Post New

https://www.quantifiedstrategies.com/triple-levera... has an interesting backtest I or we have talked about in part or something akin to it before.

The core proposal is a strategy designed to mitigate large drawdowns while retaining much of the upside performance:

The Portfolio: Establish a portfolio consisting of an equal dollar amount of TQQQ and TMF (the triple leveraged ETF linked to the 20+ year treasury bond).
The Mechanism: This strategy works because equity TLETFs (like QQQ, SPY, and DIA) are generally negatively correlated with TLT (the underlying index for TMF).
Rebalancing: The strategy requires bimonthly rebalancing (every two months), which was found to be the optimal interval based on backtesting.
Crash filter: If TQQQ drops 20% or more in a single day, exit both TQQQ and TMF. Move 100% into IEF (a 7-10 year Treasury ETF). Stay there until TQQQ recovers and exceeds its pre-crash price. Then return to the 50/50 split.

Strategy Performance Results

Over the 10+ years of TQQQ’s existence, the 50/50 TQQQ/TMF bimonthly strategy achieved a total return in excess of 5,800%. The resulting CAGR was 44.9%.

Critically, the maximum EOM drawdown was significantly reduced to less than 25% (specifically, 24.5%). This is roughly half the drawdown observed for TQQQ alone (49.1%).

Lewis Glenn authored

10 years. Bull market.

Any thoughts on how it would do if (when) the market takes a serious turn to the downside?

https://www.quantifiedstrategies.com/triple-levera... has an interesting backtest I or we have talked about in part or something akin to it before.

The core proposal is a strategy designed to mitigate large drawdowns while retaining much of the upside performance:



According to yahoo quotes, the top TQQQ daily losses were on these days:

-34.47%   3/16/2020
-27.41%   3/12/2020
-20.31%   3/9/2020
-18.31%   4/4/2025  <<<< Not -20%

These are all adjusted for dividends, not price-only. The figures come out the same either way.


So essentially that sell only triggered once, on 3/9/2020, so the sell would be on 3/10/2020.

(Funnily enough, 3/10/2020 closed up +15.56%.)

Anyway, TQQQ closed on 3/9/20 at 15.30749989.
The pre-crash price on 3/6/20 was 19.20999908

It opened 3/10/20 at 17.69000053, so that's where you would have sold.

The next open above 19.21 was on 5/18/20 at 19.2775001525878

Between those dates, the lowest close was 8.88 on 3/20/20. (That's -54% drop.)

So this strategy had you out of TQQQ/TMF for 49 days out of 3950 days.


You can do a backtest, except for the sell & buy signals. https://testfol.io/?s=6OIvmhtuudA Be sure to check the "Logarithmic scale" box.

Eyeball the chart and draw a flat line between 3/10/20 and 5/18/20 on the TQQQ/TMF plot. During the same period, IMF was flat, value $16,720 to $16,718.

I'm sorry, but this simplistic backtest says that this strategy SUCKS.
It saved you from one (very large) drawdown at the cost of $2 million!

If you are going to put money on TQQQ you have to realize that you will have huge volatility and drawdowns.
MaxDD:
-82% TQQQ
-78% TMF/TQQQ



Another similar strategy I have read about is TQQQ/BTAL with annual rebalance. That one really reduces volatility and the MaxDD, -36% vs. -82% for TQQQ.
At the low, low cost of only $1,200,000. https://testfol.io/?s=9eGpbHvey38

Actually, that is better than TMF/TQQQ over the same period beginning 9/13/2013.
Ending at $314,500 vs. $147,300. At _lower_ MaxDD and similar volatility.

Sorry to rain on his parade.

This is an example of why you _MUST_ do your own backtests.
The writer of that article (Oddmund Groette) just accepted the claims of the originator (Lewis Glenn) at face value without checking anything.
The purported fact checker (Håkan Samuelsson) obviously didn't check anything further than spelling and punctuation.

It didn't do a heck of a job of "mitigate large drawdowns while retaining much of the upside performance". Dropped the MaxDD from -82% to -78% while giving only 13% of the return, and half the CAGR.

All together: https://testfol.io/?s=0ersxbG9Ce9

It took less than 5 minutes to create the backtest at testfol.io. It took longer to download the daily prices and plug that data into Excel to find the -20% days.

ouch!

