HEDGEFUNDIE's excellent adventure [risk parity strategy using 3x leveraged ETFs]

[physixfan]

With recent huge volatility, I am tempted to rebalance more frequently based on 20-day look back risk parity strategy. For example, monthly->weekly? Is there any study done about whether frequent rebalance is a bad thing?

[MindTheGAAP]

55% UPRO (risk parity)
80% UPRO (target 20% vol)


This sound correct?

56% right now for risk parity. 80% sounds about right for 20% volatility. Mine shows as 74% but it's due to weighting most recent days more than days > 10

Can you send me in the right direction -- or provide a quick crash course -- on how I would practically implement the 20% volatility effort? Trying to work out how I'm going to rebalance and right now, I'm set just to realign to the 60/40 split on a monthly basis. May need to tweak it be weekly though since yesterday shifted me from 60/40 to 68/32.

[chicagoan23]

I don't have any technical analysis or additional tweaks to the system but still wanted to share.....I had been waiting for a slight equities pullback to implement this strategy in my Roth IRA. I put 40% in UPRO on August 1, which is now down 11% in 3+ days(!!!) And yet with the 60% in TMF up 8.6% in that same brief time, I actually am still up overall.

Thank you HEDGEFUNDIE for your insight and creativity, and for sharing this with the group. Now I'm looking forward to that UPRO recovery.

[MoneyMarathon]

Is there any study done about whether frequent rebalance is a bad thing?

Monthly was okay. Quarterly is better when the market for the S&P 500 or for long-term treasuries has longer strings of days going up, or days going down. When things are going down for long runs, you're not rebalancing into the loser. When things are going up for long runs, you're letting the winner run. Nobody knows what the future holds, but to the extent that there is autocorrelation in the short term of 2-3 months (up follows up, down follows down), quarterly might be better. Monthly would be better if you expected 1 month performance to 'mean revert'.

[MotoTrojan]

Is there any study done about whether frequent rebalance is a bad thing?

Monthly was okay. Quarterly is better when the market for the S&P 500 or for long-term treasuries has longer strings of days going up, or days going down. When things are going down for long runs, you're not rebalancing into the loser. When things are going up for long runs, you're letting the winner run. Nobody knows what the future holds, but to the extent that there is autocorrelation in the short term of 2-3 months (up follows up, down follows down), quarterly might be better. Monthly would be better if you expected 1 month performance to 'mean revert'.

I think physixfan is looking at a volatility look-back method so you can't actually say quarterly is better (true for the passive 40/60 historically or in your noted situation). Transaction costs are another reason that frequent rebalancing would hurt. Also whip-saws. I personally wouldn't go much tighter than monthly.

[MotoTrojan]

55% UPRO (risk parity)
80% UPRO (target 20% vol)


This sound correct?

56% right now for risk parity. 80% sounds about right for 20% volatility. Mine shows as 74% but it's due to weighting most recent days more than days > 10

Can you send me in the right direction -- or provide a quick crash course -- on how I would practically implement the 20% volatility effort? Trying to work out how I'm going to rebalance and right now, I'm set just to realign to the 60/40 split on a monthly basis. May need to tweak it be weekly though since yesterday shifted me from 60/40 to 68/32.

60/40 UPRO/TMF monthly forever is a poor choice (OP is 40/60). I also wouldn't rebalance monthly unless resetting allocation for target volatility or risk parity w/ a 1 month look-back.

Funny to see how quick people are adopting these strategies without much understanding of them... if you are just in the mood for gambling and don't want to learn a bunch and improve your conviction, I would just stick with the OP's 40/60 UPRO/TMF quarterly.

[MoneyMarathon]

Is there any study done about whether frequent rebalance is a bad thing?

Monthly was okay. Quarterly is better when the market for the S&P 500 or for long-term treasuries has longer strings of days going up, or days going down. When things are going down for long runs, you're not rebalancing into the loser. When things are going up for long runs, you're letting the winner run. Nobody knows what the future holds, but to the extent that there is autocorrelation in the short term of 2-3 months (up follows up, down follows down), quarterly might be better. Monthly would be better if you expected 1 month performance to 'mean revert'.

I think physixfan is looking at a volatility look-back method so you can't actually say quarterly is better (true for the passive 40/60 historically or in your noted situation). Transaction costs are another reason that frequent rebalancing would hurt. Also whip-saws. I personally wouldn't go much tighter than monthly.

Oh, I see. I don't think the OP strategy is for gamblers more so than vol targeting. The OP strategy is more conservative and isn't reaching for a more perfect, higher performing idea that could blow up in a flash crash.

[MotoTrojan]

Is there any study done about whether frequent rebalance is a bad thing?

Monthly was okay. Quarterly is better when the market for the S&P 500 or for long-term treasuries has longer strings of days going up, or days going down. When things are going down for long runs, you're not rebalancing into the loser. When things are going up for long runs, you're letting the winner run. Nobody knows what the future holds, but to the extent that there is autocorrelation in the short term of 2-3 months (up follows up, down follows down), quarterly might be better. Monthly would be better if you expected 1 month performance to 'mean revert'.

