucky than smart.
Re: HEDGEFUNDIE's excellent adventure Part II: The next journey
I've been 100% invested in 60/40 UPRO/TMF for almost two years, but few folks have even the slightest idea of what that phrase even means. To keep things simple, whenever friends and family ask me about my investments, I tell them I've taken out a triply leveraged long bet on the United States Federal Reserve.Ramjet wrote: ↑Wed Jun 02, 2021 7:37 amThanks for linking to this video. If this is not a reaffirmation to not abandon this strategy due to inflation concerns I don't know what is. You heard it right from the Fed Chairs mouth. This is not the 1970's, policy has changed since then, and we will fulfill our mandate of keeping inflation around 2% on average. I for one am not selling . Who do I trust more the WSJ or the Fed? Don't fight the FedOohLaLa wrote: ↑Tue Jun 01, 2021 11:12 pm It's alright guys! JPow to the rescue!
Very interesting comments that Robot Monster linked in another thread:
https://www.youtube.com/watch?app=desktop&v=zG4BmzD_T8A
Comments about "tools" to keep inflation in check (I imagine some good ol' quantitative tightening). Also, his comments about the nature of the inflation right now is in line with my gut feeling, whatever that's worth. lol
Now the question is: who do you trust more, the government or WSJ?

 Posts: 4
 Joined: Wed Mar 10, 2021 3:32 pm
Re: HEDGEFUNDIE's excellent adventure Part II: The next journey
There's been some discussion about the feasibility for implementing this strategy using futures. One disadvantage of doing it is that the treasury futures have six figure notional values. I don't have that much Roth space yet to try that out.
The people behind tastyworks made a new exchange focused on futures with small notional sizes, called the small exchange. They have a future /s30y which has a notional value of $2,300, and I'm trying to figure out if it would work out as a HFEA style hedge. The future is based on the 2, 10, or 30 year interest rate. Since it is based on the interest rate, you would sell a contract to get the equivalent of buying the bonds.
They also have stock futures, but they track their own indices, the small stocks 75 and the small technology 60.
The main downside I see is that the futures have much lower liquidity than the more commonly used futures.
The people behind tastyworks made a new exchange focused on futures with small notional sizes, called the small exchange. They have a future /s30y which has a notional value of $2,300, and I'm trying to figure out if it would work out as a HFEA style hedge. The future is based on the 2, 10, or 30 year interest rate. Since it is based on the interest rate, you would sell a contract to get the equivalent of buying the bonds.
They also have stock futures, but they track their own indices, the small stocks 75 and the small technology 60.
The main downside I see is that the futures have much lower liquidity than the more commonly used futures.
Re: HEDGEFUNDIE's excellent adventure Part II: The next journey
Haha, love it!cos wrote: ↑Wed Jun 02, 2021 8:56 pmI've been 100% invested in 60/40 UPRO/TMF for almost two years, but few folks have even the slightest idea of what that phrase even means. To keep things simple, whenever friends and family ask me about my investments, I tell them I've taken out a triply leveraged long bet on the United States Federal Reserve.Ramjet wrote: ↑Wed Jun 02, 2021 7:37 amThanks for linking to this video. If this is not a reaffirmation to not abandon this strategy due to inflation concerns I don't know what is. You heard it right from the Fed Chairs mouth. This is not the 1970's, policy has changed since then, and we will fulfill our mandate of keeping inflation around 2% on average. I for one am not selling . Who do I trust more the WSJ or the Fed? Don't fight the FedOohLaLa wrote: ↑Tue Jun 01, 2021 11:12 pm It's alright guys! JPow to the rescue!
Very interesting comments that Robot Monster linked in another thread:
https://www.youtube.com/watch?app=desktop&v=zG4BmzD_T8A
Comments about "tools" to keep inflation in check (I imagine some good ol' quantitative tightening). Also, his comments about the nature of the inflation right now is in line with my gut feeling, whatever that's worth. lol
Now the question is: who do you trust more, the government or WSJ?
VT & HFEA
Re: HEDGEFUNDIE's excellent adventure [risk parity strategy using 3x leveraged ETFs]
Thanks for sharing your thoughts! I'll check out EDV.MotoTrojan wrote: ↑Sun May 30, 2021 8:50 amSince that post I had swapped to 43/57 UPRO/EDV, which I still think is as much better way to approach this moving forward (bit less leverage, lower overall expenseratio and borrowing costs, less volatility decay). The 43/57 UPRO/EDV has about the same volatility ratio as 55/45 UPRO/TMF, just with ~78% of the overall leverage.olympia_t wrote: ↑Fri May 28, 2021 12:43 pmSorry to revive this old thread  wondering if you could share any perspective for someone considering 55/45 now? Have been considering it but have been sitting on the sidelines a bit too long worrying we're in a bubble. Thanks so much!MotoTrojan wrote: ↑Mon Aug 12, 2019 6:03 pm
Yup, I am the posterchild of what not to do, but it has paid of handsomely and now I can buy & hold.
I got out entirely after a ~45% overall gain on what was initially 28% of my portfolio after feeling the expected return (assuming rates stayed flat) was not worth the risk, in particular the behavioral risk if rates started to rise swiftly. I also was gaining more knowledge and conviction in systematic value investing and wanted to make room to pursue that more deeply. I have since gone off the deepend and am a fullbore factorjunkie, rocking a 100% tilted portfolio.
As to your bubble comment, this is supposed to be a bubbleresistant strategy thanks to the longtreasuries. It is obviously still highly exposed to equity though so will still see downside pressure.
Small cap/value tilt? Would love to hear about your portfolio now. I'm still trying to figure out my overall plan. Think I might have some small segments for some of these strategies.
Re: HEDGEFUNDIE's excellent adventure Part II: The next journey
I put one of my Roth IRAs  a total of 1015% of our net worth  in a 60/40 HFEA on 09/30/20, right in time to get totally boned by TMFs drop. I got a little lucky with rebalancing a few days late in January  missing one of the ridiculous drops  and then in April  when I was at 70/30 overall  decided to skip rebalancing. Since then, I've been hovering with total returns basically identical to what would have happened if I just left it in a total stock fund the whole time.
I'm happy to say that today, finally, I'm ahead of where I would have been had I not started this adventure (and had just left it in the total stock fund). By about 0.4%. Currently at 73% UPRO/27% TMF.
Not regretting it so far, just saying, the ups and downs have been... interesting.
I'm happy to say that today, finally, I'm ahead of where I would have been had I not started this adventure (and had just left it in the total stock fund). By about 0.4%. Currently at 73% UPRO/27% TMF.
Not regretting it so far, just saying, the ups and downs have been... interesting.
Re: HEDGEFUNDIE's excellent adventure Part II: The next journey
Well, you definitely jumped into the pool at an "interesting" time for this type of strategy. 2x3x will be underwhelming in a choppy period like 2021.Raryn wrote: ↑Fri Jun 04, 2021 10:54 am I put one of my Roth IRAs  a total of 1015% of our net worth  in a 60/40 HFEA on 09/30/20, right in time to get totally boned by TMFs drop. I got a little lucky with rebalancing a few days late in January  missing one of the ridiculous drops  and then in April  when I was at 70/30 overall  decided to skip rebalancing. Since then, I've been hovering with total returns basically identical to what would have happened if I just left it in a total stock fund the whole time.
I'm happy to say that today, finally, I'm ahead of where I would have been had I not started this adventure (and had just left it in the total stock fund). By about 0.4%. Currently at 73% UPRO/27% TMF.
Not regretting it so far, just saying, the ups and downs have been... interesting.
Don't lose hope based on this, though, because when the market breaks out into an uptrend, the difference is staggering. The same can be said for a drawdown, but let's not think about that kind of scenario now, shall we? haha
Re: HEDGEFUNDIE's excellent adventure Part II: The next journey
I came to the same conclusion, eventually. I'm doing regular 60:40 in a taxable account, and I've been experimenting separately with a small allocation of 60:40 UPRO:TMF (currently 27.8% CAGR).
How should we think systematically about the risk:reward of a HFEA allocation within a "traditional" 60:40 AA? For example;
Code: Select all
60% VTI: 40% BND [Jul 2009  May 2021, q rebal] => CAGR 11.38% stddev 8.64%
54% VTI: 6% UPRO: 36% BND: 4% TMF [Jul 2009  May 2021, q rebal] => CAGR 13.87% stddev 9.78%
48% VTI: 12% UPRO: 32% BND: 8% TMF [Jul 2009  May 2021, q rebal] => CAGR 16.36% stddev 11.07%
Thoughts on a sustainable allocation to HFEA in the real world?
/FIRE55
Re: HEDGEFUNDIE's excellent adventure Part II: The next journey
Are you sure those numbers are correct? The third AA's CAGR seems way too high. I replicated it and it's not even 10%:FIRE55 wrote: ↑Mon Jun 07, 2021 7:02 pmI came to the same conclusion, eventually. I'm doing regular 60:40 in a taxable account, and I've been experimenting separately with a small allocation of 60:40 UPRO:TMF (currently 27.8% CAGR).
