WEBVTT

1
00:00:00.020 --> 00:00:00.200
Hello,

2
00:00:00.320 --> 00:00:02.441
welcome back to Papers with Backtest podcast.

3
00:00:02.703 --> 00:00:05.766
Today we dive into another algo trading research paper.

4
00:00:06.227 --> 00:00:10.352
And today we're tackling something called absolute strength momentum.

5
00:00:10.750 --> 00:00:11.008
Right.

6
00:00:11.375 --> 00:00:12.516
It's an interesting concept.

7
00:00:12.891 --> 00:00:13.391
Basically,

8
00:00:13.570 --> 00:00:17.781
the idea is that stocks that have really moved a lot recently,

9
00:00:18.438 --> 00:00:19.297
up or down.

10
00:00:19.438 --> 00:00:20.047
Big moves.

11
00:00:20.312 --> 00:00:22.125
They tend to keep going in that same direction,

12
00:00:22.126 --> 00:00:23.203
at least for a little while.

13
00:00:23.297 --> 00:00:23.750
Exactly.

14
00:00:24.156 --> 00:00:26.906
And it's important right off the bat to separate this from.

15
00:00:27.420 --> 00:00:28.701
Relative strength momentum,

16
00:00:28.741 --> 00:00:30.264
which I think more people are familiar with.

17
00:00:30.403 --> 00:00:30.543
Yeah,

18
00:00:30.662 --> 00:00:35.248
relative strength is about comparing stock A to stock B or to the market.

19
00:00:35.448 --> 00:00:35.627
Uh-huh.

20
00:00:35.768 --> 00:00:37.393
How's it doing compared to its peers?

21
00:00:37.729 --> 00:00:39.112
But absolute strength,

22
00:00:39.190 --> 00:00:40.776
that's just looking at the stock itself,

23
00:00:40.893 --> 00:00:41.651
its own history.

24
00:00:42.135 --> 00:00:44.573
Has it gone up a lot on its own or down a lot?

25
00:00:44.760 --> 00:00:45.276
Precisely.

26
00:00:45.479 --> 00:00:45.604
And,

27
00:00:45.682 --> 00:00:45.916
you know,

28
00:00:46.135 --> 00:00:48.729
it's not always obvious if the profits from,

29
00:00:48.901 --> 00:00:49.135
say,

30
00:00:49.682 --> 00:00:55.073
a standard relative momentum strategy are really because of that relative outperformance.

31
00:00:55.088 --> 00:00:56.041
Or if it's actually these...

32
00:00:56.136 --> 00:00:58.858
big absolute moves doing the heavy listing underneath?

33
00:00:59.219 --> 00:01:01.362
That's the question this research tries to unpack.

34
00:01:01.741 --> 00:01:02.783
So that's what we're digging into,

35
00:01:03.041 --> 00:01:07.487
a specific trading rule based purely on this absolute strength idea.

36
00:01:07.729 --> 00:01:09.151
It sounds simple on the surface.

37
00:01:09.432 --> 00:01:10.346
Buy the big winners,

38
00:01:10.690 --> 00:01:11.713
sell the big losers.

39
00:01:11.791 --> 00:01:12.049
Right.

40
00:01:12.127 --> 00:01:13.440
Buy stocks that have shot up,

41
00:01:13.799 --> 00:01:15.018
short stocks that have cratered.

42
00:01:15.799 --> 00:01:16.784
Seems logical enough.

43
00:01:17.315 --> 00:01:21.362
But the nuance is in defining big.

44
00:01:21.815 --> 00:01:23.143
What's a significant move?

45
00:01:23.288 --> 00:01:24.089
And that's key here,

46
00:01:24.090 --> 00:01:24.309
isn't it?

47
00:01:24.310 --> 00:01:25.670
They don't just pick a number like,

48
00:01:25.711 --> 00:01:25.910
say,

49
00:01:26.070 --> 00:01:26.832
up 50 percent.

50
00:01:26.893 --> 00:01:27.012
No,

51
00:01:27.152 --> 00:01:28.033
not arbitrary at all.

52
00:01:28.072 --> 00:01:33.701
They determine it endogenously based on a long history of actual return data.

53
00:01:33.936 --> 00:01:34.061
OK,

54
00:01:34.178 --> 00:01:38.764
so they look back decades to see what a historically large move really looks like for stocks.