Well done Ray!

My findings match your own using CLOSE to CLOSE for the twenty percent drop. There are just three and they are all in March of 2020.

Date       C/C DD
3/16/2020  -34.46%
3/12/2020  -27.41%
3/9/2020   -20.31%

I'm wondering if the creator actually used CLOSE to LOW for the -20% drop, since there were have been no moves from 2/11/2010 to 2/10/2020 - the ten years he references. Using this method it would produce the following moves.

Date       Signal  Holdings
2/11/2010  BULL    TQQQ/TMF
5/6/2010   BEAR    IEF
5/12/2010  BULL    TQQQ/TMF
8/24/2015  BEAR    IEF
8/27/2015  BULL    TQQQ/TMF
3/9/2020   BEAR    IEF
3/24/2020  BULL    TQQQ/TMF
4/10/2025  BEAR    IEF
4/25/2025  BULL    TQQQ/TMF

As you can see, this only adds a 2010 and a 2015 signal prior to the ten year mark of TQQQ.

Using this revised method (CLOSE to LOW) my results come closer to what the original author stated for returns, but not perfect by any means. More importantly, what if we bring this forward to the present.

Test          Results           ROI       CAGR    GSD    Sharpe  DDD3    BBW   Signal  Drawdown  Rebalance
Para-Test #1  Combined Results  5814.64%  29.68%  55.27  0.76    15.86%  2.01  100%    -80.95%   None
Para-Test #2  Combined Results  4792.57%  28.12%  41.34  0.83    15.25%  2     79%     -78.33%   Annually
Para-Test #3  Combined Results  3764.85%  26.21%  39.89  0.8     15.37%  1.85  79%     -78.12%   Bi-Annually
Para-Test #4  Combined Results  4241.40%  27.15%  38.75  0.83    15.28%  1.93  69%     -78.09%   Tri-Annually
Para-Test #5  Combined Results  4322.11%  27.30%  38.3   0.84    15.16%  1.95  73%     -78.10%   Quarterly
Para-Test #6  Combined Results  5046.41%  28.53%  37.98  0.87    14.94%  2.07  71%     -77.36%   6x Annually
Para-Test #7  Combined Results  4025.03%  26.74%  37.81  0.83    14.95%  1.94  67%     -76.77%   Monthly

TQQQ held as a LTBH investment has a CAGR of 43.32% from 2/11/2010 to present with a Max DD of -81.65%. 6x Annual rebalances drops this Max DD to -77.36% (the second one from the bottom).

This method gets slaughtered in 2022.

Symbol  Hold  #  PY  Entry Date  Exit Date   Entry Price  Exit Price  % G/L
TQQQ    Hold  1  1   12/13/2021  2/14/2022   $80.13       $53.43      -33.32%
TMF     Hold  2  1   12/13/2021  2/14/2022   $29.87       $21.94      -26.54%
TQQQ    Hold  1  2   2/14/2022   4/18/2022   $53.43       $47.13      -11.79%
TMF     Hold  2  2   2/14/2022   4/18/2022   $21.92       $14.67      -33.08%
TQQQ    Hold  1  2   4/18/2022   6/21/2022   $47.13       $24.38      -48.27%
TMF     Hold  2  2   4/18/2022   6/21/2022   $14.64       $11.10      -24.20%
TQQQ    Hold  1  2   8/22/2022   10/24/2022  $32.57       $21.27      -34.69%
TMF     Hold  2  2   8/22/2022   10/24/2022  $11.61       $6.28       -45.90%

I'm wondering if the creator actually used CLOSE to LOW for the -20% drop, since there were have been no moves from 2/11/2010 to 2/10/2020 - the ten years he references.

I had some more thoughts on this.

Regardless of the details, since you get back into TQQQ/TMF at or slightly above the price where you got out, all you are really doing is sitting out the round-trip to the bottom. IEF is just a place to park your money, it doesn't gain much.

The 50% TMF cuts the volatility in half compared to TQQQ. But it cuts the long-term return by almost 90%. I don't consider that a good trade-off.

If you are going to stomach the volatility of a large allocation to TQQQ, why not just bite the bullet and just buy TQQQ? Going 50/50 with TMF does no good, just makes a great hand-waving argument.

