Skip to main content
No. of Recommendations: 4
By far, the best mechanical screen (and only one I still use) in my portfolio over the years has been a simple industry sector rotation suggested by tpoto here:

https://boards.fool.com/etf-sector-rotation-movement-3266321...

I wonder how it would work if leveraged ETFs were used instead of the 9 original unleveraged ETFs (all are 3x leverage except I could only find 2x for materials)?

XLK - TQQQ ProShares UltraPro QQQ 3x
XLV - CURE Direxion Daily Healthcare 3x
XLY - WANT Direxion Daily Consumer Discretionary 3x
XLP - NEED Direxion Daily Consumer Staples 3x
XLB - UYM ProShares Ultra Basic Materials 2x
XLU - UTSL Direxion Daily Utilities 3x
XLE - GUSH Direxion Daily S&P Oil & Gas Exploration & Production 3x
XLI - DUSL Direxion Daily Industrials 3x
XLF - FAS Direxion Daily Financial 3x

Any of you backtesting gurus interested in checking it out?
Print the post Back To Top
No. of Recommendations: 2
Somewhat off topic, but most (or at least, the ones I invested in) of the standard MI screens in the last few years have greatly underperformed the S&P 500. Also underperformed their backtested returns.

Maybe this is due to MSFT, AMZN, AAPL, FB, & GOOG dominating the weightings? Therefore SPY return has been essentially the return of those stocks and MI screens have not picked those stocks?
Print the post Back To Top
No. of Recommendations: 4
Yeah, but a simple monthly sector rotation based on those 9 sectors (IF = [SMA-30]/[SMA-200]>1 THEN SORT desc RS and pick top three) has beaten SPY, both in terms of volatility and CAGR, and for several years (as well as all of my other screens combined).

To me that's very MI and on topic. I suspect that a leveraged version would do even better in terms of CAGR, and maybe without compromising the Sharpe too much.
Print the post Back To Top
No. of Recommendations: 2
I suspect that a leveraged version would do even better in terms of CAGR, and maybe without compromising the Sharpe too much.

This is bit wild for my taste.
Might be best if you stay with the most common sponsor(of leveraged funds) with at least a net asset filter. Also why not include Inverse?

The following list includes 13 of 20 available sector funds by "DIREXION DAILY".
Results in 4 of 13 with positive Market Total Returns (1 Yr). Best 80%, worst -90%.

ID  Symbol  ETP Name -Sectors:Leveraged / Inverse                        Market Price    Net Assets    Leveraged / Inverse        Equity: Sector Objective  Index Composition  Inception  Number     Total
Date of Basket Returns
0 Holdings (1 Yr)
1 TECL DIREXION DAILY TECHNOLOGY BULL 3X SHARES $291.28 $1.8B 3x, Leveraged Information Technology Cap-Weighted 12/17/08 72 79.59%
2 LABU DIREXION DAILY S&P BIOTECH BULL 3X SHARES $59.65 $455.4M 3x, Leveraged Health Care Equal-Weighted 05/28/15 133 42.67%
3 SOXL DIREXION DAILY SEMICONDUCTOR BULL 3X SHARES $238.38 $1.5B 3x, Leveraged Information Technology Cap-Weighted 03/11/10 31 31.76%
4 CURE DIREXION DAILY HEALTHCARE BULL 3X SHARES $63.48 $124.5M 3x, Leveraged Health Care Cap-Weighted 06/15/11 63 21.93%
5 NAIL DIREXION DAILY HOMEBUILDERS & SUPPLIES BULL 3X S... $53.14 $431.4M 3x, Leveraged Consumer Discretionary Cap-Weighted 08/19/15 45 -18.42%
6 FAS DIREXION DAILY FINANCIAL BULL 3X SHARES $38.28 $1.5B 3x, Leveraged Financials Cap-Weighted 11/06/08 240 -52.03%
7 DRN DIREXION DAILY REAL ESTATE BULL 3X SHARES $11.83 $62.9M 3x, Leveraged Real Estate Cap-Weighted 07/16/09 179 -56.58%
8 FAZ DIREXION DAILY FINANCIAL BEAR 3X SHARES $13.72 $292.1M -3x, Inverse, Leveraged Financials Cap-Weighted 11/06/08 239 -58.65%
9 DFEN DIREXION DAILY AEROSPACE & DEFENSE BULL 3X SHARES $12.41 $225.0M 3x, Leveraged Industrials Cap-Weighted 05/03/17 34 -80.70%
10 LABD DIREXION DAILY S&P BIOTECH BEAR 3X SHARES $55.07 $62.7M -3x, Inverse, Leveraged Health Care Equal-Weighted 05/28/15 132 -85.38%
11 TECS DIREXION DAILY TECHNOLOGY BEAR 3X SHARES $14.23 $68.8M -3x, Inverse, Leveraged Information Technology Cap-Weighted 12/17/08 71 -87.21%
12 DPST DIREXION DAILY REGIONAL BANKS BULL 3X SHARES $53.66 $127.4M 3x, Leveraged Financials Equal-Weighted 08/19/15 130 -87.80%
13 SOXS DIREXION DAILY SEMICONDUCTOR BEAR 3X SHARES $42.50 $97.0M -3x, Inverse, Leveraged Information Technology Cap-Weighted 03/11/10 30 -90.03%