55
00:01:38.865 --> 00:01:39.389
Exactly.

56
00:01:39.451 --> 00:01:41.920
And that's crucial because it avoids look ahead bias.

57
00:01:42.264 --> 00:01:44.951
You're not using future info to define past extremes.

58
00:01:45.029 --> 00:01:45.467
Makes sense.

59
00:01:45.529 --> 00:01:47.811
Can't use tomorrow's data to trade today.

60
00:01:48.092 --> 00:01:52.076
So our mission here is to really get the rules down for this absolute strength strategy.

61
00:01:52.220 --> 00:01:53.802
Understand how it works step by step.

62
00:01:53.922 --> 00:01:56.543
And then critically look at the backtest results.

63
00:01:57.043 --> 00:01:57.906
Did it actually work?

64
00:01:58.047 --> 00:01:59.406
Let's dive into those details.

65
00:02:00.090 --> 00:02:00.410
First,

66
00:02:00.449 --> 00:02:01.472
the data they used.

67
00:02:01.746 --> 00:02:03.410
It's pretty standard stuff for U.S.

68
00:02:03.488 --> 00:02:04.269
equity research.

69
00:02:04.730 --> 00:02:06.676
Common stocks from NYSE,

70
00:02:07.277 --> 00:02:07.574
Amex,

71
00:02:07.715 --> 00:02:10.504
and Nasdaq from the CRSP database.

72
00:02:10.832 --> 00:02:11.644
And the time period.

73
00:02:11.754 --> 00:02:15.269
The main analysis covers 1965 through 2016.

74
00:02:15.347 --> 00:02:16.566
So a good long stretch.

75
00:02:16.910 --> 00:02:17.910
Over 50 years.

76
00:02:18.066 --> 00:02:18.285
Yeah.

77
00:02:19.051 --> 00:02:19.347
But,

78
00:02:19.707 --> 00:02:21.785
and this is important for defining those big moves,

79
00:02:22.272 --> 00:02:27.156
They used data going all the way back to 1927 to build the historical context.

80
00:02:27.258 --> 00:02:27.496
Wow.

81
00:02:27.598 --> 00:02:27.719
OK.

82
00:02:28.258 --> 00:02:32.781
Nearly a century of data for the baseline and the strategy itself.

83
00:02:32.922 --> 00:02:33.703
What are the mechanics?

84
00:02:33.820 --> 00:02:36.664
They focused on what they call the 11-1-1 strategy.

85
00:02:36.789 --> 00:02:37.430
11-1-1.

86
00:02:37.469 --> 00:02:37.586
OK.

87
00:02:37.891 --> 00:02:38.375
Break that down.

88
00:02:38.563 --> 00:02:41.906
So each month you look back at a stock's return over the past year,

89
00:02:42.094 --> 00:02:43.328
but not quite the whole year.

90
00:02:43.719 --> 00:02:46.172
You look from month T12 up to month T2.

91
00:02:46.953 --> 00:02:48.500
So the last 11 months,

92
00:02:48.563 --> 00:02:49.766
but skipping the most recent month,

93
00:02:49.938 --> 00:02:50.063
T1.

94
00:02:50.359 --> 00:02:50.875
Exactly.

95
00:02:51.084 --> 00:02:53.686
That skip of the last month is pretty common in momentum research.

96
00:02:53.827 --> 00:02:55.790
Helps avoid short-term reversal effects.

97
00:02:55.868 --> 00:02:56.110
Right.

98
00:02:56.149 --> 00:02:57.629
So you rank based on that

99
00:02:58.149 --> 00:02:58.790
11-month return.

100
00:02:59.196 --> 00:02:59.993
What gets excluded?

101
00:03:00.571 --> 00:03:01.516
Low-priced stocks,

102
00:03:01.735 --> 00:03:02.774
typically below a dollar.

103
00:03:03.274 --> 00:03:05.797
And stocks that didn't have enough return data during that

104
00:03:06.157 --> 00:03:06.657
11-month window,

105
00:03:06.821 --> 00:03:08.602
they required at least eight observations.

106
00:03:08.665 --> 00:03:08.962
Got it.

107
00:03:09.243 --> 00:03:10.040
Standard filters,

108
00:03:10.258 --> 00:03:11.212
and you form the portfolio.

109
00:03:11.602 --> 00:03:11.837
Yes.