Yeah, there is the issue that TQQQ has only been during a long-running bull market. Maybe handle that by doing some sort of timing, maybe the 90-day-no-new-high of QQQ.

In the QQQ backtest, QQQ investment was a high on 1/3/2022 at $91,000. Quick drop to $74,000 on 3/11/2022.
Now, 2/24/2025 it is $143,000.

Leveraged investment on a gain like that could be attractive. If you can stand it.



As you can see, this only adds a 2010 and a 2015 signal prior to the ten year mark of TQQQ.

I have found that a lot of papers & articles claim much better results that I can duplicate with my own backtest.
There's even aname for this: "replication crisis"

Since 2010 TQQQ had an annualized return of 42%!, Max DD of -81%!, and Sortino of 1.26. I experimented, using P123, with various methods of limiting the Max DD. Everything I tried hugely stunted its growth.

The best I could do was a 25% annualized return, Max DD of -35%, and Sortino of 1.35 using the following P123 script checked daily:

(Close(0,GetSeries("$VIX"))<SMA(200,0,GetSeries("$VIX"))) // Market volatility calm
[OR] (Close(0,GetSeries("$VIX")))<15
(Close(0,GetSeries("$VIX")))<30
Close (0,33069) > Close (17,33069)
StockID = 33069 //TQQQ


Over this same period SP500 had a 13% annualized return, Max DD of -33%, and Sortino of 1.21.


Taz

I experimented, using P123, with various methods of limiting the Max DD. Everything I tried hugely stunted its growth.

I can't find my spreadsheet right now where I played with timing TQQQ.

But just remember the results. Which I coded up to automate it.

Best timing scheme I found was:
Note that this uses QQQ to time TQQQ, and it only trades once a month.

Each day, get the high of QQQ
Each week, compute the high for that week, and note which month it is in.

# Save the high of each month and the closing price of the month.
# Then
# For each month, see if this close is higher than the close of the
# previous 2 months & this month. If so, flag it.
# For each month, see if this month & the previous 2 are flagged.
# If so, "buy"
# If not, "sell"
# This is: if each of the latest 3 months saw a 3 month high = buy/keep.
# Decide buy/sell on the next-last weekday of the month, trade on last day of month.

Not sure how you could to this at P123. Let us know if you manage it.

Probably no new 90 day high in any of the last 90 days would work ok.

I don't recall trying no new 13 week high in the last 13 weeks, and trade weekly.

As you see from the chart, you want to hold TQQQ most of the time since inception.
Not sure how TQQQ would do outside of a bull market.
TQQQ Inception Date was 2/9/2010.

I show for 2/7/2010 - 6/11/2023:
CAGR: TQQQ 40.6% Timed-TQQQ 38.2%
MaxDD: -69% & -54% (testfol.io says -82% for TQQQ, so we differ there.)

TQQQ B&H and with timing.
Weekly 2/8/2010 to 11/3/2025

Timing rule:
For each week, see if QQQ hit a new high in the last 13 weeks.
For each week, see if any of the last 13 weeks itself had a new 13 week QQQ high.
If yes, buy/hold.
If no, sell at the Monday open.

TQQQ      B&H       Timed
CAGR      43.0%     49.2%
stdev     57.6%     55.5%
MaxDD     -80%      -69%
# weeks   820       820
# DD >50%  89        56
# DD >25% 232       185
       $281,164  $551,159 Growth of $1,000


Date      Action
2/15/10   Buy
 4/8/13   Sell
 6/6/16   Buy
 7/4/16   Sell
 5/30/22  Buy
10/24/22  Sell
 2/20/23  Buy
 2/27/23  Sell

Yeah, so I went back and re-read my code and couldn't make heads or tails of it. The perils of not documenting your code properly.

I don't remember why I figured the timing monthly instead of weekly or daily. Probably because all the other screens I use are run monthly. Also probably because not wanting to get frequent whipsaws if run daily.

So I went back and redid the calculations for weekly, using the data from Yahoo.
Only 4 out periods. Still many nosebleed periods.

---------------------------------
For comparison:

QLD       B&H       Timed
CAGR      33.1%     36.5%

QQQ       B&H       Timed
CAGR      19.5%     20.8%

Post New

Berkshire Hathaway FAQ

---

Contact Shrewd'm
Contact the developer of these message boards.