How should we think systematically about the risk:reward of a HFEA allocation within a "traditional" 60:40 AA? For example;
Is it ultimately about human tolerance to the eventual drawdown? How much more would I freak out if my portfolio declined 11% vs. 8%? Risk Parity and Maximization of Sharpe/Sortino Ratio seem to be leading statistical approaches.Code: Select all
60% VTI: 40% BND [Jul 2009  May 2021, q rebal] => CAGR 11.38% stddev 8.64% 54% VTI: 6% UPRO: 36% BND: 4% TMF [Jul 2009  May 2021, q rebal] => CAGR 13.87% stddev 9.78% 48% VTI: 12% UPRO: 32% BND: 8% TMF [Jul 2009  May 2021, q rebal] => CAGR 16.36% stddev 11.07%
Thoughts on a sustainable allocation to HFEA in the real world?
/FIRE55
https://www.portfoliovisualizer.com/bac ... tion8_2=32
Your approach is definitely sustainable, IMO. I highly doubt that 1.4x total exposure, like in your 3rd option, will cause failure (whether longterm underperformance or total loss). It's a very light version of the original HFEA, and even further diversified.
Re: HEDGEFUNDIE's excellent adventure Part II: The next journey
You went back ten years earlier in time; FIRE55 used UPRO and TMF directly. The big outperformance was basically from 2009 or start of 2012...OohLaLa wrote: ↑Mon Jun 07, 2021 8:16 pm Are you sure those numbers are correct? The third AA's CAGR seems way too high. I replicated it and it's not even 10%:
https://www.portfoliovisualizer.com/bac ... tion8_2=32
Re: HEDGEFUNDIE's excellent adventure Part II: The next journey
Huh... would you look at that. I did. Thanks for spotting that. Must be mellowed out with my herb tea.Hydromod wrote: ↑Mon Jun 07, 2021 8:45 pmYou went back ten years earlier in time; FIRE55 used UPRO and TMF directly. The big outperformance was basically from 2009 or start of 2012...OohLaLa wrote: ↑Mon Jun 07, 2021 8:16 pm Are you sure those numbers are correct? The third AA's CAGR seems way too high. I replicated it and it's not even 10%:
https://www.portfoliovisualizer.com/bac ... tion8_2=32
Btw, FIRE55, you can go all the way back to 1992 with my setup, if you want to get a general idea of your AAs.
Re: HEDGEFUNDIE's excellent adventure Part II: The next journey
Hedgefundie originally recommended to make a onetime contribution, not to risk so much that if it goes south it will materially change your retirement picture (he did 10%), and to view it as a lottery ticket separately from your portfolio. This is a really good way to compartmentalize a strategy that has high volatility. Personally, this works for me. I don't want this mixed in with my regular investments. If you are doing a light version of the Excellent Adventure, maybe PSLDX or some of the other Pimco Stocks Plus funds would work for you?FIRE55 wrote: ↑Mon Jun 07, 2021 7:02 pmI came to the same conclusion, eventually. I'm doing regular 60:40 in a taxable account, and I've been experimenting separately with a small allocation of 60:40 UPRO:TMF (currently 27.8% CAGR).
How should we think systematically about the risk:reward of a HFEA allocation within a "traditional" 60:40 AA? For example;
Is it ultimately about human tolerance to the eventual drawdown? How much more would I freak out if my portfolio declined 11% vs. 8%? Risk Parity and Maximization of Sharpe/Sortino Ratio seem to be leading statistical approaches.Code: Select all
60% VTI: 40% BND [Jul 2009  May 2021, q rebal] => CAGR 11.38% stddev 8.64% 54% VTI: 6% UPRO: 36% BND: 4% TMF [Jul 2009  May 2021, q rebal] => CAGR 13.87% stddev 9.78% 48% VTI: 12% UPRO: 32% BND: 8% TMF [Jul 2009  May 2021, q rebal] => CAGR 16.36% stddev 11.07%
Thoughts on a sustainable allocation to HFEA in the real world?
/FIRE55
Last edited by Ramjet on Tue Jun 08, 2021 8:11 am, edited 1 time in total.
VT & HFEA
Re: HEDGEFUNDIE's excellent adventure Part II: The next journey
My 1 tip  Don't try to time the rebalancing.Raryn wrote: ↑Fri Jun 04, 2021 10:54 am I put one of my Roth IRAs  a total of 1015% of our net worth  in a 60/40 HFEA on 09/30/20, right in time to get totally boned by TMFs drop. I got a little lucky with rebalancing a few days late in January  missing one of the ridiculous drops  and then in April  when I was at 70/30 overall  decided to skip rebalancing. Since then, I've been hovering with total returns basically identical to what would have happened if I just left it in a total stock fund the whole time.
I'm happy to say that today, finally, I'm ahead of where I would have been had I not started this adventure (and had just left it in the total stock fund). By about 0.4%. Currently at 73% UPRO/27% TMF.
Not regretting it so far, just saying, the ups and downs have been... interesting.
Re: HEDGEFUNDIE's excellent adventure Part II: The next journey
I hear ya. That's what I've been doing so far  this can go to zero and I won't cringe too much. But if we fundamentally believe in the traditional 60:40 portfolio, then why is this so fundamentally different? I've been in HFEA for a couple of years, so I absolutely don't buy the usual arguments of "daily use only" or "leverage drag" or whatever. I'm sitting on 27.8% CAGR right now. I am 100% going to run my taxable account as a rebalanced 60:40 portfolio (broadly) so why not add some leverage using a portion of UPRO:TMF?Ramjet wrote: ↑Tue Jun 08, 2021 7:54 amHedgefundie originally recommended to make a onetime contribution, not to risk so much that if it goes south it will materially change your retirement picture (he did 10%), and to view it as a lottery ticket separately from your portfolio. This is a really good way to compartmentalize a strategy that has high volatility. Personally, this works for me. I don't want this mixed in with my regular investments. If you are doing a light version of the Excellent Adventure, maybe PSLDX or some of the other Pimco Stocks Plus funds would work for you?FIRE55 wrote: ↑Mon Jun 07, 2021 7:02 pm
I came to the same conclusion, eventually. I'm doing regular 60:40 in a taxable account, and I've been experimenting separately with a small allocation of 60:40 UPRO:TMF (currently 27.8% CAGR).
How should we think systematically about the risk:reward of a HFEA allocation within a "traditional" 60:40 AA? For example;
Is it ultimately about human tolerance to the eventual drawdown? How much more would I freak out if my portfolio declined 11% vs. 8%? Risk Parity and Maximization of Sharpe/Sortino Ratio seem to be leading statistical approaches.Code: Select all
60% VTI: 40% BND [Jul 2009  May 2021, q rebal] => CAGR 11.38% stddev 8.64% 54% VTI: 6% UPRO: 36% BND: 4% TMF [Jul 2009  May 2021, q rebal] => CAGR 13.87% stddev 9.78% 48% VTI: 12% UPRO: 32% BND: 8% TMF [Jul 2009  May 2021, q rebal] => CAGR 16.36% stddev 11.07%
Thoughts on a sustainable allocation to HFEA in the real world?
/FIRE55
Other than the collapse of the ETF itself, I'm still struggling to see where the risk of total failure may come from. Underperformance  sure. I'm prepared to take that risk. But failure? Not feeling it.
So  how much leverage to allocate? That must be a broader topic within Asset Allocation, right?
/FIRE55
Re: HEDGEFUNDIE's excellent adventure Part II: The next journey
I think there is a much higher chance of slight underperformance or slight overperformance compared to the S&P 500 than of a total failure. Stagflation is the real worry with this strategy and I just am not very concerned with that happening. I remember Hedgefundie mentioning that if LTT rates are yielding more that 3X ITT rates that would likely be a warning sign that economic conditions have changed and this strategy is in danger of not performing well in the future. I am in agreement with you that a total failure isn't likely. Besides stagflation, human emotion is of course the enemy as well. Some will have a hard time hitting that rebalance button in a recession.FIRE55 wrote: ↑Tue Jun 08, 2021 11:19 amI hear ya. That's what I've been doing so far  this can go to zero and I won't cringe too much. But if we fundamentally believe in the traditional 60:40 portfolio, then why is this so fundamentally different? I've been in HFEA for a couple of years, so I absolutely don't buy the usual arguments of "daily use only" or "leverage drag" or whatever. I'm sitting on 27.8% CAGR right now. I am 100% going to run my taxable account as a rebalanced 60:40 portfolio (broadly) so why not add some leverage using a portion of UPRO:TMF?Ramjet wrote: ↑Tue Jun 08, 2021 7:54 amHedgefundie originally recommended to make a onetime contribution, not to risk so much that if it goes south it will materially change your retirement picture (he did 10%), and to view it as a lottery ticket separately from your portfolio. This is a really good way to compartmentalize a strategy that has high volatility. Personally, this works for me. I don't want this mixed in with my regular investments. If you are doing a light version of the Excellent Adventure, maybe PSLDX or some of the other Pimco Stocks Plus funds would work for you?FIRE55 wrote: ↑Mon Jun 07, 2021 7:02 pm
I came to the same conclusion, eventually. I'm doing regular 60:40 in a taxable account, and I've been experimenting separately with a small allocation of 60:40 UPRO:TMF (currently 27.8% CAGR).