I think physixfan is looking at a volatility look-back method so you can't actually say quarterly is better (true for the passive 40/60 historically or in your noted situation). Transaction costs are another reason that frequent rebalancing would hurt. Also whip-saws. I personally wouldn't go much tighter than monthly.

Oh, I see. I don't think the OP strategy is for gamblers more so than vol targeting. The OP strategy is more conservative and isn't reaching for a more perfect, higher performing idea that could blow up in a flash crash.

My point there was that a lot of the people adopting this now don't seem interested in learning the underlying principals but more so just want a get-rich-quick scheme and want to know the basic mechanics of implementation. If you don't understand the merits of vol targeting then I think you're less likely to stick with it, and thus I would suggest the more passive OP approach. I personally understand the risks (and potential gains) and find the comparison experiment interesting.

[MindTheGAAP]

55% UPRO (risk parity)
80% UPRO (target 20% vol)


This sound correct?

56% right now for risk parity. 80% sounds about right for 20% volatility. Mine shows as 74% but it's due to weighting most recent days more than days > 10

Can you send me in the right direction -- or provide a quick crash course -- on how I would practically implement the 20% volatility effort? Trying to work out how I'm going to rebalance and right now, I'm set just to realign to the 60/40 split on a monthly basis. May need to tweak it be weekly though since yesterday shifted me from 60/40 to 68/32.

60/40 UPRO/TMF monthly forever is a poor choice (OP is 40/60). I also wouldn't rebalance monthly unless resetting allocation for target volatility or risk parity w/ a 1 month look-back.

Funny to see how quick people are adopting these strategies without much understanding of them... if you are just in the mood for gambling and don't want to learn a bunch and improve your conviction, I would just stick with the OP's 40/60 UPRO/TMF quarterly.

That was my misstep - I'm 40/60 - the same as OP. Was just going with monthly vs. quarterly rebalancing. Apologies.

[caklim00]

Wow, my 20 day volatility went to 45% UPRO, 55% TMF after yesterday. I'm weighting day 1-10 70% total (so 1-5 35% and 6-10 35%, 11-15 20% total, 16-20 10% total). Looks like a standard non-weighted one month 20% is now at UPRO 51%, TMF 49%.

20 day lookback for risk parity is 48% UPRO, 52% TMF.

If it were me I wouldn't be checking the recommended AA unless it was time to rebalance, just asking for trouble. Not a great start to the month but we will see how things shake out.

There is an elegance to the OPs mindless quarterly rebalance though, absolutely.

Well, I'm still waiting on my money to make it over to M1, so I want it invested at whatever current values are. Sent partial transfer request 8 days ago. Still waiting...

[HEDGEFUNDIE]

I don't have any technical analysis or additional tweaks to the system but still wanted to share.....I had been waiting for a slight equities pullback to implement this strategy in my Roth IRA. I put 40% in UPRO on August 1, which is now down 11% in 3+ days(!!!) And yet with the 60% in TMF up 8.6% in that same brief time, I actually am still up overall.

Thank you HEDGEFUNDIE for your insight and creativity, and for sharing this with the group. Now I'm looking forward to that UPRO recovery.

From one Chicagoan to another - you're welcome!

[MotoTrojan]

Wow, my 20 day volatility went to 45% UPRO, 55% TMF after yesterday. I'm weighting day 1-10 70% total (so 1-5 35% and 6-10 35%, 11-15 20% total, 16-20 10% total). Looks like a standard non-weighted one month 20% is now at UPRO 51%, TMF 49%.

20 day lookback for risk parity is 48% UPRO, 52% TMF.

If it were me I wouldn't be checking the recommended AA unless it was time to rebalance, just asking for trouble. Not a great start to the month but we will see how things shake out.

There is an elegance to the OPs mindless quarterly rebalance though, absolutely.

Well, I'm still waiting on my money to make it over to M1, so I want it invested at whatever current values are. Sent partial transfer request 8 days ago. Still waiting...

Maybe get it transferred over at 40/60 then adjust.

[MotoTrojan]

That was my misstep - I'm 40/60 - the same as OP. Was just going with monthly vs. quarterly rebalancing. Apologies.

Quarterly did better historically and the reason makes sense (catching a following knife on the way down and not letting momentum run on the way up).

[MindTheGAAP]

That was my misstep - I'm 40/60 - the same as OP. Was just going with monthly vs. quarterly rebalancing. Apologies.

Quarterly did better historically and the reason makes sense (catching a following knife on the way down and not letting momentum run on the way up).

Thanks - looking into the target volatility strategy now since that makes sense to me at a theoretical level (IE at what mix do you have your targeted volatility given the last 20-days/ x-days of movement) but trying to ensure I'm fully kosher on how I'd implement it -- which was basis to my question earlier. Don't want to dive into the target-vol effort without fully thinking through the practical application aspects of it. Am playing with PV now but maybe you can validate what I'm finding:

Target volatility: 20%
Downside volatility: no
Out of market asset: TMF
Trade execution: at month end price
time period #1: 20 days - 100%
allocation: UPRO 100% --- should this be UPRO 80% TMF 20% if we're looking to cap at 80% UPRO? would make sense to me but validating.