110
00:03:12.290 --> 00:03:13.399
Based on that ranking,

111
00:03:13.446 --> 00:03:15.680
you form your winner and loser portfolios,

112
00:03:15.727 --> 00:03:17.227
and you hold them for a month.

113
00:03:17.743 --> 00:03:18.524
Just one month.

114
00:03:18.587 --> 00:03:20.399
And then rebalance and do it all again next month.

115
00:03:20.548 --> 00:03:21.309
Rinse and repeat.

116
00:03:21.449 --> 00:03:21.630
Okay,

117
00:03:21.631 --> 00:03:22.930
so that's the 11-month rank,

118
00:03:23.311 --> 00:03:23.991
one-month skip,

119
00:03:24.553 --> 00:03:25.253
one-month hold.

120
00:03:25.495 --> 00:03:25.632
Now,

121
00:03:25.753 --> 00:03:26.655
the crucial part,

122
00:03:27.514 --> 00:03:31.237
defining the absolute winners and losers from that rank,

123
00:03:31.440 --> 00:03:32.120
how they do that.

124
00:03:32.440 --> 00:03:35.940
This is where that long 1927 onwards history comes in.

125
00:03:36.620 --> 00:03:37.448
They take a stock's

126
00:03:37.948 --> 00:03:38.542
11-month return.

127
00:03:38.667 --> 00:03:40.386
The T12 to T2 return.

128
00:03:40.714 --> 00:03:40.995
Yes.

129
00:03:41.401 --> 00:03:43.745
And they compare it not just to other stocks right now,

130
00:03:43.776 --> 00:03:47.542
but to the entire historical distribution of all non-overlapping

131
00:03:48.073 --> 00:03:49.354
11-month returns they have on record.

132
00:03:49.888 --> 00:03:54.393
So comparing today's moves against like the entire history of 11 month moves.

133
00:03:54.472 --> 00:03:54.972
Exactly.

134
00:03:55.073 --> 00:03:59.038
Think of it like a giant historical database of 11 month stock performance chunks.

135
00:03:59.460 --> 00:04:02.343
Where does the stock's recent performance fall in that grand scheme?

136
00:04:02.499 --> 00:04:03.280
And winners are.

137
00:04:03.444 --> 00:04:07.585
Winners are the stocks whose recent 11 month return puts them in the top 10%

138
00:04:07.586 --> 00:04:08.944
of that historical distribution.

139
00:04:09.351 --> 00:04:11.522
Truly exceptional performance compared to history.

140
00:04:11.632 --> 00:04:13.835
And losers are the bottom 10%

141
00:04:13.913 --> 00:04:15.194
of that historical distribution.

142
00:04:15.556 --> 00:04:15.896
Correct.

143
00:04:16.076 --> 00:04:20.739
Stocks that have performed exceptionally poorly relative to what's happened historically.

144
00:04:20.899 --> 00:04:27.360
That feels much more grounded than just picking the top 10 percent of this month's performers,

145
00:04:27.399 --> 00:04:27.782
doesn't it?

146
00:04:27.875 --> 00:04:28.258
It does.

147
00:04:28.282 --> 00:04:32.766
It anchors the definition of winner and loser to long term market behavior.

148
00:04:33.516 --> 00:04:34.547
And interestingly,

149
00:04:34.625 --> 00:04:38.422
they found these historical breakpoints were quite stable.

150
00:04:38.485 --> 00:04:38.828
table

151
00:04:38.984 --> 00:04:39.284
How so?

152
00:04:39.465 --> 00:04:39.605
Well,

153
00:04:39.625 --> 00:04:40.085
in the U.S.

154
00:04:40.145 --> 00:04:40.425
market,

155
00:04:40.426 --> 00:04:44.671
the cutoff for being an absolute winner over 11 months averaged around a 64%

156
00:04:45.049 --> 00:04:45.331
gain.

157
00:04:45.549 --> 00:04:45.772
Wow.

158
00:04:46.272 --> 00:04:46.952
And for losers,

159
00:04:46.953 --> 00:04:48.296
it was around a 43%

160
00:04:48.452 --> 00:04:48.796
loss.

161
00:04:49.194 --> 00:04:51.217
These numbers didn't wildly swing around.