How should we think systematically about the risk:reward of a HFEA allocation within a "traditional" 60:40 AA? For example;
Is it ultimately about human tolerance to the eventual drawdown? How much more would I freak out if my portfolio declined 11% vs. 8%? Risk Parity and Maximization of Sharpe/Sortino Ratio seem to be leading statistical approaches.Code: Select all
60% VTI: 40% BND [Jul 2009  May 2021, q rebal] => CAGR 11.38% stddev 8.64% 54% VTI: 6% UPRO: 36% BND: 4% TMF [Jul 2009  May 2021, q rebal] => CAGR 13.87% stddev 9.78% 48% VTI: 12% UPRO: 32% BND: 8% TMF [Jul 2009  May 2021, q rebal] => CAGR 16.36% stddev 11.07%
Thoughts on a sustainable allocation to HFEA in the real world?
/FIRE55
Other than the collapse of the ETF itself, I'm still struggling to see where the risk of total failure may come from. Underperformance  sure. I'm prepared to take that risk. But failure? Not feeling it.
So  how much leverage to allocate? That must be a broader topic within Asset Allocation, right?
/FIRE55
VT & HFEA
Re: HEDGEFUNDIE's excellent adventure Part II: The next journey
Bookmarked, very cool, thanks. While the absolute CAGR from 1999 is lower as expected, the relative advantage of 23% CAGR is still there when we suballocate a portion of HFEA. I'm still trying to think sensibly about "how much" in terms of risk:reward.
/FIRE55
Re: HEDGEFUNDIE's excellent adventure Part II: The next journey
Agreed. Here's my attitude to rebalancing in a down market: viewtopic.php?p=5883897#p5883897
/FIRE55
Re: HEDGEFUNDIE's excellent adventure Part II: The next journey
Yeah, definitely. Lower levels of exposure (sub 2x), seem to be able to eke out that 23% with similar or lower drawdowns. I think we always need to watch out because leverage costs can go up longerterm, inflation can wreck havoc on LTT and can cause increased correlation between LTT and stocks to increase, bond rates can go up (no gravy train like over the last 10+ years) etc. All things that can cause a leveraged position to underperform, where that 23% might not be in our favor.FIRE55 wrote: ↑Tue Jun 08, 2021 11:37 amBookmarked, very cool, thanks. While the absolute CAGR from 1999 is lower as expected, the relative advantage of 23% CAGR is still there when we suballocate a portion of HFEA. I'm still trying to think sensibly about "how much" in terms of risk:reward.
/FIRE55
Like you, I strongly doubt that a level of leverage like that can cause total loss.
I never bookmarked it, but some people had shared a good study of what the optimal leverage was, at least historically, for different indices. Can anybody please share that again?
Re: HEDGEFUNDIE's excellent adventure Part II: The next journey
The one thing that was brought up and discussed a while back that you should consider when your intent is to maintain leverage as a small part of the portfolio: you'd probably be better off just bumping up the leverage on the treasuries and increasing the allocation to unlevered equities. That's conceptually fairly close to what NTSX/NTSI/NTSE does.
For example, replacing a 60/40 VTI/BND allocation with something like 75/25 VTI/EDV or 90/10 VTI/TMF.
That way you get higher returns from the VTI with a similar treasury insurance policy, and maintain simplicity.
For example, replacing a 60/40 VTI/BND allocation with something like 75/25 VTI/EDV or 90/10 VTI/TMF.
That way you get higher returns from the VTI with a similar treasury insurance policy, and maintain simplicity.
Re: HEDGEFUNDIE's excellent adventure Part II: The next journey
I do recall that, now that you mention it. A somewhat simpler approach but the same basic question remains  how to choose those % allocations sensibly?Hydromod wrote: ↑Tue Jun 08, 2021 1:06 pm The one thing that was brought up and discussed a while back that you should consider when your intent is to maintain leverage as a small part of the portfolio: you'd probably be better off just bumping up the leverage on the treasuries and increasing the allocation to unlevered equities. That's conceptually fairly close to what NTSX/NTSI/NTSE does.
For example, replacing a 60/40 VTI/BND allocation with something like 75/25 VTI/EDV or 90/10 VTI/TMF.
That way you get higher returns from the VTI with a similar treasury insurance policy, and maintain simplicity.
75/25 VTI/EDV pairs stocks with long duration treasurys, but not leveraged  right?
/FIRE55
Re: HEDGEFUNDIE's excellent adventure Part II: The next journey
I have these in my bookmarks.
http://ddnum.com/articles/leveragedETFs.php
https://rhsfinancial.com/2017/06/20/lin ... leverage/
/FIRE55
Re: HEDGEFUNDIE's excellent adventure Part II: The next journey
As has been mentioned before, I'd HIGHLY caution against using these as a historical reference for leveraged ETF performance, because (based on the Simulating LETF Returns thread) they likely overstate performance. Their best fit models don't adequately account for cost of financing leverage among other expenses. Today, that amount is relatively inconsequential, but historically it would've been a huge drain on performance that's not really represented in these backtests.FIRE55 wrote: ↑Tue Jun 08, 2021 3:21 pmI have these in my bookmarks.
http://ddnum.com/articles/leveragedETFs.php
https://rhsfinancial.com/2017/06/20/lin ... leverage/
/FIRE55
Re: HEDGEFUNDIE's excellent adventure Part II: The next journey
Taking a look at the monthly reset leveraged mutual funds. Direxion's 2x monthly products were originally 2.5x and were changed to 2x during the depths of the financial crisis, explaining part of the underperformance. Aside from that, they perform very closely with the daily reset products, which isn't exactly how it should be. Rydex's RMQHX has outperformed DXQLX. Over a 2 percent higher CAGR. Not bad at all. RMQHX has smoked it's daily leveraged counterpart QLD despite the higher ER. Around 3.5 percent higher CAGR since 2015.
I wish Rydex would introduce monthly reset S&P 500/600/400 index funds.
I wish Rydex would introduce monthly reset S&P 500/600/400 index funds.
Last edited by Bearbug on Tue Jun 08, 2021 4:15 pm, edited 1 time in total.
Re: HEDGEFUNDIE's excellent adventure Part II: The next journey
EDV is not leveraged, it's extra long duration treasuries instead, so even more volatility than TLT. It has roughly mimicked TYD (3x ITTs), but with only nominal ER. My rule of thumb is that 2.3 TYD has the approximate volatility of 1 TMF when doing risk parity, so I'd expect about the same for EDV.FIRE55 wrote: ↑Tue Jun 08, 2021 3:14 pmI do recall that, now that you mention it. A somewhat simpler approach but the same basic question remains  how to choose those % allocations sensibly?Hydromod wrote: ↑Tue Jun 08, 2021 1:06 pm The one thing that was brought up and discussed a while back that you should consider when your intent is to maintain leverage as a small part of the portfolio: you'd probably be better off just bumping up the leverage on the treasuries and increasing the allocation to unlevered equities. That's conceptually fairly close to what NTSX/NTSI/NTSE does.
For example, replacing a 60/40 VTI/BND allocation with something like 75/25 VTI/EDV or 90/10 VTI/TMF.
That way you get higher returns from the VTI with a similar treasury insurance policy, and maintain simplicity.
75/25 VTI/EDV pairs stocks with long duration treasurys, but not leveraged  right?
/FIRE55
I personally am using a risk budget approach to balance leveraged funds in my Roth account, which allocates a fraction of risk to equities and the remainder to treasuries. My tuning knob is the fraction of risk assigned to equities. I find that 2/3 to 3/4 of risk to equities would have given a fairly smooth ride since the 1980s, with the individual allocations changing by a factor of two or so over time. For many of the combinations I've tried, I generally have calculated positive moving 2year CAGRs throughout the last 25 to 35 years. Perhaps the synthetic data is optimistic.
A second knob that some people use is based on limiting volatility. That would raise and lower the leverage in order to maintain a fixed level of portfolio volatility that you are comfortable with, while keeping the risk allocations constant between equities and treasuries. In essence you would calculate your risk allocation between equities and treasuries, then adjust the UPRO and EDV/TMF fractions up and down based on recent volatility to keep the overall volatility moreorless fixed. Or take some of the portfolio to cash. This will tend to smoothly reduce leverage when volatility rises.
A bit complex, but once programmed the math works out the allocations in a dispassionate way for both alternatives and I don't have to do much thinking. This is not buyandhold the way I do it, although you could get historical data to estimate an average set of allocations. I haven't played with the second approach much, to be honest, because it doesn't do as well IMO.
Hope this helps.
Re: HEDGEFUNDIE's excellent adventure Part II: The next journey
The monthly vs daily difference may be luck in rebalancing timing. A contrasting example is DXSLX (monthly S&P 500) vs SSO (daily S&P 500). The former trails by 3.4% CAGR over the 13years they've both been around, mostly because the 2008 crash had a lot of months with many red days which favored daily rebalancing to reduce exposure each time. At the bottom, DXSLX was almost down to half of SSO. (In contrast, March 2020 had almost as many green days as red days, so daily rebalancing didn't give much of an edge.)Bearbug wrote: ↑Tue Jun 08, 2021 3:50 pm Taking a look at the monthly reset leveraged mutual funds. Direxion's 2x monthly products were originally 2.5x and were changed to 2x during the depths of the financial crisis, explaining part of the underperformance. Aside from that, they perform very closely with the daily reset products, which isn't exactly how it should be. Rydex's RMQHX has outperformed DXQLX. Over a 2 percent higher CAGR. Not bad at all. RMQHX has smoked it's daily leveraged counterpart QLD despite the higher ER. Around 3.5 percent higher CAGR since 2015.