Gives me a sharpe ratio of 1.52 and us market correlation is 0.35 -- CAGR is 36.80% -- link https://www.portfoliovisualizer.com/tes ... total1=100

Can you let me know - if not too much trouble - if I'm going about this the right way?

[MotoTrojan]

That was my misstep - I'm 40/60 - the same as OP. Was just going with monthly vs. quarterly rebalancing. Apologies.

Quarterly did better historically and the reason makes sense (catching a following knife on the way down and not letting momentum run on the way up).

Thanks - looking into the target volatility strategy now since that makes sense to me at a theoretical level (IE at what mix do you have your targeted volatility given the last 20-days/ x-days of movement) but trying to ensure I'm fully kosher on how I'd implement it -- which was basis to my question earlier. Don't want to dive into the target-vol effort without fully thinking through the practical application aspects of it. Am playing with PV now but maybe you can validate what I'm finding:

Target volatility: 20%
Downside volatility: no
Out of market asset: TMF
Trade execution: at month end price
time period #1: 20 days - 100%
allocation: UPRO 100% --- should this be UPRO 80% TMF 20% if we're looking to cap at 80% UPRO? would make sense to me but validating.

Gives me a sharpe ratio of 1.52 and us market correlation is 0.35 -- CAGR is 36.80% -- link https://www.portfoliovisualizer.com/tes ... total1=100

Can you let me know - if not too much trouble - if I'm going about this the right way?

You sure are. PV won’t allow you to cap UPRO at 80% while also not impacting it when the target would be less than 80% so no your idea there is off; that would target your vol based on 80/20. Someone above manually did this though to see the impact.

Lastly you should take the time to download the 1987+ data, making decisions based on this raging bull is flawed.

[EfficientInvestor]

Timestamp 38:10 https://bogleheads.podbean.com/

I listeded to the bit on multifactor and risk parity. For multifactor, nothing new, didn't really answer the question as to whether Vanguard Multifactor was good approach, just was really vague saying Vanguard is light on factors. Was more of a blanket Vanguard statement and not focused on the facotr funds specifically. Risk Parity discussion made sense, but this seems different.

I haven't listened to the whole thing yet, but the part about risk parity prompted me to put together the following analysis that looks at, "what would happen if you combine factor investing and risk parity?"

The first link below shows a comparison of a Vanguard Balanced (60/40) portfolio vs a variety of 50/50 (stock/intermediate treasury) portfolios that use factor tilts. P1 stock is tilted toward small cap value, P2 is tilted towards large cap growth, and P3 is a blend of the two. The tilts toward the higher risk factors allows you to move from a 60/40 allocation to a 50/50 allocation (either P1 or P2) and actually have better returns and risk-adjusted returns than the 60/40. As Larry stated, these riskier factors don't necessarily have higher risk-adjusted returns than the overall market. However, due to the risk premium in the stock factors, you are able to move some of your stock allocation over to bonds and move closer to risk parity. The move closer to risk parity is what is driving the Sharpe ratio higher. Then, if you combine the small cap value and large cap growth, you add a diversification benefit on top of the risk premium you are already getting above the S&P. That leads to a Sharpe Ratio of 0.78 for P3 vs the 0.65 of the 60/40.

https://www.portfoliovisualizer.com/bac ... total3=100

Now...what if you move even further toward risk parity? The main thing to point out in the backtest below is P1 vs P2. Both are a 25/75 blend that are leveraged by a factor of 2. The stock exposure is still 50% overall, but the bonds are leveraged to achieve a better risk parity. P1 uses the factor tilts and P2 just uses S&P 500 for the stock portion. P1 represents the combination of the factor tilts with risk parity. As expected, P2 has a better return and Sharpe ratio (0.90) than either the S&P 500 or the 60/40 blend. Then, when you add factor tilts on top of that, you get an even better return and Sharpe ratio (0.96).

https://www.portfoliovisualizer.com/bac ... total3=100

Here is a selected summary of results:

Mar 1993 - Jul 2019
60 LC/40 ITT - CAGR = 8.05%, SD = 8.87%, Max DD = -32.6%, Sharpe = 0.65
25 SCV/25 LCG/50 ITT - CAGR = 9.18%, SD = 8.7%, Max DD = -23.5%, Sharpe = 0.78
50 LC/150 ITT - CAGR = 10.56%, SD = 9.03%, Max DD = -16.1%, Sharpe = 0.90
25 SCV/25 LCG/150 ITT - CAGR = 11.95%, SD = 9.85%, Max DD = -17.0%, Sharpe = 0.96

I think this makes a strong case for factor tilting AND risk parity.

[MindTheGAAP]

Thanks - looking into the target volatility strategy now since that makes sense to me at a theoretical level (IE at what mix do you have your targeted volatility given the last 20-days/ x-days of movement) but trying to ensure I'm fully kosher on how I'd implement it -- which was basis to my question earlier. Don't want to dive into the target-vol effort without fully thinking through the practical application aspects of it. Am playing with PV now but maybe you can validate what I'm finding:

Target volatility: 20%
Downside volatility: no
Out of market asset: TMF
Trade execution: at month end price
time period #1: 20 days - 100%
allocation: UPRO 100% --- should this be UPRO 80% TMF 20% if we're looking to cap at 80% UPRO? would make sense to me but validating.