Leveraged / Inverse
Leveraged, Inverse
165
Sponsor
AdvisorShares, Direxion Shares & 3 more...
67
Net Assets
$53.68M and Above
35
Equity: Sector Objective
Communication Services, Consumer Discretionary & 10 more...
20

Equity: Sector Objective
Communication Services
0out of 15
Consumer Discretionary
1out of 28
Consumer Staples
0out of 27
Energy
2out of 63
Financials
3out of 47
Health Care
3out of 51
Industrials
1out of 35
Information Technology
5out of 76
Materials
4out of 35
Multi-Sector
0out of 114
Real Estate
1out of 44
Utilities
0out of 19

Industry Exposure
None selected
20
Index Composition
None selected
20
Inception Date
28 – 3 Years
14
Number of Basket Holdings
None selected
14
Market Total Returns (1 Yr)
None selected
14
Geography Objective
Domestic
13


GD_
Print the post Back To Top
No. of Recommendations: 11
Regarding the Ultra Sector ETFs- post 279013

I could get only these four to have
and decent history.

Direxion Daily Healthcare Bull 3X ETF (CURE)	
Direxion Daily Financial Bull 3X ETF (FAS)	
ProShares UltraPro QQQ (TQQQ)	
ProShares Ultra Basic Materials (UYM)

Starting with $10K picking the one
that did the best over the last 6 months.

Great when there's a bull market (in this case),
otherwise... you guessed it...
(Don't know if there are inverse ones, but
not going to look them up!!!)

		Ultra 	SPY	Ultra	        SPY
2012		38.74%	15.82%	$13,874 	$11,582 
2013		139.26%	32.18%	$33,194 	$15,309 
2014		49.77%	13.51%	$49,714 	$17,377 
2015		14.72%	1.25%	$57,035 	$17,595 
2016		3.85%	11.82%	$59,229 	$19,674 
2017		42.20%	21.67%	$84,225 	$23,938 
2018		10.23%	-4.52%	$92,843 	$22,856 
2019		87.64%	31.33%	$174,214 	$30,016 
2020		75.15%	9.65%	$305,143 	$32,913
Print the post Back To Top
No. of Recommendations: 2
Print the post Back To Top
No. of Recommendations: 1
Perhaps, but tpoto's version of sector rotation has significantly underperformed the last few years.
Average rank of 3, 6, 12 month returns, 2x the 6 months, invest in top average rank, hold for a month.

(https://boards.fool.com/etf-sector-switching-update-30752046...)

Not saying it can't work - just that the successful factors change over time. Top 3 by 1 month SMA/10 month SMA ratio is yet another momentum factor
Print the post Back To Top
No. of Recommendations: 7
.... but tpoto's version of sector rotation has
significantly underperformed the last few years


As long as few MI people made some bucks early on, I'm okay with that.

And just maybe because all the top hedge fund managers read
my post and started to follow it thus weakening the strategy.
<:^)
Print the post Back To Top
No. of Recommendations: 5
My actual 3-year returns at Etrade since 2017 are 51% for my sector screen vs. 33% for SPY, and the chart shows that the downdrafts were significantly less for the sector screen.

I use etfscreen to compare the performance of 9 select sector ETFs (XLB,XLE,XLF,XLI,XLK,XLP,XLU,XLV,XLY) with the formula _SMAR2 = SMA-30/SMA-200>1, SORT desc RSf.

I just ran 5-year backtests on etfscreen for that screen and for the equivalent using the leveraged ETFs (CURE,DUSL,FAS,GUSH,NEED,TQQQ,UTSL,UYM,WANT) and got these results:

SCREEN CAGR SD
SPY 13.9 19
SECTOR 13.4 17
LEVERAGED SECTOR 13.9 75

So, whoa! Not good. Same CAGR as SPY, but massive increase in volatility.

In comparison, just holding leveraged ETFs yields:

UPRO 25.9 79
TQQQ 52.3 84

So yet again, after decades of trying, my conclusion is KISS: as an individual investor, I can't beat the index. Even so, I think I'll continue to use the basic sector rotation for a corner of my portfolio. Despite what the backtest shows, my actual sector returns beat SPY in recent years (which is probably due to XLK being the top ETF pick for quite some time).

PS. After multiple 403 forbidden errors, updating my windows, rebooting several times, I discovered that you cannot use brackets in these posts. Strange!
Print the post Back To Top
No. of Recommendations: 3
I didn't know you couldn't use [brackets]
But it didn't work for me until I made them bold.
Print the post Back To Top
No. of Recommendations: 12
BAGoldman:
I wonder how it would work if leveraged ETFs were used instead of the 9 original unleveraged ETFs (all are 3x leverage except I could only find 2x for materials)?