I wish Rydex would introduce monthly reset S&P 500/600/400 index funds.
Re: HEDGEFUNDIE's excellent adventure Part II: The next journey
I was under the impression that the larger 2008 drawdown was due to DXSLX being 2.5x leveraged at that time and then being switched to 2x at/near the bottom, causing the greater losses to be locked in.Semantics wrote: ↑Tue Jun 08, 2021 11:43 pmThe monthly vs daily difference may be luck in rebalancing timing. A contrasting example is DXSLX (monthly S&P 500) vs SSO (daily S&P 500). The former trails by 3.4% CAGR over the 13years they've both been around, mostly because the 2008 crash had a lot of months with many red days which favored daily rebalancing to reduce exposure each time. At the bottom, DXSLX was almost down to half of SSO. (In contrast, March 2020 had almost as many green days as red days, so daily rebalancing didn't give much of an edge.)Bearbug wrote: ↑Tue Jun 08, 2021 3:50 pm Taking a look at the monthly reset leveraged mutual funds. Direxion's 2x monthly products were originally 2.5x and were changed to 2x during the depths of the financial crisis, explaining part of the underperformance. Aside from that, they perform very closely with the daily reset products, which isn't exactly how it should be. Rydex's RMQHX has outperformed DXQLX. Over a 2 percent higher CAGR. Not bad at all. RMQHX has smoked it's daily leveraged counterpart QLD despite the higher ER. Around 3.5 percent higher CAGR since 2015.
I wish Rydex would introduce monthly reset S&P 500/600/400 index funds.
Re: HEDGEFUNDIE's excellent adventure Part II: The next journey
You're right, my bad, I misread and thought only DXQLX was formerly 2.5x. It should be possible to compute based on daily/monthly returns and volatility which has historically been the better rebalancing frequency, but I would be surprised if one turned out to be better in the long run.Bearbug wrote: ↑Wed Jun 09, 2021 12:08 amI was under the impression that the larger 2008 drawdown was due to DXSLX being 2.5x leveraged at that time and then being switched to 2x at/near the bottom, causing the greater losses to be locked in.Semantics wrote: ↑Tue Jun 08, 2021 11:43 pmThe monthly vs daily difference may be luck in rebalancing timing. A contrasting example is DXSLX (monthly S&P 500) vs SSO (daily S&P 500). The former trails by 3.4% CAGR over the 13years they've both been around, mostly because the 2008 crash had a lot of months with many red days which favored daily rebalancing to reduce exposure each time. At the bottom, DXSLX was almost down to half of SSO. (In contrast, March 2020 had almost as many green days as red days, so daily rebalancing didn't give much of an edge.)Bearbug wrote: ↑Tue Jun 08, 2021 3:50 pm Taking a look at the monthly reset leveraged mutual funds. Direxion's 2x monthly products were originally 2.5x and were changed to 2x during the depths of the financial crisis, explaining part of the underperformance. Aside from that, they perform very closely with the daily reset products, which isn't exactly how it should be. Rydex's RMQHX has outperformed DXQLX. Over a 2 percent higher CAGR. Not bad at all. RMQHX has smoked it's daily leveraged counterpart QLD despite the higher ER. Around 3.5 percent higher CAGR since 2015.
I wish Rydex would introduce monthly reset S&P 500/600/400 index funds.
Re: HEDGEFUNDIE's excellent adventure Part II: The next journey
As somewhat of an aside, it occurs to me that drawdowns are sometimes a bit misleading when running levered funds. This popped up as I was assessing some backtests experiencing 50% drops in 1987 and 1990, but with portfolios that were running roughly 40% afterwards.
Consider a 50% drawdown.
If the portfolio is returning 25% on average, a 50% drawdown is basically giving back 3.1 years of growth. At the 40% rate, that's just 2 years.
That's a far different thing than if the portfolio is returning 10% on average; a 50% drawdown means giving back 7.3 years. Or if returns are 6% on average, you're giving back 11.9 years.
It seems to me that to compare apples to apples, you'd compare the same giveback duration.
So if you are giving back 5 years, this would correspond to 25%, 38%, 67%, and 87% drawdowns for CAGRs of 6, 10, 25, and 50%, respectively.
Make sense? Thoughts?
Consider a 50% drawdown.
If the portfolio is returning 25% on average, a 50% drawdown is basically giving back 3.1 years of growth. At the 40% rate, that's just 2 years.
That's a far different thing than if the portfolio is returning 10% on average; a 50% drawdown means giving back 7.3 years. Or if returns are 6% on average, you're giving back 11.9 years.
It seems to me that to compare apples to apples, you'd compare the same giveback duration.
So if you are giving back 5 years, this would correspond to 25%, 38%, 67%, and 87% drawdowns for CAGRs of 6, 10, 25, and 50%, respectively.
Make sense? Thoughts?

 Posts: 306
 Joined: Sun Feb 04, 2018 7:44 pm
Re: HEDGEFUNDIE's excellent adventure Part II: The next journey
I don't think so. If you are comparing the same giveback duration then you are in effect saying the risk is always equal, which it is not. I think what you might contemplate instead is that the drawdown sizes are different instead which is more accurate.Hydromod wrote: ↑Sat Jun 12, 2021 4:33 pm As somewhat of an aside, it occurs to me that drawdowns are sometimes a bit misleading when running levered funds. This popped up as I was assessing some backtests experiencing 50% drops in 1987 and 1990, but with portfolios that were running roughly 40% afterwards.
Consider a 50% drawdown.
If the portfolio is returning 25% on average, a 50% drawdown is basically giving back 3.1 years of growth. At the 40% rate, that's just 2 years.
That's a far different thing than if the portfolio is returning 10% on average; a 50% drawdown means giving back 7.3 years. Or if returns are 6% on average, you're giving back 11.9 years.
It seems to me that to compare apples to apples, you'd compare the same giveback duration.
So if you are giving back 5 years, this would correspond to 25%, 38%, 67%, and 87% drawdowns for CAGRs of 6, 10, 25, and 50%, respectively.
Make sense? Thoughts?
Re: HEDGEFUNDIE's excellent adventure Part II: The next journey
I'm afraid I miss your point.perfectuncertainty wrote: ↑Sun Jun 13, 2021 7:26 am I don't think so. If you are comparing the same giveback duration then you are in effect saying the risk is always equal, which it is not. I think what you might contemplate instead is that the drawdown sizes are different instead which is more accurate.
Assume that a portfolio has a trailing volatility measured on 60 trading days, has a value of $10 at a starting time, rises to a peak after two years, and returns to $10 after four years. This gives a giveback duration of four years. It doesn't matter what the 60day trailing volatility does, nor does it matter what the peak was (although the peak determines drawdown).
I was a little unclear in what motivated the thought. I think of it as a way of framing that may help with emotional control during unusual events.
My thinking is as follows.
A common (and desired) effect of the UPRO/TMF dynamics is that TMF rises after UPRO starts to crash (flight to safety). I noticed that in some cases, TMF spikes after a crash enough to more than compensate for the UPRO loss, then returns to normal. Strictly speaking, the portfolio drawdown from the crash is measured from the peak of the TMF spike, which sets a higher bar for recovery.
More generally, a crash tends to follow a bubble. A bubble implies that gains are irrational and excessive (illusory); a more rational market would not have provided the same gains. 2000 and 2008, for example. Again, the drawdown measure doesn't take the illusions of the bubble gains into account.
The giveback duration idea mitigates the effect of spikes and bubbles, which temporarily give gains only to pull them away again. In these cases, it may be more fair to say "I'm back where I was X months ago" than "I lost Y percent from the alltime high". A drop isn't bad in the grand scheme if the portfolio is back where it was 3 months ago, even if the portfolio halved after a spike.
The financebased definition of risk is in terms of volatility, or the standard deviation of a time series of portfolio values. Normally this is measured over some trailing duration (e.g., the last 60 trading days).
The idea of the giveback duration measures a similar concept, except that it's more like tracing a given return in time. I suppose the giveback duration definition would be "the longest trailing duration with 0 return."
My physicsbased interpretation is that this is a little like the difference between Eulerian and Lagrangian measures of flow velocity. In Eulerian measures, you sit at a point in space and track the changes in velocity at the point. In Lagrangian measures, you follow a particle in the flow and track how rapidly it changes position over time. Both measures are valid and useful.
In essence, volatility is getting at how much the portfolio fluctuates and giveback duration is getting at how long it takes to lock in a gain.
Re: HEDGEFUNDIE's excellent adventure Part II: The next journey
Ok, I get what you said now.
Yes, in effect I'm saying that two portfolios have equivalent risks, if the giveback duration is equal, when expressed in the metric of fraction of the portfolio gains that are lost per unit time. Here the metric is how much of my lifetime has been lost retracing previous gains, rather than how much of my money has been lost from its last peak.
The two risks are complementary.