Gives me a sharpe ratio of 1.52 and us market correlation is 0.35 -- CAGR is 36.80% -- link https://www.portfoliovisualizer.com/tes ... total1=100

Can you let me know - if not too much trouble - if I'm going about this the right way?

You sure are. PV won’t allow you to cap UPRO at 80% while also not impacting it when the target would be less than 80% so no your idea there is off; that would target your vol based on 80/20. Someone above manually did this though to see the impact.

Lastly you should take the time to download the 1987+ data, making decisions based on this raging bull is flawed.
[/quote]

Absolutely will -- just wanted to make sure I was at least conceptually looking at it the correct way. Appreciate the feedback and guidance. I have a load of free trades with TDAmeritrade where I'm implementing this in my Roth so happy to go to the monthly rebalance based on target vol. if it is going to provide stronger risk-adjusted results.

[MotoTrojan]

Timestamp 38:10 https://bogleheads.podbean.com/

I listeded to the bit on multifactor and risk parity. For multifactor, nothing new, didn't really answer the question as to whether Vanguard Multifactor was good approach, just was really vague saying Vanguard is light on factors. Was more of a blanket Vanguard statement and not focused on the facotr funds specifically. Risk Parity discussion made sense, but this seems different.

I haven't listened to the whole thing yet, but the part about risk parity prompted me to put together the following analysis that looks at, "what would happen if you combine factor investing and risk parity?"

The first link below shows a comparison of a Vanguard Balanced (60/40) portfolio vs a variety of 50/50 (stock/intermediate treasury) portfolios that use factor tilts. P1 stock is tilted toward small cap value, P2 is tilted towards large cap growth, and P3 is a blend of the two. The tilts toward the higher risk factors allows you to move from a 60/40 allocation to a 50/50 allocation (either P1 or P2) and actually have better returns and risk-adjusted returns than the 60/40. As Larry stated, these riskier factors don't necessarily have higher risk-adjusted returns than the overall market. However, due to the risk premium in the stock factors, you are able to move some of your stock allocation over to bonds and move closer to risk parity. The move closer to risk parity is what is driving the Sharpe ratio higher. Then, if you combine the small cap value and large cap growth, you add a diversification benefit on top of the risk premium you are already getting above the S&P. That leads to a Sharpe Ratio of 0.78 for P3 vs the 0.65 of the 60/40.

https://www.portfoliovisualizer.com/bac ... total3=100

Now...what if you move even further toward risk parity? The main thing to point out in the backtest below is P1 vs P2. Both are a 25/75 blend that are leveraged by a factor of 2. The stock exposure is still 50% overall, but the bonds are leveraged to achieve a better risk parity. P1 uses the factor tilts and P2 just uses S&P 500 for the stock portion. P1 represents the combination of the factor tilts with risk parity. As expected, P2 has a better return and Sharpe ratio (0.90) than either the S&P 500 or the 60/40 blend. Then, when you add factor tilts on top of that, you get an even better return and Sharpe ratio (0.96).

https://www.portfoliovisualizer.com/bac ... total3=100

Here is a selected summary of results:

Mar 1993 - Jul 2019
60 LC/40 ITT - CAGR = 8.05%, SD = 8.87%, Max DD = -32.6%, Sharpe = 0.65
25 SCV/25 LCG/50 ITT - CAGR = 9.18%, SD = 8.7%, Max DD = -23.5%, Sharpe = 0.78
50 LC/150 ITT - CAGR = 10.56%, SD = 9.03%, Max DD = -16.1%, Sharpe = 0.90
25 SCV/25 LCG/150 ITT - CAGR = 11.95%, SD = 9.85%, Max DD = -17.0%, Sharpe = 0.96

I think this makes a strong case for factor tilting AND risk parity.

Since when is large cap growth considered a factor that should outperform? Seems you’re overfitting to past data? Maybe momentum would be a better option fundamentally?

[ocrtech]

Absolutely will -- just wanted to make sure I was at least conceptually looking at it the correct way. Appreciate the feedback and guidance. I have a load of free trades with TDAmeritrade where I'm implementing this in my Roth so happy to go to the monthly rebalance based on target vol. if it is going to provide stronger risk-adjusted results.

I noticed when I was modeling the targeted volatility approach using the simulated data, the returns are very sensitive to start date. I ran the model 20 times pushing the start date forward by one day for each run. The graph below shows the results. The blue line at the very top of the graph is the one associated with the initial model run without any data shifting. Targeted volatility was 20% and no cap on percentage of UPRO allowed.

[MotoTrojan]

Absolutely will -- just wanted to make sure I was at least conceptually looking at it the correct way. Appreciate the feedback and guidance. I have a load of free trades with TDAmeritrade where I'm implementing this in my Roth so happy to go to the monthly rebalance based on target vol. if it is going to provide stronger risk-adjusted results.