If the Fama/French twelve sectors (https://mba.tuck.dartmouth.edu/pages/faculty/ken.french/ftp/...) are an acceptable substitute, then you can easily backtest both 1x and 3x versions of sector momentum (with the timing step) back to 1927. You can reduce the 12 sectors down to a smaller subset of sectors by eliminating any symbols you don't want (for example, it probably makes sense to eliminate F1212, "Other") from the custom universe code that looks like this (for 1x):

{!F1201T}{!F1202T}{!F1203T}{!F1204T}{!F1205T}{!F1206T}{!F1207T}{!F1208T}{!F1209T}{!F1210T}{!F1211T}{!F1212T}

Note that only ten of the twelve sectors go all the way back to 1927.

Backtest results using all twelve sectors are as follows:

                                  Top 3 of 12 by TR252 (1x)                             Top 3 of 12 by TR252 (3x)
^S5T Avg Min Max SD Avg Min Max SD
CAGR: 9.96 12.18
11.57 12.80 0.34 30.37 27.49 32.76 1.39
TR: 727,648 4,951,918 2,857,459 7,978,739 1,418,335 9,586,677,383,168 760,124,145,664 33,901,691,011,072 8,665,666,420,736
SAWR(20; 0.95): 4.63 7.71 7.12 8.41 0.34 7.18 4.68 9.00 1.11
GSD(20): 20.85 16.85 16.54 17.54 0.22 63.51 61.85 67.16 1.16
DIGSD(20; 0%): 23.25 19.85 19.43 21.01 0.34 80.44 77.74 86.90 1.97
LDD(20; 0%): 13.32 11.05 10.77 11.72 0.20 37.64 36.78 39.99 0.72
LDDD3: 13.51 9.06 8.43 10.04 0.34 26.63 25.22 30.07 1.02
MDD: -84.60 -58.44 -64.49 -53.74 3.15 -96.46 -97.60 -95.08 0.79
UI(20): 21.89 14.34 12.83 15.80 0.79 39.89 35.32 43.57 2.12
Sharpe(20): 0.42 0.62 0.59 0.66 0.02 0.78 0.73 0.82 0.02
Beta(20): 1.00 0.51 0.50 0.52 0.01 1.46 1.43 1.49 0.02
TI(20): 7.96 18.54 17.55 19.73 0.60 23.54 22.13 24.86 0.76
AT: 0.00 2.20 2.12 2.29 0.04 2.29 2.22 2.34 0.03

