DRAFT Data Dredging Momentum Strategies

W.G.Paseman
04/30/21 15:54:25 PDT-0700
Prior Work

In July of 2020 I presented momentum results from Momentum202007. Momentum202007 ran a modified version of the Jegadeesh/Titman/Rowenhorst momentum algorithm. This algorithm computes btmN, topN, topNx (topN - ^IRX), and timedTopNx on portfolios containing 1-6 members on the NASDAQ, DOW, DAX, S&P 100 and Russell 1000 indices. Unlike the original papers which used days, Momentum202007 used months (as per Don Maurer). Momentum202007's price database comes from 22 years of Yahoo's price adjusted data. Momentum202007 did not take into account the evolution of index composition through time and so was subject to survivorship bias. Momentum202007 determined that the result of a strategy was significant if, when compared to its benchmark, it had an independent t-test > 3.64. Among other results, Momentum202007 showed significant results over 20 years for momentum portfolios utilizing lags of 01_03_06_12mos containing 6 members (Fundx) and a lag of 11mos containing 3 members.

Current Approach

Momentum202104 is similar to Momentum202007, but accounts for the evolution of index composition for the universes shown below (Thanks to Don Maurer fo allowing me to run Momentum202104 on his Norgate Machine).

UniverseName Benchmark NorgateKey UniverseStartDate IdxStartDate Price
SPx500 $SPX S&P 500 19500103 19280131 17.57
DOW30 $DJI Dow Jones Industrial Average 19500103 18960529 40.63
SPx100 $OEX S&P 100 19820730 19820730 54.95
RUx1000 $RUI Russell 1000 19870629 19870629 159.84
NDx100 $NDX NASDAQ 100 19850930 19850930 112.14

Momentum202007 runs 1351 lag strategies formed by taking Harris (1,3,6,6) plus all 1,2 and 3 length combinations of the Portfolio Visualizer months.
months=[1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 15, 18, 21, 24, 27, 30, 33, 36]
It then generates a data cube of 777,600 cases for the cross product of 6 universes, 2 methods (topN, btmN), 6 memberCounts (1-6), 1350 lag strategies and timeframes coving (5,10,20...70 years).

Results

Momentum202104 is far enough along that I have some results.
Items of Note:

Future work

This document is a DRAFT. These strategies were discovered, not proposed and then checked post hoc. This is the very definition of p-hacking or (Data Dredging). In order to validate the results, I need to add topNx and timedTopNx methods and compare a run of Momentum202104 with Yahoo data to a Momentum202007 with the same data

Bibliography

Figures

01_03_06_12mos_FundX_01_03_06_06mos_Harris_06_10_11mos_03_06_12mos_03_12mos_02_10_12mos_07_10mos_06_08_09mos_03_11mos_04_08_09mos_03_10_12mos_06_09_11mos_04_09_11mos_03_07_11mos_07_08_11mos_06_09_12mos_01mos_02mos_03mos_04mos_05mos_06mos_07mos_08mos_09mos_10mos_11mos_12mos

Monthly GeoRtn/Std (Risk) by Universe for t>3.65 significance=99.95 - 11 rows

( Interactive Figure)


GeoRtn Sharpe
timeframe 30 yrs 35 yrs 70 yrs 30 yrs 35 yrs 70 yrs
universe method lags memberCount
NDx100 topN 03_06_12mos 3 0.031969 6.845456
03_11mos 4 0.030371 6.973197
11mos 1 0.037423 0.036179 5.708621 5.966486
SPx500 topN 03_10_12mos 5 0.015457 6.820589
04_09_11mos 6 0.015375 7.067429
06_09_11mos 5 0.015485 6.832753
07_08_11mos 4 0.015554 6.660135
10mos 4 0.015424 6.548809
12mos 1 0.016459 5.357520
2 0.016225 5.893879

Monthly GeoRtn/Std (Risk) by Universe for t>3.28 significance=99.95 - 94 rows

( Interactive Figure)


GeoRtn Sharpe
timeframe 30 yrs 35 yrs 60 yrs 65 yrs 70 yrs 30 yrs 35 yrs 60 yrs 65 yrs 70 yrs
universe method lags memberCount
NDx100 topN 01_03_06_12mos_FundX 3 0.030327 6.566831
4 0.028627 6.606154
5 0.028607 6.791325
03_06_12mos 2 0.029815 6.061332
3 0.030437 0.031969 6.170859 6.845456
4 0.029413 6.775020
5 0.028733 6.808961
03_07_11mos 2 0.029702 6.133242
03_10_12mos 4 0.028899 6.495292
03_11mos 2 0.029046 6.036762
3 0.029770 6.494589
4 0.030371 6.973197
03_12mos 2 0.030167 6.156851
3 0.030560 6.374765
4 0.028872 6.521374
5 0.028205 6.586765
6 0.027865 6.715320
06_09_12mos 4 0.028759 6.604533
11mos 1 0.037423 0.036179 5.708621 5.966486
2 0.030134 6.064248
12mos 1 0.029047 5.149817
2 0.028654 5.858606
SPx500 topN 02_10_12mos 4 0.014389 6.219528
03_07_11mos 4 0.014336 6.147665
5 0.014240 6.445464
03_10_12mos 2 0.015136 5.660785
3 0.014644 5.983555
4 0.014166 0.014940 5.905552 6.412043
5 0.014618 0.015457 6.247855 6.820589
6 0.014000 0.014769 6.207951 6.783519
04_08_09mos 3 0.014565 5.981010
5 0.014148 6.468007
04_09_11mos 3 0.014634 6.004964
4 0.014273 0.014891 5.992959 6.458921
5 0.014053 0.014422 6.083175 6.498836
6 0.014676 0.014990 0.015375 6.181890 6.592276 7.067429
06_09_11mos 3 0.014524 5.939655
4 0.014771 0.015092 6.093870 6.472815
5 0.015170 0.015485 6.417688 6.832753
06_09_12mos 3 0.014514 5.900386
4 0.014465 0.015111 5.954913 6.427521
5 0.014219 0.014713 6.074011 6.526663
6 0.014215 0.014657 6.227642 6.691004
06_10_11mos 3 0.014889 6.098744
4 0.014293 0.014946 5.918002 6.408080
5 0.014800 0.015181 6.280645 6.702922
6 0.014098 6.463914
07_08_11mos 3 0.015006 0.015583 5.865755 6.283813
4 0.015202 0.015554 6.257771 6.660135
5 0.014227 6.352155
6 0.014788 0.015058 6.481259 6.887904
07_10mos 4 0.014525 0.015130 6.113574 6.599018
5 0.014299 6.420092
6 0.014436 6.697504
08mos 3 0.014476 5.927464
09mos 2 0.014662 0.014888 5.309838 5.571811
10mos 2 0.015184 5.646737
3 0.015004 6.110439
4 0.014533 0.015424 5.983242 6.548809
5 0.014872 6.545254
11mos 1 0.015037 5.005444
2 0.014761 5.526804
4 0.014297 0.014893 5.797317 6.263876
5 0.014289 6.316538
6 0.014392 6.509692
12mos 1 0.016438 0.015895 0.016459 4.946032 5.067434 5.357520
2 0.015538 0.016225 5.506118 5.893879
4 0.014918 6.179889

Monthly GeoRtn/Std (Risk) by Universe for t>2.81 significance=99.75 - 392 rows

( Interactive Figure)


Monthly GeoRtn/Std (Risk) by Universe for t>2.58 significance=99.50 - 613 rows

( Interactive Figure)