I noticed when I was modeling the targeted volatility approach using the simulated data, the returns are very sensitive to start date. I ran the model 20 times pushing the start date forward by one day for each run. The graph below shows the results. The blue line at the very top of the graph is the one associated with the initial model run without any data shifting. Targeted volatility was 20% and no cap on percentage of UPRO allowed.

Interesting. Hydromod did this and got different results I thought.

Could you provide basic return and performance values?

[ocrtech]

Absolutely will -- just wanted to make sure I was at least conceptually looking at it the correct way. Appreciate the feedback and guidance. I have a load of free trades with TDAmeritrade where I'm implementing this in my Roth so happy to go to the monthly rebalance based on target vol. if it is going to provide stronger risk-adjusted results.

I noticed when I was modeling the targeted volatility approach using the simulated data, the returns are very sensitive to start date. I ran the model 20 times pushing the start date forward by one day for each run. The graph below shows the results. The blue line at the very top of the graph is the one associated with the initial model run without any data shifting. Targeted volatility was 20% and no cap on percentage of UPRO allowed.

Here is the same model running with a 80% cap on UPRO


And this is the same thing but with a 60% cap on UPRO

[MotoTrojan]

I noticed when I was modeling the targeted volatility approach using the simulated data, the returns are very sensitive to start date. I ran the model 20 times pushing the start date forward by one day for each run. The graph below shows the results. The blue line at the very top of the graph is the one associated with the initial model run without any data shifting. Targeted volatility was 20% and no cap on percentage of UPRO allowed.

Would you mind doing the same exercise, using the same method, for the OPs 40/60 quarterly method as a spot check?

[EfficientInvestor]

Since when is large cap growth considered a factor that should outperform? Seems you’re overfitting to past data? Maybe momentum would be a better option fundamentally?

I was wondering if someone was going to call me out on that. I should have just left it at small cap value. I was ultimately just trying to show the benefit of additional diversification on the stock side. I agree that momentum would be better, but you can't really implement it with Leveraged ETFs. For small cap value, you can get pretty close with a Russell 2000 fund. But there aren't any leveraged momentum funds that I am aware of. Maybe you could use options on a momentum ETF? As an alternative, I chose to go with another option that has shown to have risk premium that allows you to put less allocation towards stocks and more towards bonds. I decided that the best alternative would be the Nasdaq (large cap growth).

[MoneyMarathon]

I agree that momentum would be better, but you can't really implement it with Leveraged ETFs.

Momentum (MTUM) has 20% standard deviation on its own, higher than the S&P 500 and closer to the volatility of emerging markets (VWO), so it could be a 'conservative' holding for leveraged ETF fans.

[Hydromod]

I had done an investigation of rebalancing frequency with the UPROSIM/TMFSIM dataset. I concluded that rebalancing every day or two gave some long-term benefit of an extra 1 to 2 percent of CAGR. You won't see this performance very easily for durations of less than 5 or 10 years though.

There wasn't very much difference between biweekly through quarterly rebalancing, although the spread in results with different starting date increased significantly.

I just did a similar quick calculation, implementing bands and monthly/bimonthly/quarterly rebalancing. In all cases I used the risk parity weights. With bands, I looked at UPRO different from the current daily value of the risk parity weight by various percentages as the trigger, up to 15 percent either way.

It looks like bands based on UPRO changing more than a few percent from allocated weight give nearly daily rebalancing, which gives a similar improvement to CAGR as daily rebalancing.

Bands of 10 percent give approximately monthly rebalancing on average, but occasionally require daily or near-daily rebalancing. Overall CAGR was slightly better than monthly rebalance.

Some decades, the bands performed better than periodic rebalancing. Some decades, periodic rebalancing performed better or was comparable.

My suspicion is that very frequent rebalancing may give a bit of overall performance during rapidly changing conditions, but it very likely isn't worth the work and expense of chasing this incremental performance for anyone on this forum.

There's more noise with 20-day lookback than 60-day lookback.

For this case, monthly rebalancing with 20-day volatility period gave 16.6 CAGR for the 1987-2019 period, daily gave 18.3, quarterly gave 17.1, and almost all of the bands I checked gave between 17 and 18 CAGR.

Monthly rebalancing with 60-day volatility period gave 17.3 CAGR for the 1987-2019 period, daily gave 18.4, quarterly gave 16.5, and almost all of the bands I checked gave between 17 and 18 CAGR.

So for those that are asking about rebalancing frequency, I'm sticking with monthly for now but quarterly will likely do just fine as well.

Take it for what it's worth. Hope this helps.

[ocrtech]

I noticed when I was modeling the targeted volatility approach using the simulated data, the returns are very sensitive to start date. I ran the model 20 times pushing the start date forward by one day for each run. The graph below shows the results. The blue line at the very top of the graph is the one associated with the initial model run without any data shifting. Targeted volatility was 20% and no cap on percentage of UPRO allowed.

Would you mind doing the same exercise, using the same method, for the OPs 40/60 quarterly method as a spot check?

Here you go. By the way, all graphs end on the last trading day of Dec 2018.

[willthrill81]

I was up 4% since the beginning of July with my current 80/20 allocation, but now I'm down 4%. It's a multiple G-force roller coaster!