Total Return for Year Ending:
19271230 35.49 51.21 48.99 53.50 1.21 223.80 211.50 237.13 5.99
19281231 39.23 59.07 51.26 64.73 3.91 254.78 194.78 311.54 31.52
19291231 -8.54 -13.49 -28.68 -7.14 4.71 -66.56 -81.24 -57.17 5.18
19301231 -26.07 2.22 2.22 2.22 0.00 2.22 2.22 2.22 0.00
19311231 -45.45 1.11 1.11 1.11 0.00 1.11 1.11 1.11 0.00
19321230 -8.80 -13.48 -21.61 -5.80 4.53 -45.63 -61.29 -31.31 9.27
19331229 50.51 27.92 13.59 55.40 11.50 33.35 -5.20 100.55 30.75
19341231 -1.00 -2.18 -7.37 1.48 2.44 -10.99 -21.87 1.95 6.63
19351231 45.06 54.70 48.14 59.43 3.33 208.68 152.29 275.56 38.18
19361231 33.09 35.64 31.59 43.62 2.93 105.65 84.31 129.63 12.60
19371231 -35.95 -6.92 -10.28 -2.86 1.88 -24.36 -31.22 -17.11 3.93
19381230 28.56 6.17 3.80 9.21 1.40 18.49 11.37 27.28 4.84
19391229 -0.85 -8.97 -14.58 -3.17 2.58 -23.20 -38.27 -9.05 7.21
19401231 -9.67 -22.51 -26.30 -17.58 2.87 -57.59 -64.28 -49.23 5.09
19411231 -11.51 -12.78 -20.63 -5.97 3.83 -31.60 -45.11 -20.88 6.26
19421231 20.86 20.86 18.99 23.12 1.19 72.55 60.40 82.70 6.16
19431231 26.81 31.51 26.10 36.20 2.92 113.21 89.06 132.69 12.04
19441229 21.23 25.52 23.61 27.53 1.35 91.14 74.26 108.11 7.96
19451231 36.77 41.94 38.41 45.08 1.99 175.61 156.39 189.37 8.35
19461231 -9.24 5.01 -1.02 12.98 4.12 10.33 -7.21 38.62 13.08
19471231 4.81 4.30 2.27 5.88 0.94 9.83 5.13 16.84 3.00
19481231 5.22 -12.06 -16.85 -8.06 2.81 -35.87 -46.80 -26.95 5.99
19491230 18.29 17.80 14.27 19.49 1.32 60.00 46.00 65.93 4.93
19501229 32.75 35.22 33.19 37.93 1.33 123.49 110.75 137.98 6.88
19511231 23.37 17.64 15.49 19.77 1.46 54.18 40.72 66.54 6.84
19521231 18.98 18.32 17.00 20.79 1.03 65.97 58.94 75.65 4.83
19531231 -1.73 -2.29 -4.34 -0.75 1.06 -8.94 -15.29 -4.57 2.93
19541231 52.69 53.24 47.93 59.30 3.76 250.71 231.03 278.65 17.85
19551230 31.02 30.76 19.46 34.26 3.55 97.02 56.41 121.49 19.41
19561231 6.30 3.97 -0.51 9.16 2.64 9.33 -0.26 21.30 5.88
19571231 -10.62 -4.08 -11.03 3.39 3.90 -16.72 -33.89 3.87 10.35
19581231 43.75 30.46 26.15 33.61 1.97 120.83 109.93 135.92 8.53
19591231 12.37 20.12 13.96 22.56 2.33 71.55 48.96 85.29 9.02
19601230 0.86 -4.65 -10.67 2.29 3.78 -13.64 -28.36 2.80 9.66
19611229 27.32 34.71 33.18 37.25 0.96 135.21 114.50 144.65 8.02
19621231 -8.61 -9.61 -19.68 -4.28 3.74 -29.43 -51.15 -16.26 8.71
19631231 22.59 25.22 22.96 27.31 1.44 93.00 83.66 103.08 6.33
19641231 16.68 21.09 17.21 22.98 1.20 72.28 61.83 84.77 7.32
19651231 12.41 15.08 14.12 16.76 0.79 50.81 44.23 56.57 3.46
19661230 -10.10 3.71 -0.82 7.54 1.97 2.28 -10.31 14.42 6.65
19671229 23.56 23.55 18.96 28.49 2.59 79.41 60.49 109.76 11.62
19681231 10.92 9.05 6.09 12.08 1.65 26.47 15.74 36.59 6.00
19691231 -8.34 -11.31 -16.18 -5.72 3.14 -36.05 -45.83 -23.10 6.76
19701231 4.12 17.20 14.84 19.05 1.41 43.98 33.54 50.92 5.69
19711231 14.50 6.70 2.53 9.66 2.30 16.27 3.39 25.04 7.03
19721229 19.15 16.40 13.55 19.72 1.77 57.28 49.48 69.69 5.79
19731231 -14.79 -10.47 -14.70 -3.59 3.25 -33.95 -42.65 -19.86 6.93
19741231 -26.54 7.82 7.82 7.82 0.00 7.82 7.82 7.82 0.00
19751231 37.01 16.39 10.48 23.78 3.62 48.94 28.14 72.47 12.25
19761231 23.97 33.55 28.86 38.32 2.52 119.05 97.74 137.00 12.96
19771230 -7.42 -1.23 -4.72 1.51 1.68 -10.21 -19.27 -3.37 4.39
19781229 6.35 2.89 -0.38 5.70 1.35 0.04 -7.90 6.08 3.54
19791231 18.60 28.73 25.58 33.00 1.74 108.81 89.19 130.92 9.62
19801231 32.58 27.36 19.62 33.70 4.14 87.84 55.20 120.73 20.85
19811231 -4.72 -3.36 -9.56 6.85 4.99 -14.51 -30.39 13.94 13.58
19821231 21.95 24.89 15.66 29.71 3.59 57.47 27.59 79.68 13.04
19831230 22.36 25.03 23.48 27.84 1.50 75.75 62.82 87.80 7.29
19841231 6.33 0.36 -1.41 1.86 1.11 -7.76 -15.52 -2.19 3.59
19851231 31.99 34.60 31.28 36.32 1.47 143.07 134.30 149.38 4.36
19861231 18.26 20.13 17.11 23.68 1.88 60.42 51.18 81.04 8.38
19871231 5.06 -0.15 -7.10 6.58 3.65 -32.24 -43.39 -19.67 6.83
19881230 17.09 6.59 3.04 10.69 2.13 19.88 9.28 27.62 4.88
19891229 31.38 38.42 35.06 41.43 1.55 156.17 144.10 169.48 6.51
19901231 -3.26 -6.58 -13.76 1.37 4.51 -26.78 -43.82 -5.28 12.09
19911231 30.56 28.15 22.60 33.88 2.79 96.03 73.80 121.87 12.00
19921231 7.74 9.11 3.13 11.90 2.46 30.45 10.13 48.29 10.27
19931231 10.00 20.74 17.22 22.73 1.48 71.29 57.52 80.81 5.92
19941230 1.20 -2.46 -6.38 1.37 2.22 -15.13 -23.79 -4.32 5.06
19951229 37.58 43.27 41.46 44.93 0.98 184.19 172.31 191.57 4.90
19961231 23.03 17.55 13.85 21.86 2.13 51.09 35.26 76.06 11.96
19971231 33.50 27.42 21.26 32.33 3.22 91.78 67.26 114.63 14.22
19981231 29.16 23.86 17.68 31.26 3.73 66.08 36.83 89.76 14.43
19991231 20.82 35.79 22.90 52.81 7.51 110.97 76.81 180.57 34.51
20001229 -9.39 -10.30 -21.62 -2.63 5.31 -39.31 -55.34 -24.85 9.30
20011231 -11.61 3.50 3.50 3.50 0.00 3.50 3.50 3.50 0.00
20021231 -21.99 2.20 -0.43 4.36 1.14 3.36 -3.28 9.93 3.32
20031231 28.74 21.55 18.91 25.07 1.76 73.81 54.91 92.33 9.51
20041231 10.88 7.70 2.42 10.94 2.22 19.35 7.72 31.56 5.86
20051230 5.00 18.38 16.95 20.39 0.87 53.96 46.21 61.51 3.80
20061229 15.78 14.96 9.82 20.35 2.49 38.16 19.31 61.48 11.33
20071231 5.64 16.89 13.08 19.55 1.95 42.18 28.09 56.34 8.88
20081231 -36.64 0.60 -2.81 1.63 1.41 -0.90 -10.59 2.28 4.06
20091231 26.18 24.41 19.10 28.59 2.95 82.17 64.90 103.46 10.07
20101231 14.99 9.65 1.07 18.40 5.02 12.96 -8.73 34.61 12.41
20111230 1.74 -12.67 -15.83 -9.31 2.03 -35.25 -42.02 -28.46 4.28
20121231 16.20 14.06 11.68 17.38 1.73 45.29 34.28 61.52 6.94
20131231 32.38 34.41 29.93 38.25 2.01 136.79 113.87 151.71 10.04
20141231 13.67 12.18 9.95 14.12 1.12 34.27 26.33 41.10 3.93
20151231 1.31 -2.29 -7.07 2.56 2.51 -11.09 -23.80 -0.19 6.49
20161230 11.85 -2.08 -8.94 1.81 2.49 -10.91 -25.64 -0.18 7.02
20171229 22.04 19.31 15.67 22.72 2.03 70.48 54.67 84.96 8.65
20181231 -4.33 0.44 -3.76 5.67 2.95 -4.02 -18.28 14.21 8.75
20191231 31.33 12.47 8.75 15.92 1.78 33.42 19.95 46.69 7.72
20200918 4.50 -6.89 -15.33 0.61 3.78 -37.80 -57.04 -20.40 9.02