If two portfolios have the same giveback duration, the same amount of invested money has resulted in the same gain (zero) over the same period of time. So at that moment, a rational investor would be indifferent to which portfolio was selected. If they could have been substituted without change, they had the same risk over the interval.
I'd say that a given drawdown expresses a much different risk, expressed in terms of my lifetime lost, if it represents 3 months of gains versus 2 years of gains.
I guess this is an argument for using the largest number of years lost during a backtest as a summary complement to maximum drawdown.
Although perhaps the ulcer index tells the story at least as well.
Yes, in effect I'm saying that two portfolios have equivalent risks, if the giveback duration is equal, when expressed in the metric of fraction of the portfolio gains that are lost per unit time. Here the metric is how much of my lifetime has been lost retracing previous gains, rather than how much of my money has been lost from its last peak.
The two risks are complementary.
If two portfolios have the same giveback duration, the same amount of invested money has resulted in the same gain (zero) over the same period of time. So at that moment, a rational investor would be indifferent to which portfolio was selected. If they could have been substituted without change, they had the same risk over the interval.
I'd say that a given drawdown expresses a much different risk, expressed in terms of my lifetime lost, if it represents 3 months of gains versus 2 years of gains.
I guess this is an argument for using the largest number of years lost during a backtest as a summary complement to maximum drawdown.
Although perhaps the ulcer index tells the story at least as well.
 RovenSkyfall
 Posts: 266
 Joined: Wed Apr 01, 2020 11:40 am
Re: HEDGEFUNDIE's excellent adventure Part II: The next journey
Interesting. There are some complexities which you may have overlooked. To have this be a helpful metric you would need to incorporate all drops (not just the largest drops) and recoveries, and note the frequency of each. If the one largest drop had a fast recovery then LETF will likely look superior on your metric, but a long drawdown and slow recovery might favor the benchmark. Both are important to the risk/return weighting of using LETFs, and only looking at one may mislead you on the risk of a particular fund.Hydromod wrote: ↑Sun Jun 13, 2021 9:38 pm Ok, I get what you said now.
Yes, in effect I'm saying that two portfolios have equivalent risks, if the giveback duration is equal, when expressed in the metric of fraction of the portfolio gains that are lost per unit time. Here the metric is how much of my lifetime has been lost retracing previous gains, rather than how much of my money has been lost from its last peak.
The two risks are complementary.
If two portfolios have the same giveback duration, the same amount of invested money has resulted in the same gain (zero) over the same period of time. So at that moment, a rational investor would be indifferent to which portfolio was selected. If they could have been substituted without change, they had the same risk over the interval.
I'd say that a given drawdown expresses a much different risk, expressed in terms of my lifetime lost, if it represents 3 months of gains versus 2 years of gains.
I guess this is an argument for using the largest number of years lost during a backtest as a summary complement to maximum drawdown.
Although perhaps the ulcer index tells the story at least as well.
Because of the daily resetting, the drops and recovery differ between LETFs and its benchmark depending on the timeline of the drop and subsequent recovery. Since each crash and recovery has its own unique timeline, each crash/recovery cycle may have different relationships, making it a difficult to interpret parameter (i.e. 2008 A is better than B, 2000 A is equal to B and 2020 B is better than A).
CAGR is much simpler and incorporates the recovery into the return. Admittedly it is pathway dependent (but this metric is also dependent on the timeline chosen).
I saved my money, but it can't save me  The Chariot
Re: HEDGEFUNDIE's excellent adventure Part II: The next journey
Actually, I have been leaning on trailing 1yr, 2yr, and 3yr CAGR as a roughandready criterion for assessing portfolio composition and rebalancing strategies using 3x LETFs. I agree moving CAGR is a really good type of metric.RovenSkyfall wrote: ↑Mon Jun 14, 2021 1:17 pm Interesting. There are some complexities which you may have overlooked. To have this be a helpful metric you would need to incorporate all drops (not just the largest drops) and recoveries, and note the frequency of each. If the one largest drop had a fast recovery then LETF will likely look superior on your metric, but a long drawdown and slow recovery might favor the benchmark. Both are important to the risk/return weighting of using LETFs, and only looking at one may mislead you on the risk of a particular fund.
Because of the daily resetting, the drops and recovery differ between LETFs and its benchmark depending on the timeline of the drop and subsequent recovery. Since each crash and recovery has its own unique timeline, each crash/recovery cycle may have different relationships, making it a difficult to interpret parameter (i.e. 2008 A is better than B, 2000 A is equal to B and 2020 B is better than A).
CAGR is much simpler and incorporates the recovery into the return. Admittedly it is pathway dependent (but this metric is also dependent on the timeline chosen).
My thought was that it would be a good design target to have a portfolio that generally recovers within a year, two max, which gives a criterion of moving 1yr CAGR for the entire backtest.
Some candidate 3x LETF portfolios with trailing 1yr CAGR > 0 throughout suffered drawdowns of 50%, but these were just blips because they recovered so quickly.
The contrast between substantial (but temporary) drawdown and smooth trendline led to the giveback duration idea as a complement to drawdown.
Re: HEDGEFUNDIE's excellent adventure Part II: The next journey
This research paper might be of interest to you. All that Glitters Is Not Gold: Comparing Backtest and OutofSample Performance on a Large Cohort of Trading Algorithms. It seems to support volatility and maximum drawdown to have better predictive value than Sharpe Ratio. IMO, the current challenge in assessing performance is the sustained bull market that has overshadowed any signal that exhibit variations.
Moving CAGR sounds similar to holding period returns that I use as one of the evaluation criteria. It is basically how a portfolio would have performed if started in different years and held for different number of years.
For example, a 60/40 UPRO/TMF quarterly rebalanced backtest gives lowest annualized return of 27.4% if started in Jan 2015 and balanced quarterly till now versus highest 53.5% if started in Jan 2019. For 1 year holding period, range is from 14.3% (2018) to +83.5% (2019), for 2 year holding period from +8.0% (2015 +2) to +79.7% (2019 +2), for 3 year holding period from +16.2% (2016 +3) to +53.5% (2019 +3).
I find the data about Worst drawdown periods and stress events to be more indicative of effectiveness of strategy. For the above portfolio, for 5 worst drawdown periods, the Net drawdown ranged from 18.32% to 45.35% with recovery duration ranging from 50 to 300 days. LETFs tend to handle recovery from unknown/sudden events much better than from lingering events.
Some of my preferred evaluation criteria are Holding Period Return table, Performance during stress events, Distribution of monthly returns, underwater plots, top 5 drawdown periods, and comparison between In sample and out of sample annual volatility, and max drawdown.
Moving CAGR sounds similar to holding period returns that I use as one of the evaluation criteria. It is basically how a portfolio would have performed if started in different years and held for different number of years.
For example, a 60/40 UPRO/TMF quarterly rebalanced backtest gives lowest annualized return of 27.4% if started in Jan 2015 and balanced quarterly till now versus highest 53.5% if started in Jan 2019. For 1 year holding period, range is from 14.3% (2018) to +83.5% (2019), for 2 year holding period from +8.0% (2015 +2) to +79.7% (2019 +2), for 3 year holding period from +16.2% (2016 +3) to +53.5% (2019 +3).
I find the data about Worst drawdown periods and stress events to be more indicative of effectiveness of strategy. For the above portfolio, for 5 worst drawdown periods, the Net drawdown ranged from 18.32% to 45.35% with recovery duration ranging from 50 to 300 days. LETFs tend to handle recovery from unknown/sudden events much better than from lingering events.
Some of my preferred evaluation criteria are Holding Period Return table, Performance during stress events, Distribution of monthly returns, underwater plots, top 5 drawdown periods, and comparison between In sample and out of sample annual volatility, and max drawdown.
Hydromod wrote: ↑Mon Jun 14, 2021 2:09 pm Actually, I have been leaning on trailing 1yr, 2yr, and 3yr CAGR as a roughandready criterion for assessing portfolio composition and rebalancing strategies using 3x LETFs. I agree moving CAGR is a really good type of metric.
My thought was that it would be a good design target to have a portfolio that generally recovers within a year, two max, which gives a criterion of moving 1yr CAGR for the entire backtest.
Some candidate 3x LETF portfolios with trailing 1yr CAGR > 0 throughout suffered drawdowns of 50%, but these were just blips because they recovered so quickly.
The contrast between substantial (but temporary) drawdown and smooth trendline led to the giveback duration idea as a complement to drawdown.
Re: HEDGEFUNDIE's excellent adventure Part II: The next journey
I am converging toward just thinking about CAGR of the benchmark, under the assumption that over long periods of time the volatility will not be significantly different. If volatility and leverage ratio are fixed, then the excess CAGR due to leverage depends *only* on the unleveraged CAGR (a mathematical consequence of Ito's Lemma). If the underlying CAGR is really poor, the leverage will underperform. Unfortunately, expected future returns are currently low  so it is extremely unlikely that HFEA will perform the way it has the past decade going forward.RovenSkyfall wrote: ↑Mon Jun 14, 2021 1:17 pmInteresting. There are some complexities which you may have overlooked. To have this be a helpful metric you would need to incorporate all drops (not just the largest drops) and recoveries, and note the frequency of each. If the one largest drop had a fast recovery then LETF will likely look superior on your metric, but a long drawdown and slow recovery might favor the benchmark. Both are important to the risk/return weighting of using LETFs, and only looking at one may mislead you on the risk of a particular fund.Hydromod wrote: ↑Sun Jun 13, 2021 9:38 pm Ok, I get what you said now.