[Hydromod]

I noticed when I was modeling the targeted volatility approach using the simulated data, the returns are very sensitive to start date. I ran the model 20 times pushing the start date forward by one day for each run. The graph below shows the results. The blue line at the very top of the graph is the one associated with the initial model run without any data shifting. Targeted volatility was 20% and no cap on percentage of UPRO allowed.

Yes, I found that, with periodic rebalancing, CAGR for the 1987-2019 period could by different by up to a percent based on starting date within the first quarter. Which seems like a lot, but there it is.

[ocrtech]

I noticed when I was modeling the targeted volatility approach using the simulated data, the returns are very sensitive to start date. I ran the model 20 times pushing the start date forward by one day for each run. The graph below shows the results. The blue line at the very top of the graph is the one associated with the initial model run without any data shifting. Targeted volatility was 20% and no cap on percentage of UPRO allowed.

Yes, I found that, with periodic rebalancing, CAGR for the 1987-2019 period could by different by up to a percent based on starting date within the first quarter. Which seems like a lot, but there it is.

I was really excited by Targeted Volatility when I originally ran the model without shifting dates. Unfortunately, that one run seems to be an extreme outlier towards the positive. I guess it is a good reminder that we really aren't dealing with sufficient data to truly model any of these approaches.

[Hydromod]

Well, we do know that we need to check various starting dates to get a fairer picture. And this tells us that we shouldn't expect to get returns exactly like anyone else, even those starting close to each other. We can expect that all of the folks starting this up based on the thread are going to get somewhat different returns even if they followed exactly the same strategy.

[MotoTrojan]

I noticed when I was modeling the targeted volatility approach using the simulated data, the returns are very sensitive to start date. I ran the model 20 times pushing the start date forward by one day for each run. The graph below shows the results. The blue line at the very top of the graph is the one associated with the initial model run without any data shifting. Targeted volatility was 20% and no cap on percentage of UPRO allowed.

Yes, I found that, with periodic rebalancing, CAGR for the 1987-2019 period could by different by up to a percent based on starting date within the first quarter. Which seems like a lot, but there it is.

I was really excited by Targeted Volatility when I originally ran the model without shifting dates. Unfortunately, that one run seems to be an extreme outlier towards the positive. I guess it is a good reminder that we really aren't dealing with sufficient data to truly model any of these approaches.

To be clear it still seems to beat the OP strategy, no?

[Hydromod]

I was really excited by Targeted Volatility when I originally ran the model without shifting dates. Unfortunately, that one run seems to be an extreme outlier towards the positive. I guess it is a good reminder that we really aren't dealing with sufficient data to truly model any of these approaches.

When you do the comparisons, I'd suggest comparing against the OP strategy with exactly the same start dates and rebalance dates. At least this is an apples to apples comparison. I think you'll find that you will consistently get one or two percentage points higher with the targeted volatility when comparing side-by-side with the same start date.

[MotoTrojan]

I was really excited by Targeted Volatility when I originally ran the model without shifting dates. Unfortunately, that one run seems to be an extreme outlier towards the positive. I guess it is a good reminder that we really aren't dealing with sufficient data to truly model any of these approaches.

When you do the comparisons, I'd suggest comparing against the OP strategy with exactly the same start dates and rebalance dates. At least this is an apples to apples comparison. I think you'll find that you will consistently get one or two percentage points higher with the targeted volatility when comparing side-by-side with the same start date.

Excellent point. Should be easy enough to plot the delta between them which could be a very interesting plot.

[ocrtech]

To be clear it still seems to beat the OP strategy, no?

For raw returns, certainly. For risk adjusted returns, I'm not so sure. Take a look at the drawdown chart. It compares results from the original 60/40 approach, inverse volatility (66% cap), EWMA weighted inverse volatility (66% cap and 94% lambda), and targeted volatility (20% target, no cap on UPRO). The drawdown for target volatility is significantly higher than any of the other three approaches.

[ocrtech]

When you do the comparisons, I'd suggest comparing against the OP strategy with exactly the same start dates and rebalance dates. At least this is an apples to apples comparison. I think you'll find that you will consistently get one or two percentage points higher with the targeted volatility when comparing side-by-side with the same start date.

This is a really good idea.

[physixfan]

With 3x leverage, OP chose UPRO/TMF.

What about the situation for lower leverage with the same philosophy?

1x leverage: VOO/TLT? or VOO/EDV?

2x leverage: SSO/UBT?

[MotoTrojan]

With 3x leverage, OP chose UPRO/TMF.

What about the situation for lower leverage with the same philosophy?

1x leverage: VOO/TLT? or VOO/EDV?

2x leverage: SSO/UBT?

Less return but the efficiency improvements should work out still. For me I would see something more like 80/20 equity/bond with the equity being broad or perhaps sliced with some momentum or small-value, and then 20% EDV to help boost efficiency. 40/60 will be a bit weak in returns.

The 2x example I think would have all the downsides of leverage (high ER, borrowing costs) but not enough return to be worth it.

[MotoTrojan]

To be clear it still seems to beat the OP strategy, no?