^S5T (S&P 500 Market Capitalization-Weighted Total Return index): http://gtr1.net/2013/?s19261231::!S5T
Top 3 of 12 by TR252 (1x): http://gtr1.net/2013/?s19261231::BULL:et1:trp%281,252%29tn3:...
Top 3 of 12 by TR252 (3x): http://gtr1.net/2013/?s19261231::BULL:et1:trp%281,252%29tn3:...


Robbie Geary
Print the post Back To Top
No. of Recommendations: 0
Hey Robbie,

Those results are amazing. Definitely suggests sector rotation beats the index, and more so if leveraged. The drawdowns aren't as bad as I would have guessed. Other than 1929, the 3X holds up remarkably well during downturns. Does that really say $9.6 trillion vs. <$1M total return??? On what initial investment? Certainly not for the faint of heart.
Print the post Back To Top
No. of Recommendations: 13
I'm saddened to hear about the spat with tpoto, since it was his posts that led me to the only MI screen I'm still using (sector rotation).

In a way, this just encapsulates the conundrum of mechanical investing. We are not mechanical, but emotional, psychological (dare I say spiritual) beings.

I understand how hard it is to be a traffic cop AND how hard it can be to get a ticket.

We'd all benefit by staying pristine with our respect for each other while severely strict with our analyses.

It's strange how this relates to my personal life today too. I caught my son stealing from me yesterday, then lying about it when I confronted him. I'm not saying that's what tpoto did (I don't know what tpoto did); but rather, suggesting this is the most extreme example of the balance. I still love him with all my heart, but I cannot tolerate that behavior, which I can now see has been an ongoing pattern. It breaks my heart and it's hard to know what to do. I don't want to call the police or banish him from my life, but the behavior must change, and how to make that happen is not easy to see. Therapists are definitely needed. My trust has been shattered and it will be difficult to rebuild... ever.
Print the post Back To Top
No. of Recommendations: 2
Those results are amazing.

I’d be very careful about interpreting those results. If I understand correctly, what Robbie did is not a realistic back test using investable funds, but rather a synthetic backtest using sectors. No corresponding funds have existed over the backtest period, so it would have been very difficult or rather impossible to replicate the result. With the 3x backtest the results are yet much more difficult to replicate because even when 3x funds exist they are notoriously poor at tracking 3x the index value over time. And the tracking error is always to the down side.