Yes, in effect I'm saying that two portfolios have equivalent risks, if the giveback duration is equal, when expressed in the metric of fraction of the portfolio gains that are lost per unit time. Here the metric is how much of my lifetime has been lost retracing previous gains, rather than how much of my money has been lost from its last peak.
The two risks are complementary.
If two portfolios have the same giveback duration, the same amount of invested money has resulted in the same gain (zero) over the same period of time. So at that moment, a rational investor would be indifferent to which portfolio was selected. If they could have been substituted without change, they had the same risk over the interval.
I'd say that a given drawdown expresses a much different risk, expressed in terms of my lifetime lost, if it represents 3 months of gains versus 2 years of gains.
I guess this is an argument for using the largest number of years lost during a backtest as a summary complement to maximum drawdown.
Although perhaps the ulcer index tells the story at least as well.
Because of the daily resetting, the drops and recovery differ between LETFs and its benchmark depending on the timeline of the drop and subsequent recovery. Since each crash and recovery has its own unique timeline, each crash/recovery cycle may have different relationships, making it a difficult to interpret parameter (i.e. 2008 A is better than B, 2000 A is equal to B and 2020 B is better than A).
CAGR is much simpler and incorporates the recovery into the return. Admittedly it is pathway dependent (but this metric is also dependent on the timeline chosen).
On the plus side, since it functions more like 1.65x leverage, there's not that much volatility drag in HFEA and it has a good chance to still outperform unleveraged even with low returns from the underlyings. My estimate is 510% CAGR over the next decade, based on the fact that an international version of HFEA would have returned 10.75% during the period 20122018 where long term treasuries were flat and VGTSX only returned 5.13%. Still, not too bad if VOO only returns half that. But that said, factor tilting and international continue to seem appealing.
I guess my overarching point being I like the idea of not looking at any backtests, past drawdowns, or other singular data points, and just model out the expected returns mathematically using various conservative assumptions.
 RovenSkyfall
 Posts: 266
 Joined: Wed Apr 01, 2020 11:40 am
Re: HEDGEFUNDIE's excellent adventure Part II: The next journey
I think looking at expected returns makes a lot of sense.Semantics wrote: ↑Mon Jun 14, 2021 9:51 pm
I am converging toward just thinking about CAGR of the benchmark, under the assumption that over long periods of time the volatility will not be significantly different. If volatility and leverage ratio are fixed, then the excess CAGR due to leverage depends *only* on the unleveraged CAGR (a mathematical consequence of Ito's Lemma). If the underlying CAGR is really poor, the leverage will underperform. Unfortunately, expected future returns are currently low  so it is extremely unlikely that HFEA will perform the way it has the past decade going forward.
On the plus side, since it functions more like 1.65x leverage, there's not that much volatility drag in HFEA and it has a good chance to still outperform unleveraged even with low returns from the underlyings. My estimate is 510% CAGR over the next decade, based on the fact that an international version of HFEA would have returned 10.75% during the period 20122018 where long term treasuries were flat and VGTSX only returned 5.13%. Still, not too bad if VOO only returns half that. But that said, factor tilting and international continue to seem appealing.
I guess my overarching point being I like the idea of not looking at any backtests, past drawdowns, or other singular data points, and just model out the expected returns mathematically using various conservative assumptions.
One thing that comes to mind, however, is that the volatility is also important to consider. I think TNA is a good example of why this is important. That would mean expected returns and expected future volatility should be considered.
The other point worth considering is that a lot of the recent performance of the SP500 has been due to increasing valuations. How this has influenced the returns and future expected returns seems less predictable. This would lead me to feel more confident in a LETF of a SCV index for future expected returns (despite the volatility) compared to UPRO. This post offers some good thoughts on the matter.
I saved my money, but it can't save me  The Chariot

 Posts: 1857
 Joined: Tue Mar 05, 2019 10:29 pm
 Location: Colorado
Re: HEDGEFUNDIE's excellent adventure Part II: The next journey
Am I the only one left on the UPRO/EDV train ever since MotoTrojan jumped off? I've been invested in that strategy for 22 months now and I'm up 87%, compared to total return for the S&P500 of 43% over the same time period.
Next rebal is July 1.
Next rebal is July 1.
Re: HEDGEFUNDIE's excellent adventure Part II: The next journey
This discussion of the right way to assess risk for leveraged portfolios made me sit down and simply work out what all the optimal leveraged portfolios look like.
I make the usual simplifying assumptions that you have an underlying asset with arithmetic return μ and volatility σ (following a lognormal distribution). The geometric return is then μ  σ^2/2. Leveraging up by leverage ratio L gives an asset with arithmetic return Lμ and volatility Lσ. The geometric return of the leveraged asset is Lμ  (Lσ)^2/2. It's easy to see that the leverage ratio that maximizes geometric return (i.e. CAGR) is L = μ/σ^2.
Plugging this in, you get that the optimal leveraged portfolio has arithmetic return μ^2/σ^2 and volatility μ/σ. Its geometric return is 0.5(μ/σ)^2. Notice that in these formulas, μ and σ always appear together as s = μ/σ. In other words, given an underlying asset with arithmetic return μ and volatility σ, the optimal leveraged portfolio that you can derive from it only depends on the Sharpe ratio s = μ/σ.
If the underlying asset is "weak" meaning it has a low Sharpe ratio s, the optimal leveraged version of it will have low volatility s and a low geometric return 0.5*s^2. Furthermore, the ratio between the geometric return and volatility is 0.5*s, which will be low. If you think about it, this means that the portfolio takes a long time to recover from drawdowns (hopefully shedding some light on the question about "duration risk" mentioned above).
Conversely, if the underlying asset is "strong" meaning it has a high Sharpe, the optimal leveraged version of it will have high volatility and a high geometric return. But despite the high volatility, it will recover quickly from drawdowns.
I make the usual simplifying assumptions that you have an underlying asset with arithmetic return μ and volatility σ (following a lognormal distribution). The geometric return is then μ  σ^2/2. Leveraging up by leverage ratio L gives an asset with arithmetic return Lμ and volatility Lσ. The geometric return of the leveraged asset is Lμ  (Lσ)^2/2. It's easy to see that the leverage ratio that maximizes geometric return (i.e. CAGR) is L = μ/σ^2.
Plugging this in, you get that the optimal leveraged portfolio has arithmetic return μ^2/σ^2 and volatility μ/σ. Its geometric return is 0.5(μ/σ)^2. Notice that in these formulas, μ and σ always appear together as s = μ/σ. In other words, given an underlying asset with arithmetic return μ and volatility σ, the optimal leveraged portfolio that you can derive from it only depends on the Sharpe ratio s = μ/σ.
If the underlying asset is "weak" meaning it has a low Sharpe ratio s, the optimal leveraged version of it will have low volatility s and a low geometric return 0.5*s^2. Furthermore, the ratio between the geometric return and volatility is 0.5*s, which will be low. If you think about it, this means that the portfolio takes a long time to recover from drawdowns (hopefully shedding some light on the question about "duration risk" mentioned above).
Conversely, if the underlying asset is "strong" meaning it has a high Sharpe, the optimal leveraged version of it will have high volatility and a high geometric return. But despite the high volatility, it will recover quickly from drawdowns.
Re: HEDGEFUNDIE's excellent adventure Part II: The next journey
I'm trying to understand this a bit better, along with the implications.langlands wrote: ↑Tue Jun 15, 2021 10:30 am This discussion of the right way to assess risk for leveraged portfolios made me sit down and simply work out what all the optimal leveraged portfolios look like.
I make the usual simplifying assumptions that you have an underlying asset with arithmetic return μ and volatility σ (following a lognormal distribution). The geometric return is then μ  σ^2/2. Leveraging up by leverage ratio L gives an asset with arithmetic return Lμ and volatility Lσ. The geometric return of the leveraged asset is Lμ  (Lσ)^2/2. It's easy to see that the leverage ratio that maximizes geometric return (i.e. CAGR) is L = μ/σ^2.
Plugging this in, you get that the optimal leveraged portfolio has arithmetic return μ^2/σ^2 and volatility μ/σ. Its geometric return is 0.5(μ/σ)^2. Notice that in these formulas, μ and σ always appear together as s = μ/σ. In other words, given an underlying asset with arithmetic return μ and volatility σ, the optimal leveraged portfolio that you can derive from it only depends on the Sharpe ratio s = μ/σ.
If the underlying asset is "weak" meaning it has a low Sharpe ratio s, the optimal leveraged version of it will have low volatility s and a low geometric return 0.5*s^2. Furthermore, the ratio between the geometric return and volatility is 0.5*s, which will be low. If you think about it, this means that the portfolio takes a long time to recover from drawdowns (hopefully shedding some light on the question about "duration risk" mentioned above).
Conversely, if the underlying asset is "strong" meaning it has a high Sharpe, the optimal leveraged version of it will have high volatility and a high geometric return. But despite the high volatility, it will recover quickly from drawdowns.