162
00:04:51.397 --> 00:04:52.553
Unlike relative strength,

163
00:04:52.796 --> 00:04:53.678
where the top 10%

164
00:04:53.741 --> 00:04:56.038
performer might actually have a negative return in a bad market,

165
00:04:56.085 --> 00:04:56.225
right?

166
00:04:56.226 --> 00:04:56.616
Precisely.

167
00:04:56.975 --> 00:04:58.241
Or the bottom 10%

168
00:04:58.350 --> 00:05:01.210
relative performer might still be up in a roaring bull market.

169
00:05:01.428 --> 00:05:03.303
Absolute strength avoids that ambiguity.

170
00:05:03.632 --> 00:05:03.803
Okay,

171
00:05:03.835 --> 00:05:05.272
so the definition seems robust.

172
00:05:05.428 --> 00:05:05.616
Now,

173
00:05:05.866 --> 00:05:06.553
the big question,

174
00:05:07.178 --> 00:05:08.007
did the strategy...

175
00:05:08.300 --> 00:05:11.203
buying these historical winners and selling historical losers.

176
00:05:11.482 --> 00:05:12.303
They call it Abs Mom.

177
00:05:12.365 --> 00:05:12.646
Yes,

178
00:05:12.744 --> 00:05:13.064
Abs Mom.

179
00:05:13.103 --> 00:05:15.025
Did Abs Mom actually make money at the back test?

180
00:05:15.287 --> 00:05:15.646
It did,

181
00:05:15.849 --> 00:05:16.592
quite significantly,

182
00:05:16.631 --> 00:05:16.928
actually.

183
00:05:17.310 --> 00:05:20.553
Over that May 1965 to 2016 period,

184
00:05:20.896 --> 00:05:24.834
the average monthly risk-adjusted return was 2.50%.

185
00:05:25.131 --> 00:05:26.224
Two and a half percent per month,

186
00:05:26.412 --> 00:05:27.006
risk-adjusted.

187
00:05:27.193 --> 00:05:27.334
Yep.

188
00:05:27.537 --> 00:05:29.834
And the monthly Sharpe ratio was 0.34.

189
00:05:30.068 --> 00:05:30.959
Pretty solid numbers.

190
00:05:30.960 --> 00:05:31.943
That's really strong.

191
00:05:32.318 --> 00:05:33.318
And was it consistent?

192
00:05:34.279 --> 00:05:36.061
Or just driven by a few good years.

193
00:05:36.202 --> 00:05:37.042
Very consistent.

194
00:05:37.302 --> 00:05:39.546
They specifically highlighted the more recent period,

195
00:05:39.663 --> 00:05:41.007
2000 to 2016.

196
00:05:41.128 --> 00:05:43.971
Which includes the dot-com bust and the global financial crisis.

197
00:05:44.065 --> 00:05:44.612
Exactly.

198
00:05:44.768 --> 00:05:46.307
Tough times for many strategies.

199
00:05:46.675 --> 00:05:49.378
But Abs Mom still pulled a 1.86%

200
00:05:49.379 --> 00:05:50.768
risk-adjusted return per month,

201
00:05:51.112 --> 00:05:52.690
with a sharp 0.20%

202
00:05:52.878 --> 00:05:53.362
even then.

203
00:05:53.378 --> 00:05:54.862
That suggests real persistence.

204
00:05:54.878 --> 00:05:57.081
It wasn't just a fluke of the earlier decades.

205
00:05:57.143 --> 00:05:57.378
Right.

206
00:05:57.628 --> 00:06:02.440
It weathered some major storms and still showed profitability based on this absolute momentum concept.

207
00:06:02.612 --> 00:06:02.721
Now,

208
00:06:02.722 --> 00:06:03.971
they also mentioned something about...

209
00:06:05.132 --> 00:06:05.252
The

210
00:06:05.592 --> 00:06:07.093
Kvigorov-Smirnov test,

211
00:06:07.214 --> 00:06:07.513
KSD,

212
00:06:07.654 --> 00:06:08.275
what was that about?

213
00:06:08.615 --> 00:06:08.755
Ah,

214
00:06:08.876 --> 00:06:09.036
yes,

215
00:06:09.037 --> 00:06:09.978
the KSD test.

216
00:06:10.017 --> 00:06:11.997
That was for a conditional version of the strategy.