Robbie may wish to correct me if I’m wrong.

Elan
Print the post Back To Top
No. of Recommendations: 16
I'm saddened to hear about the spat with tpoto, since it was his posts that led me to the only MI screen I'm still using (sector rotation).

This is not the first time tpoto has threatened to take his marbles and leave. In fact, I seem to recall that he has been exceptionally thin skinned in the past. Robbie, on the other hand, has been harsh in his criticism many a time, and I’ve never known him to be factually wrong in his criticism of others’ mistakes. In this case tpoto has persisted in making a familiar methodological error and Robbie, a perfectionist if there ever was one, became understandably irate.

As I’ve watched this controversy unfold, I’ve thought about how fortunate we are that Robbie isn’t so thin skinned. Just imagine for a moment, if you will, what would happen if Robbie was as thin skinned as tpoto.

Elan
Print the post Back To Top
No. of Recommendations: 1
BAGoldman:
I wonder how it would work if leveraged ETFs were used instead of the 9 original unleveraged ETFs (all are 3x leverage except I could only find 2x for materials)?

Robbie:
If the Fama/French twelve sectors (https://mba.tuck.dartmouth.edu/pages/faculty/ken.french/ftp/...) are an acceptable substitute, then you can easily backtest both 1x and 3x versions of sector momentum (with the timing step) back to 1927. You can reduce the 12 sectors down to a smaller subset of sectors by eliminating any symbols you don't want (for example, it probably makes sense to eliminate F1212, "Other") from the custom universe code that looks like this (for 1x):


Question are the Fama/French twelve sectors Cap-weighted or Equal-weighted?

SSGA Funds Management, Inc, SSGA Funds Management, Inc / SPDR StrategicFactors equal-weighted would be the closest set of funds available (11 of 17) with History 15y.

GD_
Print the post Back To Top
No. of Recommendations: 1
Those results are amazing. Definitely suggests sector rotation beats the index, and more so if leveraged. The drawdowns aren't as bad as I would have guessed. Other than 1929, the 3X holds up remarkably well during downturns. Does that really say $9.6 trillion vs. <$1M total return??? On what initial investment? Certainly not for the faint of heart.

Yeah, that 96.5% MDD is certainly intimidating.

Tails
Print the post Back To Top
No. of Recommendations: 0
...even when 3x funds exist they are notoriously poor at tracking 3x the index value over time. And the tracking error is always to the down side.

Hmmm. TQQQ is a 3x fund that has been around a bit over ten years. Using Yahoo's historical numbers:

TQQQ 112.64/2.72 = 41.5
QQQ 263.86/52.18 = 5.1

DB2
Print the post Back To Top
No. of Recommendations: 4
Hmmm. TQQQ is a 3x fund that has been around a bit over ten years. Using Yahoo's historical numbers:


TQQQ 112.64/2.72 = 41.5
QQQ 263.86/52.18 = 5.1


Yup. So a 41.5x gain over 10 years is a CAGR of 45.1%. And a 5.1x return over 10 years is a CAGR of 17.7%. Thus less than 3x.

Elan
Print the post Back To Top
No. of Recommendations: 0
test
Print the post Back To Top
No. of Recommendations: 0
sector rotation bullet 19981222 with timing
vs. S&P and vs. SPY with timing

cagr 14.33 9.2598 11.5
at 8.61 0.529 .72

gprc(1)>0
tr(1,10) bottom 5
tr(1,300) top 2
tr(1,200) top 1
Print the post Back To Top
No. of Recommendations: 0
Print the post Back To Top
No. of Recommendations: 17
elann:
I’d be very careful about interpreting those results. If I understand correctly, what Robbie did is not a realistic back test using investable funds, but rather a synthetic backtest using sectors. No corresponding funds have existed over the backtest period, so it would have been very difficult or rather impossible to replicate the result... Robbie may wish to correct me if I’m wrong.


I too would be extremely cautious in drawing conclusions about trading real-world S&P Select Sector ETFs/futures from the backtests I provided, especially when it comes to 3x ETFs/futures trading.

In regards to the 1x GTR1 Fama/French sector indexes, I don't think replication is the problem. I.e., I think it's quite plausible that they could have been replicated within real portfolios all the way back to 1926 due to the weighting by market capitalization, which would have kept transaction and management costs low. Yes, the indexes are "syntehtic", but so is the S&P 500, which passive mutual funds have closely tracked at least as far back as 1980. I don't see why they couldn't have done so all the way back to 1926--all that was missing was the bright idea to do so. There were in fact a plethora of active funds (investment trusts) in the 1920s--so many, in fact, that they are often cited as a major cause of the 1929 stock market boom and bust.

The Fama/French sectors aren't really any different from the S&P 500 in this respect. If one wants to simulate fund expenses for them, then I would suggest comparing the CAGR of VFINX in the 1980s with the CAGR of ^S5T over the same period; whatever VFINX's shortfall is, assume the sector portfolios would have had the same drag, or perhaps a bit more, since they would not have been as concentrated in the mega-caps.