I'm running a sample portfolio from 1994 to present with 4 assets, 3 equities (IWM, QQQ, RWR, all positively correlated with each other) plus TLT. This period has the crashes in 2000, 2008, and 2020, but not 1987 or 1990. I use a rebalancing scheme that minimizes variance and accounts for correlations.
The scheme gives me a CAGR of 0.11, and annualized volatility sigma of 0.093. This gives me an arithmetic return of mu = 0.114, for a Sharpe ratio of 0.114/0.093 = 1.22. I get an optimal leverage of L = 0.114/0.093^2 = 13.1.
I ran the same scheme with L = 3 on each asset (using URTY/TQQQ/DRN/TMF and extending the 1x fund to create a synthetic 3x fund where needed). I expect that the extension doesn't account for all borrowing and trading costs, so it is somewhat optimistic.
I calculate a CAGR of 0.52, sigma = 0.29, mu = 0.57, Sharpe = 1.97, and optimal leverage of 6.8. The more recent returns, without synthetic ETFs, has the 3x portfolio producing about 3x the CAGR of the 1x portfolio.
So (i) I may be calculating stuff wrong, (ii) there may be something about calculating the synthetic 3x LETFs that was quite optimistic (CAGR for the 3x LETF portfolios drops by about a factor of two after 2010, when the real LETFs began), or (iii) there's something about the leverage implementation or anticorrelated volatility that amplified returns, or (iii) there's something about fat tails that amplifies returns.
As a test, just running VFINX from 1980present gives me CAGR = 0.108, sigma = 0.18, mu = 0.12, Sharpe = 0.69, and L = 3.8. This L is about twice what others seem to think appropriate for the S&P 500. The calculated VFINX ulcer index is 24 times larger than the 1x portfolio and about twice the 3x portfolio.
I don't think I quite understand the L calculation yet, but I'm interpreting the L=13 for the 1x portfolio as suggesting that the scheme should be very robust using 3x LETFs, even if the calculation misrepresents L by a factor of two. Does that sound reasonable?
Re: HEDGEFUNDIE's excellent adventure Part II: The next journey
Hi Hydromod,Hydromod wrote: ↑Tue Jun 15, 2021 1:25 pm I'm trying to understand this a bit better, along with the implications.
I'm running a sample portfolio from 1994 to present with 4 assets, 3 equities (IWM, QQQ, RWR, all positively correlated with each other) plus TLT. This period has the crashes in 2000, 2008, and 2020, but not 1987 or 1990. I use a rebalancing scheme that minimizes variance and accounts for correlations.
The scheme gives me a CAGR of 0.11, and annualized volatility sigma of 0.093. This gives me an arithmetic return of mu = 0.114, for a Sharpe ratio of 0.114/0.093 = 1.22. I get an optimal leverage of L = 0.114/0.093^2 = 13.1.
I ran the same scheme with L = 3 on each asset (using URTY/TQQQ/DRN/TMF and extending the 1x fund to create a synthetic 3x fund where needed). I expect that the extension doesn't account for all borrowing and trading costs, so it is somewhat optimistic.
I calculate a CAGR of 0.52, sigma = 0.29, mu = 0.57, Sharpe = 1.97, and optimal leverage of 6.8. The more recent returns, without synthetic ETFs, has the 3x portfolio producing about 3x the CAGR of the 1x portfolio.
So (i) I may be calculating stuff wrong, (ii) there may be something about calculating the synthetic 3x LETFs that was quite optimistic (CAGR for the 3x LETF portfolios drops by about a factor of two after 2010, when the real LETFs began), or (iii) there's something about the leverage implementation or anticorrelated volatility that amplified returns, or (iii) there's something about fat tails that amplifies returns.
As a test, just running VFINX from 1980present gives me CAGR = 0.108, sigma = 0.18, mu = 0.12, Sharpe = 0.69, and L = 3.8. This L is about twice what others seem to think appropriate for the S&P 500. The calculated VFINX ulcer index is 24 times larger than the 1x portfolio and about twice the 3x portfolio.
I don't think I quite understand the L calculation yet, but I'm interpreting the L=13 for the 1x portfolio as suggesting that the scheme should be very robust using 3x LETFs, even if the calculation misrepresents L by a factor of two. Does that sound reasonable?
So the assumptions of my calculation above are that you are levering up a fixed portfolio with a constant arithmetic return and a constant volatility for all time where every day the return outcome is normally distributed and every day is independent of the next. In reality, none of these hold unfortunately.
If your portfolio is made up of say 55% UPRO/45% TMF, the assumption then is that that ratio is kept fixed every day with no drift in percentages, i.e. daily rebalancing. If you're not rebalancing everyday, then you don't have a fixed portfolio and the analysis gets more complicated.
Regarding the discrepancy between levering up the portfolio of 1x assets vs. the portfolio of 3x assets, I suspect it's most likely because your rebalancing mechanism doesn't behave identically on the two portfolios. As I said above, if you're not rebalancing daily, there will likely be a drift in composition between the two portfolios and so it's not surprising to see quite different outcomes. For example, how heavily you're weighted towards equities right before and during a crash will have a massive effect. Also, during placid bull markets like from 20092019, a slowly rebalanced portfolio of 3x assets will likely be more slanted towards equities than a portfolio of 1x assets. For instance, a 55% SPY/45% TLT might move to 60% SPY/40% TLT before you rebalance while a 55% UPRO/45% TMF might move to 70% UPRO/30% TMF.
The main other thing to keep in mind is that of varying volatility and fat tails. Return distribution is not normal and volatility is not constant. In reality, the L = 13 value you computed above is no where near optimal because of this. Both fat tails and the inability to know the actual volatility should lead you to lower leverage. (There's an analogous concept in the betting literature known as fractional Kelly, where instead of betting the full amount given by the Kelly criterion, one only bets say half of that because of errors in estimation). But yes, the fact that under ideal conditions the optimal leverage is so high does give me some confidence that even under realistic scenarios, L = 3 shouldn't be too bad. But this is of course assuming the future continues looking like the past. Questions of inflation and how that might affect the correlations/returns of the assets have been discussed ad nauseam.
A general thing regarding Sharpe: a Sharpe of 2 or above is considered ungodly good unless you're an elite hedge fund. A rule of thumb says that the S&P itself has a Sharpe of about 0.5. A SPY/TLT portfolio could plausibly have a long run Sharpe of about 0.75. Any claims of a constructed portfolio having a Sharpe > 1 seem optimistic to me unless you think you have some secret sauce or something. It's very easy to see these high Sharpes in backtesting of course, but just be aware you're overfitting. Maybe a good rule of thumb is to just halve whatever backtested Sharpe you got.
Re: HEDGEFUNDIE's excellent adventure Part II: The next journey
My Sharpe for both VFINX alone and VFINX/VUSTX are about 40% high compared to 0.5 and 0.75, but not all that far off. It suggests scaling back a bit though. I'm not an elite hedge fund, although I did stay at a Holiday Inn Express once.langlands wrote: ↑Tue Jun 15, 2021 3:22 pm Hi Hydromod,
So the assumptions of my calculation above are that you are levering up a fixed portfolio with a constant arithmetic return and a constant volatility for all time where every day the return outcome is normally distributed and every day is independent of the next. In reality, none of these hold unfortunately.
If your portfolio is made up of say 55% UPRO/45% TMF, the assumption then is that that ratio is kept fixed every day with no drift in percentages, i.e. daily rebalancing. If you're not rebalancing everyday, then you don't have a fixed portfolio and the analysis gets more complicated.
Regarding the discrepancy between levering up the portfolio of 1x assets vs. the portfolio of 3x assets, I suspect it's most likely because your rebalancing mechanism doesn't behave identically on the two portfolios. As I said above, if you're not rebalancing daily, there will likely be a drift in composition between the two portfolios and so it's not surprising to see quite different outcomes. For example, how heavily you're weighted towards equities right before and during a crash will have a massive effect. Also, during placid bull markets like from 20092019, a slowly rebalanced portfolio of 3x assets will likely be more slanted towards equities than a portfolio of 1x assets. For instance, a 55% SPY/45% TLT might move to 60% SPY/40% TLT before you rebalance while a 55% UPRO/45% TMF might move to 70% UPRO/30% TMF.
The main other thing to keep in mind is that of varying volatility and fat tails. Return distribution is not normal and volatility is not constant. In reality, the L = 13 value you computed above is no where near optimal because of this. Both fat tails and the inability to know the actual volatility should lead you to lower leverage. (There's an analogous concept in the betting literature known as fractional Kelly, where instead of betting the full amount given by the Kelly criterion, one only bets say half of that because of errors in estimation). But yes, the fact that under ideal conditions the optimal leverage is so high does give me some confidence that even under realistic scenarios, L = 3 shouldn't be too bad. But this is of course assuming the future continues looking like the past. Questions of inflation and how that might affect the correlations/returns of the assets have been discussed ad nauseam.
A general thing regarding Sharpe: a Sharpe of 2 or above is considered ungodly good unless you're an elite hedge fund. A rule of thumb says that the S&P itself has a Sharpe of about 0.5. A SPY/TLT portfolio could plausibly have a long run Sharpe of about 0.75. Any claims of a constructed portfolio having a Sharpe > 1 seems optimistic to me unless you think you have some secret sauce or something. It's very easy to see these high Sharpes in backtesting of course, but just be aware you're overfitting. Maybe a good rule of thumb is to just halve whatever backtested Sharpe you got.