217
00:06:12.462 --> 00:06:12.900
Basically,

218
00:06:12.939 --> 00:06:16.079
the KSD test measures how different the distribution of recent

219
00:06:16.423 --> 00:06:19.322
11-month returns looks compared to the historical distribution.

220
00:06:19.728 --> 00:06:21.072
So not just individual stocks,

221
00:06:21.103 --> 00:06:24.072
but the overall shape of recent returns versus history.

222
00:06:24.244 --> 00:06:24.790
Exactly.

223
00:06:25.275 --> 00:06:30.103
Sometimes the whole market might be skewed towards extreme losses or extreme gains.

224
00:06:30.623 --> 00:06:32.645
The KSD flags those periods.

225
00:06:32.725 --> 00:06:35.168
And what do they do when the KSD flags such a period?

226
00:06:35.289 --> 00:06:38.410
When the KSD measure was very high in its top quintile,

227
00:06:38.473 --> 00:06:41.051
meaning recent returns looked very different from history,

228
00:06:41.691 --> 00:06:43.918
they suggested maybe switching the strategy off.

229
00:06:44.277 --> 00:06:44.996
Switching it off,

230
00:06:45.137 --> 00:06:45.434
meaning?

231
00:06:45.777 --> 00:06:48.184
Holding the risk-free asset instead.

232
00:06:48.887 --> 00:06:51.621
The idea is that in these highly unusual periods,

233
00:06:51.965 --> 00:06:55.793
the balance between winners and losers might be severely distorted.

234
00:06:55.949 --> 00:06:59.043
Like you might have tons of losers and almost no winners.

235
00:06:59.247 --> 00:06:59.947
or vice versa.

236
00:07:00.047 --> 00:07:00.567
Precisely,

237
00:07:00.647 --> 00:07:01.548
they give examples.

238
00:07:01.828 --> 00:07:03.707
End of February 2009,

239
00:07:03.708 --> 00:07:04.570
deep in the crisis.

240
00:07:05.167 --> 00:07:07.089
Far more absolute losers than winners.

241
00:07:07.507 --> 00:07:08.648
The market was just crushed.

242
00:07:08.726 --> 00:07:09.007
Looks so.

243
00:07:09.273 --> 00:07:09.750
Conversely,

244
00:07:09.929 --> 00:07:11.312
end of February 2004,

245
00:07:11.313 --> 00:07:12.570
lots more winners than losers.

246
00:07:13.132 --> 00:07:20.250
So the conditional strategy only ran the long-short absmom when the market's return distribution looked somewhat normal relative to history.

247
00:07:20.453 --> 00:07:22.359
Did that conditional approach perform differently?

248
00:07:22.687 --> 00:07:24.765
The paper suggests it helps manage risk,

249
00:07:25.187 --> 00:07:27.500
particularly around potential imbalances.

250
00:07:27.999 --> 00:07:31.321
But the core profitability comes from the Abs Mom signal itself.

251
00:07:31.642 --> 00:07:35.185
They also showed robustness by requiring a minimum number of stocks,

252
00:07:35.364 --> 00:07:36.067
like thirder,

253
00:07:36.325 --> 00:07:37.106
in both legs.

254
00:07:37.325 --> 00:07:37.489
Right.

255
00:07:37.505 --> 00:07:40.067
So it wasn't driven by just a few outliers.

256
00:07:40.192 --> 00:07:40.552
Correct.

257
00:07:40.810 --> 00:07:41.856
Even with that constraint,

258
00:07:41.950 --> 00:07:44.294
Abs Mom outperformed relative momentum.

259
00:07:44.356 --> 00:07:45.481
And this wasn't just a U.S.

260
00:07:45.544 --> 00:07:46.294
stock phenomenon,

261
00:07:46.356 --> 00:07:46.653
was it?

262
00:07:46.654 --> 00:07:47.700
They tested it elsewhere.

263
00:07:47.731 --> 00:07:48.044
They did.

264
00:07:48.106 --> 00:07:48.872
That's a crucial check.

265
00:07:48.903 --> 00:07:50.919
They looked at industry portfolios,

266
00:07:50.981 --> 00:07:51.747
corporate bonds,

267
00:07:51.903 --> 00:07:52.544
currencies,

268
00:07:53.091 --> 00:07:54.403
global equity indices.

269
00:07:54.450 --> 00:07:55.591
Different asset classes.

270
00:07:55.700 --> 00:07:55.966
Yes.