For raw returns, certainly. For risk adjusted returns, I'm not so sure. Take a look at the drawdown chart. It compares results from the original 60/40 approach, inverse volatility (66% cap), EWMA weighted inverse volatility (66% cap and 94% lambda), and targeted volatility (20% target, no cap on UPRO). The drawdown for target volatility is significantly higher than any of the other three approaches.

Strange that drawdown for 20% vol target is so much lower in PV...?

[MoneyMarathon]

With 3x leverage, OP chose UPRO/TMF.

What about the situation for lower leverage with the same philosophy?

1x leverage: VOO/TLT? or VOO/EDV?

2x leverage: SSO/UBT?

Easiest way to get started with these ideas is to go back to page 49 and read from there:

viewtopic.php?f=10&t=272007&start=2400

Several posts examined various ideas here.

Without any specific purpose in mind (i.e. if you are happy with a mix of equity risk, credit risk, and term risk and don't have any particular desired ratio of them that you'd like), and if there is enough space in tax advantaged, PSLDX may suit people looking for less than 3x leverage overall (it's about 2.15x levered overall). NSTX might work in taxable for people who like its 1.5x leverage approach. Both are reviewed in the recent pages of the thread.

1x leverage is usually just called long-only. EDV has a kind of 1.5x "accounting leverage" over TLT, because it gets 1.5x the duration exposure, but it's not actually using any borrowing or financial leverage. It's achieving this 1.5x long term treasuries by way of extended duration: long bonds with the coupons stripped (payable at maturity). If you want some leverage and want long term treasuries, but don't want much leverage, it's probably what you look at first. However, keep in mind, with a long-only portfolio, you're possibly not getting the return of the market, if you're just buying something like VOO and VTI and mixing it with something else. You need to apply leverage on the stock side to maintain 1x exposure to the market (while still being invested in broad market indexes). PSLDX does this if it's your only holding (maintains about 1x exposure to the S&P 500).

Or UPRO does this, in a mix with other holdings. You can reduce UPRO to 30% or 35% of the portfolio (0.9x or 1.05x exposure to the S&P 500) if you don't want as much exposure to the market as in the OP's strategy, then fill out the rest of the portfolio with things other than market beta.

[MoneyMarathon]

The 2x example I think would have all the downsides of leverage (high ER, borrowing costs) but not enough return to be worth it.

I believe the conclusion for roll-your-own with less leverage, which still is applicable here, is to mix 3x and 1x but avoid 2x, all else being equal.

It is worth it, for the same reason 3x is worth it: excess returns exceed costs. I'm not sure what you're saying. I understand not everyone would want to do it, especially those who prefer to construct a 3x portfolio. The main differences will be the kinds of risks taken (mostly the same for SSO/UBT) & their degree of effect.

[physixfan]

With 3x leverage, OP chose UPRO/TMF.

What about the situation for lower leverage with the same philosophy?

1x leverage: VOO/TLT? or VOO/EDV?

2x leverage: SSO/UBT?

Easiest way to get started with these ideas is to go back to page 49 and read from there:

viewtopic.php?f=10&t=272007&start=2400

Several posts examined various ideas here.

Without any specific purpose in mind (i.e. if you are happy with a mix of equity risk, credit risk, and term risk and don't have any particular desired ratio of them that you'd like), and if there is enough space in tax advantaged, PSLDX may suit people looking for less than 3x leverage overall (it's about 2.15x levered overall). NSTX might work in taxable for people who like its 1.5x leverage approach. Both are reviewed in the recent pages of the thread.

1x leverage is usually just called long-only. EDV has a kind of 1.5x "accounting leverage" over TLT, because it gets 1.5x the duration exposure, but it's not actually using any borrowing or financial leverage. It's achieving this 1.5x long term treasuries by way of extended duration: long bonds with the coupons stripped (payable at maturity). If you want some leverage and want long term treasuries, but don't want much leverage, it's probably what you look at first. However, keep in mind, with a long-only portfolio, you're possibly not getting the return of the market, if you're just buying something like VOO and VTI and mixing it with something else. You need to apply leverage on the stock side to maintain 1x exposure to the market (while still being invested in broad market indexes). PSLDX does this if it's your only holding (maintains about 1x exposure to the S&P 500).

Or UPRO does this, in a mix with other holdings. You can reduce UPRO to 30% or 35% of the portfolio (0.9x or 1.05x exposure to the S&P 500) if you don't want as much exposure to the market as in the OP's strategy, then fill out the rest of the portfolio with things other than market beta.

Thank you for your explanation! I think the risk parity philosophy is useful even without the leverage part. I made a comparison in Portfolio Visualizer: SPY/TLT 40/60 v.s. the classical SPY/BND 80/20.



It seems the risk parity portfolio (SPY/TLT 40/60) performed very well in 2008/2009. If I use a 20-day volatility look back adaptive risk parity strategy, the overall performance is even better slightly. I think the performance over a really bad time is really important to me in today's environment.

[MotoTrojan]

The 2x example I think would have all the downsides of leverage (high ER, borrowing costs) but not enough return to be worth it.

I believe the conclusion for roll-your-own with less leverage, which still is applicable here, is to mix 3x and 1x but avoid 2x, all else being equal.