Instead, my main concerns are as follows:

1. The fact that assigning SIC codes, and industrial classification in general, is a very inexact science. (See https://boards.fool.com/gtr1-database-update-11214-31063732....) It's extremely rare for any two given sources of SIC codes (e.g., Dunn & Bradstreet, Thomson-Reuters, Morningstar, etc) to agree on all four digits for more than 50% of stocks. In fact, the same source can disagree with itself. There are three different ways to access SIC codes from Morningstar, and they all differ!

2. Related to (1), I'm far from happy with my own SIC-assigning algorithm. It's been trained to match the SIC codes assigned to stocks in the historical database at the end of 2013. The algorithm is automatically used to assign SIC codes to all new stocks, as well as to any old stocks whose SIC codes are "unlocked" from their 2013 values due to certain corporate actions. I've tested it out-of-sample on the 2017 edition of the historical database; I don't recall what the 4-digit match rate was, but I doubt that it was a lot higher than 50%.

3. Whether the SIC codes in the historical database are truly point-in-time, that is, free of all crystal ball contamination. (Surprisingly, the documentation for the historical database says very little about this. I think the data is fairly safe to use, only because it is one of the main sources in academia.)

4. Whether Fama/French's SIC-based sectorization is remotely similar enough to GICS-based sectorization (as used by the S&P Select Sector Indexes that major ETFs and futures are based on) to enable conclusions about trading systems based on one to carry over to those based on the other.

5. In the case of 3x ETFs, the fact that ETF issuers have demonstrated a tendency to reduce leverage down to 2x or 1.5x exactly when traders need the leverage most, i.e, when recovering losses during sharp declines. (We saw this with volatility-related products during "Volmageddon" in February 2018, and again with certain commodity-related ETFs, e.g., GUSH/DRIP, NUGT/DUST, etc, during the COVID-19 panic in 2020. SOXL/SOXS also came very close to having their leverage reduced in 2020. If TQQQ/SQQQ had been invented in 1999, I think there is no doubt that the leverage on these ETFs would have been reduced at one of the many troughs during the subsequent bear market. Likewise for FAS/FAZ during the 2007-2009 bear market.)

Because of (5), I would not go anywhere near 3x sector ETFs in accounts where you could not also trade the futures with the same 3x exposure as back-up (or even as the primary instruments), keeping in mind that margin requirements on futures are almost guaranteed to be ratcheted up at the exact moment when you most desperately need the leverage.

To elaborate on (2) a little more, consider GOOG and GOOGL. Prior to the Alphabet/Google restructuring in 2015, both of these stocks had SIC code 7375 (Information Retrieval Services) inherited from the historical database, but the restructuring "unlocked" their SIC codes. My algorithm determined that 7375 was the best fit for GOOGL but assigned 8999 (miscellaneous services that include such specializations as Data Retrieval and Information Management, but also Poetry and Ice Carving!). I'm not really happy with either of those designations (Google is much more of an information gatherer than an information retriever, but more than anything else, it's still an advertising services company, going by the source of most of its revenue). The point of the algorithm isn't to be "right", whatever that would mean, but to be consistent with the methodology of the historical database. It succeeded with GOOGL but failed with GOOG. Furthermore, it's simply nonsensical for two classes of common stock representing equity in the same business to have different SIC codes; while that only happens with a small handful of companies, it's something I need to fix.


With the 3x backtest the results are yet much more difficult to replicate because even when 3x funds exist they are notoriously poor at tracking 3x the index value over time. And the tracking error is always to the down side.


This is a major misconception. The 3x ETFs, especially those linked to the major indexes (TQQQ, TNA, UPRO, UDOW) are quite successful at behaving according to prospectus, that is, returning three times the daily return of the 1x ETFs (QQQ, IWM, SPY and DIA). There are always minor discrepancies at the close each day, but these are usually reversed the next day. There are a few rare exceptions, however, and the biggest exceptions on record occurred during COVID-19 panic this year, when the 3x ETFs deviated quite noticeably from three times the daily returns of the 1x ETFs on a few of the worst days.

What most people call the 3x ETF's "tracking error" isn't that at all; rather, it's just their own bad mathematical intuition, or ignorance about the path-dependence of long-term 3x ETF returns.


Yup. So a 41.5x gain over 10 years is a CAGR of 45.1%. And a 5.1x return over 10 years is a CAGR of 17.7%. Thus less than 3x.


This is totally the wrong way to evaluate a 3x ETF's performance versus its underlying 1x ETF. There is absolutely no mathematical reason to expect that compounding the tripled daily returns of QQQ would result in three times the CAGR of QQQ over any period, except one market day. Over longer periods than one market day, its entirely possible, in fact, for the 1x ETF to have a positive CAGR while the properly-behaving 3x ETF has a negative CAGR. And even vice versa!