271
00:07:56.539 --> 00:07:58.460
And also international stock markets,

272
00:07:58.760 --> 00:07:58.920
UK,

273
00:07:59.739 --> 00:08:00.139
Europe,

274
00:08:00.360 --> 00:08:00.879
Japan,

275
00:08:01.001 --> 00:08:01.442
Canada.

276
00:08:01.700 --> 00:08:02.282
And the results.

277
00:08:02.762 --> 00:08:03.360
Consistently,

278
00:08:03.540 --> 00:08:06.243
absolute strength momentum seemed to perform better,

279
00:08:06.985 --> 00:08:10.782
more robustly than relative strength momentum across these different domains.

280
00:08:11.063 --> 00:08:13.383
That's pretty powerful evidence for the concept itself.

281
00:08:13.540 --> 00:08:15.118
What about other robustness checks,

282
00:08:15.180 --> 00:08:16.430
different time periods,

283
00:08:16.508 --> 00:08:16.837
rules?

284
00:08:16.946 --> 00:08:17.087
Oh,

285
00:08:17.149 --> 00:08:17.258
yeah,

286
00:08:17.290 --> 00:08:18.368
they really tried to break it.

287
00:08:18.477 --> 00:08:19.805
They looked at the pre-1965 data,

288
00:08:19.806 --> 00:08:21.118
1927,

289
00:08:21.119 --> 00:08:23.243
1964.

290
00:08:23.244 --> 00:08:24.102
The earlier period.

291
00:08:24.639 --> 00:08:28.579
Used different ways to calculate the historical breakpoints rolling windows,

292
00:08:29.059 --> 00:08:31.700
only using NYSE stocks for the breakpoints.

293
00:08:32.379 --> 00:08:33.540
Started the analysis later,

294
00:08:33.602 --> 00:08:34.782
like 1978.

295
00:08:34.783 --> 00:08:35.485
Changed the timing.

296
00:08:35.641 --> 00:08:35.805
Yep.

297
00:08:36.118 --> 00:08:37.899
Varied the ranking and holding periods,

298
00:08:38.040 --> 00:08:38.602
three months,

299
00:08:38.704 --> 00:08:39.243
six months,

300
00:08:39.422 --> 00:08:39.758
nine,

301
00:08:39.821 --> 00:08:40.243
twelve,

302
00:08:40.696 --> 00:08:41.602
instead of just the

303
00:08:42.180 --> 00:08:44.883
11-1-1. And the conclusion through all that.

304
00:08:44.962 --> 00:08:46.821
The absolute strength effect held up.

305
00:08:47.212 --> 00:08:51.040
It seems quite robust to these variations in methodology and time period.

306
00:08:51.352 --> 00:08:51.946
That's convincing.

307
00:08:52.171 --> 00:08:54.373
Did they link it to anything fundamental about the companies?

308
00:08:54.533 --> 00:08:55.353
They did touch on that.

309
00:08:55.375 --> 00:08:58.236
They found that the absolute strength winners tended to show,

310
00:08:58.396 --> 00:08:58.638
you know,

311
00:08:58.736 --> 00:09:00.963
significantly positive growth in things like ROE,

312
00:09:01.338 --> 00:09:01.580
ROA,

313
00:09:01.923 --> 00:09:02.439
earnings,

314
00:09:02.517 --> 00:09:03.486
gross profitability.

315
00:09:03.603 --> 00:09:06.166
So real business momentum behind the stock momentum.

316
00:09:06.167 --> 00:09:07.330
Kind of supports the idea,

317
00:09:07.525 --> 00:09:07.728
yeah.

318
00:09:08.330 --> 00:09:10.111
And the losers showed the opposite,

319
00:09:10.173 --> 00:09:12.470
significantly negative growth in those fundamentals.

320
00:09:12.923 --> 00:09:13.673
And importantly,

321
00:09:13.814 --> 00:09:18.861
these fundamental patterns were stronger for absolute momentum winner closers than for relative ones.

322
00:09:19.439 --> 00:09:20.173
Any other quirks?

323
00:09:20.611 --> 00:09:21.127
Seasonality?

324
00:09:21.439 --> 00:09:22.380
Briefly mentioned,

325
00:09:22.541 --> 00:09:22.701
yeah,

326
00:09:23.161 --> 00:09:26.062
some hints that institutions might track absolute strength,

327
00:09:26.406 --> 00:09:29.468
a possible November effect for absolute losers,

328
00:09:29.890 --> 00:09:33.070
maybe some window dressing pushing up big absolute winners in December.