It is worth it, for the same reason 3x is worth it: excess returns exceed costs. I'm not sure what you're saying. I understand not everyone would want to do it, especially those who prefer to construct a 3x portfolio. The main differences will be the kinds of risks taken (mostly the same for SSO/UBT) & their degree of effect.

I meant not worth it using 2x funds directly, as you also alluded to. Expense ratio is about the same for these 2x vs. 3x funds and as you noted, you can reduce leverage by simply diluting with some unleveraged funds and save on the expenses. It also seems some of the 3x funds are better run and more liquid than 2x at Direxion at-least (haven't compared ProShares but it seems they are superb at the 3x so I presume 2x as well).

[MotoTrojan]

It seems the risk parity portfolio (SPY/TLT 40/60) performed very well in 2008/2009. If I use a 20-day volatility look back adaptive risk parity strategy, the overall performance is even better slightly. I think the performance over a really bad time is really important to me in today's environment.

Bonds have been a tear though, so if you need equity like returns you are unlikely to see them in the future without a close to (or greater) than 1x equity exposure. TLT alone has returned 6.24% since 2003, most people would be thrilled to get that on their equities today.

[caklim00]

Wow, my 20 day volatility went to 45% UPRO, 55% TMF after yesterday. I'm weighting day 1-10 70% total (so 1-5 35% and 6-10 35%, 11-15 20% total, 16-20 10% total). Looks like a standard non-weighted one month 20% is now at UPRO 51%, TMF 49%.

20 day lookback for risk parity is 48% UPRO, 52% TMF.

If it were me I wouldn't be checking the recommended AA unless it was time to rebalance, just asking for trouble. Not a great start to the month but we will see how things shake out.

There is an elegance to the OPs mindless quarterly rebalance though, absolutely.

Well, I'm still waiting on my money to make it over to M1, so I want it invested at whatever current values are. Sent partial transfer request 8 days ago. Still waiting...

Maybe get it transferred over at 40/60 then adjust.

Not possible. Before I put in the transfer request I bought UPRO, but when I went to buy TMF it gave an error message. Called and they told me that TMF is restricted as a "protection." Next day I sold some UPRO and bought TLT in a 31/69 UPRO/TLT allocation for 10k (20 day lookback risk parity). I sold the other 20k just to move over in cash. I've watched TLT provide some assistance on down UPRO days but not like TMF (obviously) so just an waiting it out. Tomorrow will be day 9 from the transfer request so it has to be soon.

[Hydromod]

You might want to check with M1 to make sure everything is going well. I tried transferring an entire fund, part went through but not the rest. I got a notification that the transfer was made, but not that only part of the request transferred. I finally got an explanation when I followed up a couple of times. There was a good reason the transfer wasn't made, but communication about it was lacking.

[ocrtech]

Strange that drawdown for 20% vol target is so much lower in PV...?

Good catch. That was 30% targeted volatility. Here is the one for 20%.

[mrspock]

Finally got the courage to take a peak at my portfolio today.... 40/60 (UPRO/TMF) is up 35% (!) since I began this adventure (Feb 12th). It weathered this past week with flying colors. It's rather amusing to look at the account growth chart, it's rather easy to see where this started , the slope of the line is rather.... pronounced.

Now my standard Bogehead 75/25 portfolio....ya well.... let's just say it's nice to have this as a distraction. It's going to be a rough month by the looks of things.

[MoneyMarathon]

Thank you for your explanation! I think the risk parity philosophy is useful even without the leverage part. I made a comparison in Portfolio Visualizer: SPY/TLT 40/60 v.s. the classical SPY/BND 80/20.

It seems the risk parity portfolio (SPY/TLT 40/60) performed very well in 2008/2009. If I use a 20-day volatility look back adaptive risk parity strategy, the overall performance is even better slightly. I think the performance over a really bad time is really important to me in today's environment.

Using 40% stocks / 60% bonds is definitely a way to lower risk when not leveraged. If you're not comfortable taking more risk, and if it will let you meet your goals (without any expectations that the investment will grow more than slowly), it's a decent idea.

[vel]

I am about 20 days in on the orig 40/60 upro/tmf. It drifted to 33/67. I am adding some more money to this today and having trouble deciding if I should maintain the 33/67, or use the new money to rebalance it back to 40/60. Since I am only 20 days in, I am leaning toward maintaining the 33/67 (in keeping with the quarterly rebalance).

Does this sound reasonable?

[Hydromod]

I am about 20 days in on the orig 40/60 upro/tmf. It drifted to 33/67. I am adding some more money to this today and having trouble deciding if I should maintain the 33/67, or use the new money to rebalance it back to 40/60. Since I am only 20 days in, I am leaning toward maintaining the 33/67 (in keeping with the quarterly rebalance).

Does this sound reasonable?

I would think that the approach consistent with the OP would be to assume that the weights are valid at all times. So I would always use new money to get back to 40/60 if I was following the OP strategy, and either start a new quarter or keep the same schedule. Note that it doesn't necessarily hurt to rebalance more frequently, it's more the luck of the draw whether it works better or not.