Forgot the risk free rate. That’ll cut back the Sharpe some. I should really plot the excess returns.
The portfolio is rebalanced every 10 days, so composition shouldn't slide so much compared to daily rebalancing. Ten days isn't frequent enough to get much volatility bonus from daily fluctuations, but does cut back on trade slippage a bit.
The 1x and 3x compositions appear to be quite similar after rebalancing and are rebalanced on exactly the same schedule, but I hear you loud and clear on rebalance timing issues.
I do take the diagnostics with an appropriately large grain of salt, because this is not at all a fixed asset allocation approach. It's a type of clustered minimum variance. I marvel at how smooth it keeps the portfolio trajectory even with the 3x funds, it doesn't seem to follow the highs or the lows.
The weights are set to keep a fixed risk budget of 0.75 risk to equities, 0.25 to bonds, with each asset in a category accounting for an equal fraction of the category's risk budget. That way I can change the number of assets without strongly changing the risk profile.
The only fitted parameters are related to allocation of risk between equities and bonds, lookback periods for volatility (42 days) and correlation (63 days), and rebalance frequency. The results don't seem all that sensitive to these parameters.
My VFINX/VUSTX Sharpe is 1.1 over 1986present. Over this period the fraction of the portfolio to VFINX ranged from 0.25 to 0.75, but probably averaged around 0.5. Far from a fixed allocation. This variability damps some as assets are added.
I have a very hard time trusting the synthetic returns, so I strongly discount the calculated CAGRs before 2010 in the 3x portfolio. I wouldn't dream of acting on a value for L of 13. It's more of confidencebuilding thing, especially when the 3x L is quite a bit larger than 1.
Thank you much for the advice. Very helpful.

 Posts: 2
 Joined: Tue Jun 15, 2021 9:54 pm
Re: HEDGEFUNDIE's excellent adventure Part II: The next journey
Here is what I'm thinking of employing:
Buying $100 of TQQQ every week day at closing. This dollar cost averaging of TQQQ perhaps some should be a good strategy correct?
I think the rebalancing is annoying, and I'd rather just throw more money at it to try and reduce some of the impact of volatility. How does DCAing leveraged etf sound to everyone?
Buying $100 of TQQQ every week day at closing. This dollar cost averaging of TQQQ perhaps some should be a good strategy correct?
I think the rebalancing is annoying, and I'd rather just throw more money at it to try and reduce some of the impact of volatility. How does DCAing leveraged etf sound to everyone?
Re: HEDGEFUNDIE's excellent adventure Part II: The next journey
I quickly backtested investing $100 everyday in TQQQ since 20100211. I don't think I had the stomach to handle the drawdown roller coaster of this strategy. I most probably would have bailed out from the strategy pretty quickly. Just imagine yourself the day of 20110819 staring at your portfolio down 84.65% ($100 becoming $15.35). What kind of emotional reaction will you have and the rash action will you take?
Code: Select all
Worst drawdown periods Net drawdown in % Peak date Valley date Recovery date Duration
0 84.65 20110726 20110819 20120201 137
1 74.13 20200219 20200320 20200710 103
2 63.31 20180829 20181224 20191104 309
3 59.53 20120919 20121115 20130430 160
4 57.34 20120402 20120601 20120906 114
halivingston wrote: ↑Tue Jun 15, 2021 9:57 pm Here is what I'm thinking of employing:
Buying $100 of TQQQ every week day at closing. This dollar cost averaging of TQQQ perhaps some should be a good strategy correct?
I think the rebalancing is annoying, and I'd rather just throw more money at it to try and reduce some of the impact of volatility. How does DCAing leveraged etf sound to everyone?
Re: HEDGEFUNDIE's excellent adventure Part II: The next journey
Sure you can DCA into TQQQ. Note there is an ongoing debate over whether it is actually better to 'lump sum' into a stock or DCA into it. But if you're not counterbalancing it with a negatively correlated asset, you're not following HFEA.halivingston wrote: ↑Tue Jun 15, 2021 9:57 pm Here is what I'm thinking of employing:
Buying $100 of TQQQ every week day at closing. This dollar cost averaging of TQQQ perhaps some should be a good strategy correct?
I think the rebalancing is annoying, and I'd rather just throw more money at it to try and reduce some of the impact of volatility. How does DCAing leveraged etf sound to everyone?
Buy an initial portfolio of TQQQ/TMF in your chosen % allocation (say 55:45). Spend your $100 every day on either TQQQ or TMF, depending on which will nudge the asset allocation back to your chosen allocation. If one day you're at 54:46 TQQQ:TMF  put your $100 into TQQQ. If next week you're at 56:44 TQQQ:TMF  put your $100 into TMF. This is the exact same as rebalancing, without the taxable sale. Note if the imbalance gets beyond your ability to rectify it at $100/day  you should do an actual rebalance at some point.
*That* is DCAing into HFEA.
Backtest, FYI. Of course CAGR is higher, but keep an eye on max drawdown, the drawdown tab, worst year, sharpe and sortino ratios. If you want to go allin on TQQQ, knock yourself out, but beware monster drawdowns.
/FIRE55
Last edited by FIRE55 on Thu Jun 17, 2021 12:49 am, edited 1 time in total.

 Posts: 75
 Joined: Thu Feb 25, 2021 12:43 pm
Re: HEDGEFUNDIE's excellent adventure Part II: The next journey
TQQQ Would have dropped 99.97% or so during 2000. I highly doubt anyone would have continued DCAing in to TQQQ knowing that inevitably in 20 years it will come roaring back. People are deluding themselves big time. Most likely, they would have chalked it up to a massive mistake, and moved on with their life.

 Posts: 4636
 Joined: Fri Nov 06, 2020 1:41 pm
Re: HEDGEFUNDIE's excellent adventure Part II: The next journey
Those embarking on this journey need to understand that the equities portion can get completely wiped. UPRO would have gone 97% in 2008, not as bad as 99.97% but most people would struggle to hold onto it.TheDoctor91 wrote: ↑Wed Jun 16, 2021 3:53 pm TQQQ Would have dropped 99.97% or so during 2000. I highly doubt anyone would have continued DCAing in to TQQQ knowing that inevitably in 20 years it will come roaring back. People are deluding themselves big time. Most likely, they would have chalked it up to a massive mistake, and moved on with their life.
Re: HEDGEFUNDIE's excellent adventure Part II: The next journey
Haven't we seen that rebalancing daily significantly reduces your CAGR though? We rebalance quarterly for a reason!FIRE55 wrote: ↑Wed Jun 16, 2021 11:44 am
Sure you can DCA into TQQQ. Note there is an ongoing debate over whether it is actually better to 'lump sum' into a stock or DCA into it. But if you're not counterbalancing it with a negatively correlated asset, you're not following HFEA.
Buy an initial portfolio of TQQQ/TMF in your chosen % allocation (say 55:45). Spend your $100 every day on either TQQQ or TMF, depending on which will nudge the asset allocation back to your chosen allocation. If one day you're at 59:41 TQQQ:TMF  put your $100 into TQQQ. If next week you're at 61:39 TQQQ:TMF  put your $100 into TMF. This is the exact same as rebalancing, without the taxable sale. Note if the imbalance gets beyond your ability to rectify it at $100/day  you should do an actual rebalance at some point.
*That* is DCAing into HFEA.
Backtest, FYI. Of course CAGR is higher, but keep an eye on max drawdown, the drawdown tab, worst year, sharpe and sortino ratios. If you want to go allin on TQQQ, knock yourself out, but beware monster drawdowns.
/FIRE55
Edit: Nevermind, you use quarterly rebalancing in your backtest but it's not clear in your description. If you're DCAing in, you need to maintain your current ratio until the rebalance would have occurred.
Re: HEDGEFUNDIE's excellent adventure Part II: The next journey
I actually neglected that, but yes I assumed quarterly because it's what I use! OP wanted to think about daily. They should indeed look into daily vs. quarterly rebalancing, perhaps realigning the balance using $6,000/quarter inbound. That sounds pretty f'in good to me, actually.Afrofreak wrote: ↑Wed Jun 16, 2021 4:17 pm
Haven't we seen that rebalancing daily significantly reduces your CAGR though? We rebalance quarterly for a reason!
Edit: Nevermind, you use quarterly rebalancing in your backtest but it's not clear in your description. If you're DCAing in, you need to maintain your current ratio until the rebalance would have occurred.
/FIRE55
Re: HEDGEFUNDIE's excellent adventure Part II: The next journey
Not to interrupt the thread, but I see threads from 2017 and 2011 on the Permanent Portfolio Forum that talk about 3x LETFs in the context of the permanent portfolio. I just ran across these today.
Starting 12/30/2017: PP Inspired Leveraged Portfolios
Starting 1/22/2011: 20% annual returns over 40 years...interested?
Lots of thinking along some of the same lines that have been hashed out in the HFEAinspired threads.
Starting 12/30/2017: PP Inspired Leveraged Portfolios
Starting 1/22/2011: 20% annual returns over 40 years...interested?
Lots of thinking along some of the same lines that have been hashed out in the HFEAinspired threads.