329
00:09:33.750 --> 00:09:35.211
Standard seasonality patterns,

330
00:09:35.273 --> 00:09:37.054
but seen through this absolute lens.

331
00:09:37.468 --> 00:09:42.656
And how does Abs Mom stack up against other maybe more complex momentum strategies we hear about?

332
00:09:42.922 --> 00:09:43.562
Good question.

333
00:09:43.797 --> 00:09:45.797
They compared it to residual momentum,

334
00:09:46.343 --> 00:09:48.000
constant volatility momentum,

335
00:09:48.078 --> 00:09:49.765
dynamically weighted momentum.

336
00:09:50.335 --> 00:09:51.817
Some of the more advanced factors.

337
00:09:52.337 --> 00:09:53.398
Absmom held its own.

338
00:09:53.739 --> 00:09:56.040
Its Sharpe ratio was very competitive,

339
00:09:56.122 --> 00:09:58.442
often better than these more complex variations.

340
00:09:58.982 --> 00:10:00.146
Simple can be effective,

341
00:10:00.147 --> 00:10:00.646
it seems.

342
00:10:00.747 --> 00:10:01.247
And crucially,

343
00:10:01.302 --> 00:10:02.528
what about momentum crashes?

344
00:10:03.286 --> 00:10:05.810
Relative momentum is known for those painful drawdowns.

345
00:10:06.091 --> 00:10:07.130
That's a key finding.

346
00:10:07.567 --> 00:10:11.724
Absmom seemed to largely avoid the big crashes that played relative strength momentum,

347
00:10:11.833 --> 00:10:13.099
like the one in 2009.

348
00:10:13.271 --> 00:10:14.802
So smoother performance profile.

349
00:10:14.942 --> 00:10:15.614
Apparently so.

350
00:10:16.114 --> 00:10:19.505
Even when they excluded the specific months where relative momentum crashed badly,

351
00:10:19.952 --> 00:10:21.777
Abs Mom's strong performance remained.

352
00:10:21.898 --> 00:10:23.643
It suggests it's capturing a different,

353
00:10:23.684 --> 00:10:25.670
perhaps more fundamental aspect of momentum.

354
00:10:25.951 --> 00:10:27.256
So to wrap up the findings then.

355
00:10:28.254 --> 00:10:28.975
This strategy,

356
00:10:29.135 --> 00:10:32.537
defining winners and losers based on their own historical performance extremes,

357
00:10:33.119 --> 00:10:35.080
seems to be significantly profitable.

358
00:10:35.162 --> 00:10:35.299
Uh-huh.

359
00:10:35.662 --> 00:10:36.346
And robust.

360
00:10:36.526 --> 00:10:38.080
Robust across time,

361
00:10:38.283 --> 00:10:39.127
across markets,

362
00:10:39.385 --> 00:10:40.643
across different ways of setting it up,

363
00:10:41.065 --> 00:10:45.330
and potentially less prone to crashing than standard relative momentum.

364
00:10:45.596 --> 00:10:48.362
That's a pretty accurate summary of what the research indicates.

365
00:10:48.580 --> 00:10:52.080
It's a compelling alternative way to think about momentum.

366
00:10:52.174 --> 00:10:55.315
Definitely gives you something to think about beyond just comparing stocks to each other.

367
00:10:55.698 --> 00:10:57.539
It's about the magnitude of the move itself.

368
00:10:57.661 --> 00:10:58.200
Absolutely.

369
00:10:58.721 --> 00:11:01.563
Is the move historically significant on its own terms?

370
00:11:01.946 --> 00:11:03.547
That seems to be a powerful signal.

371
00:11:04.325 --> 00:11:06.786
Thank you for tuning in to Papers with Backtest podcast.

372
00:11:07.231 --> 00:11:09.411
We hope today's episode gave you useful insights.

373
00:11:09.973 --> 00:11:12.301
Join us next time as we break down more research.

374
00:11:12.770 --> 00:11:14.254
And for more papers and backtests,

375
00:11:14.379 --> 00:11:17.536
find us at https.paperswithbacktest.com.

376
00:11:18.207 --> 00:11:18.786
Happy trading.