To see just how well TQQQ conforms to prospectus, just compare TQQQ (http://gtr1.net/2013/?TQQQ) with trippling the daily returns of QQQ (http://gtr1.net/2013/?!!QlpoMTFBWSZTWQcjUZYAAYVfgCAAAgd48AhE...) over the same period.


Robbie Geary
Print the post Back To Top
No. of Recommendations: 5
rgearyiii:
There were in fact a plethora of active funds (investment trusts) in the 1920s--so many, in fact, that they are often cited as a major cause of the 1929 stock market boom and bust.

I meant to say "active leveraged funds (investment trusts) in the 1920s...".

Also, regarding the 1x versus 3x backtests, notice that while the 3x backtest produces a much higher CAGR and even a higher Sharpe(20) than the 1x backtest, the SAWR is actually lower for 3x. This implies that for a retired person, the 3x option makes zero sense. I would imagine that this is a conclusion that you can carry over from Fama/French sector backtests to S&P Select Sectors.

Robbie Geary
Print the post Back To Top
No. of Recommendations: 10
Yup. So a 41.5x gain over 10 years is a CAGR of 45.1%. And a 5.1x return over 10 years is a CAGR of 17.7%. Thus less than 3x.

Technically, a 10-year performance of "only" 2.5x long-term leverage supports your thesis. But really, it highlights how little long-term divergence there has been.

The doomsayers' predictions are generally along the lines of daily leverage being an inevitable trip down the toilet.

- Jamie
Print the post Back To Top
No. of Recommendations: 5
how fortunate we are that Robbie isn’t so thin skinned

I remember when it used to be "long-toed" on this board... :)

That one required an explanation back in the day.

Draggon
Print the post Back To Top
No. of Recommendations: 1
Except that the leverage problem / return problem is to the downside, not the upside. As has been written about extensively.

Roughly, using monthly prices up through August, last 5 years, QQQ is up 2.718. *3 is 8.154.
TQQQ is up 8.11 over same time frame - .04 difference, not really significant.

The inverse of 8.154 (the 3x bullish ETF return) is .122.
But SQQQ (the 3x inverse ETF) is .0148 the price of 5 years ago - a 99.52% drop - and, a price of $21.35 vs the $175 it would be if it tracked the inverse of the 3x return.

I don't even pretend to understand the mathematics behind this problem. Maybe it's the way the ETFs are structured.
Print the post Back To Top
No. of Recommendations: 6
Another overlooked fact about leveraged ETFs is that their leverage actually changes throughout the trading day.

For example, suppose that at some point during the day, QQQ is down -10% from its previous close. If TQQQ is behaving correctly (meaning arbitragers are doing their job), TQQQ will be down exactly -30% from its previous close. Suppose you buy TQQQ at that level and then QQQ returns to break-even for the day (a 0% return). Then TQQQ should also return to break-even. Assuming it does, then your return on TQQQ for the day is 1/0.7 ~= 42.86%, while the return on QQQ from the point where you bought TQQQ is 1/0.9 ~= 11.11%. So your effective leverage was 42.86/11.11 ~= 3.86.

Of course, once the market closes, your leverage is reset to 3 (give or a take a tiny bit, depending on how far both QQQ and TQQQ closed from their NAVs).

Robbie Geary
Print the post Back To Top
No. of Recommendations: 5
The inverse of 8.154 (the 3x bullish ETF return) is .122.
But SQQQ (the 3x inverse ETF) is .0148 the price of 5 years ago - a 99.52% drop - and, a price of $21.35 vs the $175 it would be if it tracked the inverse of the 3x return.

I don't even pretend to understand the mathematics behind this problem. Maybe it's the way the ETFs are structured.


The math is simple, once you understand the fact that the leverage is that of the DAILY return. Coupled with the bit that is also the name of a financial web site: Asymmetrical Returns.

A 10% drop is larger than a 10% gain. A 10% loss hurts you more than a 10% gain helps you. 3X leverage makes this even more extreme.

Consider if the investment loses 33% in a day. 3X that is 100% loss. To make up a 100% loss is not possible, it would take an infinite percent gain. That's the extreme, but shows you the problem.

Consider a 10% loss. 3X that is a 30% loss. It takes a 42% gain to get back to even. That's 3X of a 14% gain. One-day 10% losses are MUCH more frequent than one-day 14% gains.
Print the post Back To Top
No. of Recommendations: 0
It's strange how this relates to my personal life today too. I caught my son stealing from me yesterday, then lying about it when I confronted him. I'm not saying that's what tpoto did (I don't know what tpoto did); but rather, suggesting this is the most extreme example of the balance. I still love him with all my heart, but I cannot tolerate that behavior, which I can now see has been an ongoing pattern. It breaks my heart and it's hard to know what to do. I don't want to call the police or banish him from my life, but the behavior must change, and how to make that happen is not easy to see. Therapists are definitely needed. My trust has been shattered and it will be difficult to rebuild... ever.

How old is your son?
Print the post Back To Top