WEBVTT

1
00:00:00.040 --> 00:00:00.240
Hello.

2
00:00:00.420 --> 00:00:02.723
Welcome back to Papers with Backtest podcast.

3
00:00:02.844 --> 00:00:05.824
Today we dive into another algo trading research paper.

4
00:00:06.105 --> 00:00:07.289
And this one's quite interesting.

5
00:00:07.328 --> 00:00:13.031
It's called Acceleration Effect Combined with Momentum in Stocks by Liwen Chen and Xinyi Yu.

6
00:00:13.414 --> 00:00:13.656
Okay,

7
00:00:14.578 --> 00:00:15.312
let's unpack this.

8
00:00:15.375 --> 00:00:19.906
Have you ever looked at a stock chart that just seems to be taken off like a rocket?

9
00:00:20.203 --> 00:00:20.375
Yeah,

10
00:00:20.516 --> 00:00:21.156
absolutely.

11
00:00:21.157 --> 00:00:21.281
Well,

12
00:00:21.282 --> 00:00:28.781
the core idea here seems to be that those sort of visual patterns in historical stock prices really grab investors'

13
00:00:28.812 --> 00:00:29.281
attention.

14
00:00:29.684 --> 00:00:29.844
Right.

15
00:00:29.845 --> 00:00:33.248
And the authors argue this attention can lead to overreactions.

16
00:00:33.869 --> 00:00:35.849
And what's really interesting for us,

17
00:00:36.552 --> 00:00:38.556
potentially profitable trading strategies,

18
00:00:38.595 --> 00:00:40.275
maybe even better than standard momentum.

19
00:00:40.353 --> 00:00:42.377
And they didn't just look at a couple of years either.

20
00:00:42.556 --> 00:00:44.478
This study covers a massive period,

21
00:00:44.556 --> 00:00:46.236
January 1962,

22
00:00:46.237 --> 00:00:48.002
all the way to December 2011.

23
00:00:48.142 --> 00:00:48.283
Yeah,

24
00:00:48.284 --> 00:00:51.330
that's nearly 50 years of data across the big U.S.

25
00:00:51.361 --> 00:00:52.783
exchanges to NYSE,

26
00:00:53.080 --> 00:00:54.189
Amex and Nasdaq.

27
00:00:54.548 --> 00:00:55.423
It's a real goldmine.

28
00:00:55.592 --> 00:00:56.033
Exactly.

29
00:00:56.093 --> 00:01:01.257
So our mission today in this deep dive is to really get into the specific trading rules they came up with and,

30
00:01:01.276 --> 00:01:01.499
you know,

31
00:01:01.500 --> 00:01:03.120
the crucial part with the back test showed.

32
00:01:03.382 --> 00:01:03.921
Let's do it.

33
00:01:04.577 --> 00:01:07.843
So they start with this idea of visual attention.

34
00:01:08.764 --> 00:01:10.007
Why do these patterns matter?

35
00:01:10.429 --> 00:01:12.483
They link it to gestalt psychology.

36
00:01:12.702 --> 00:01:13.186
Gestalt.

37
00:01:13.452 --> 00:01:13.577
OK,

38
00:01:13.749 --> 00:01:14.374
refresh my memory.

39
00:01:14.405 --> 00:01:15.858
What's the connection to stock charts?

40
00:01:16.436 --> 00:01:16.577
Well,

41
00:01:16.578 --> 00:01:19.108
the gist is our brains love finding patterns.

42
00:01:19.499 --> 00:01:23.483
The law of continuity specifically suggests we expect trends to continue.

43
00:01:23.655 --> 00:01:23.780
So.

44
00:01:24.104 --> 00:01:25.945
you see a price accelerating upwards.

45
00:01:26.224 --> 00:01:28.146
You instinctively think it'll keep accelerating.

46
00:01:28.306 --> 00:01:28.666
Exactly.

47
00:01:28.884 --> 00:01:29.904
It's not just data points.

48
00:01:29.923 --> 00:01:32.908
It's a visual narrative of acceleration that people latch onto.

49
00:01:33.142 --> 00:01:33.283
Okay,

50
00:01:33.284 --> 00:01:34.189
that makes intuitive sense.

51
00:01:34.509 --> 00:01:37.150
But how did they actually measure this visual pattern?

52
00:01:37.166 --> 00:01:38.705
It wasn't just eyeballing charts,

53
00:01:38.791 --> 00:01:38.923
right?

54
00:01:38.924 --> 00:01:39.048
Oh,

55
00:01:39.087 --> 00:01:39.205
no,

56
00:01:39.384 --> 00:01:40.455
much more systematic.

57
00:01:40.611 --> 00:01:40.814
First,

58
00:01:40.815 --> 00:01:42.314
they do the standard momentum sort,

59
00:01:42.986 --> 00:01:44.330
rank stocks into five groups,

60
00:01:44.673 --> 00:01:46.330
quintiles based on past,

61
00:01:46.392 --> 00:01:46.595
say,

62
00:01:47.048 --> 00:01:47.876
J-month returns.

63
00:01:48.001 --> 00:01:48.142
Yeah.

64
00:01:48.205 --> 00:01:48.970
Your winners and losers.

65
00:01:48.986 --> 00:01:49.673
Standard stuff.

66
00:01:49.876 --> 00:01:50.705
Top 20%

67
00:01:50.736 --> 00:01:51.111
winners,

68
00:01:51.189 --> 00:01:52.095
bottom 20%

69
00:01:52.096 --> 00:01:52.548
losers.

70
00:01:52.970 --> 00:01:53.408
Got it.

71
00:01:53.824 --> 00:01:54.465
What's next?

72
00:01:54.745 --> 00:01:55.566
Here's the clever bit.

73
00:01:56.126 --> 00:01:57.308
For every single stock,

74
00:01:57.728 --> 00:02:02.013
they look at its daily prices over those past J months and run a regression.

75
00:02:02.052 --> 00:02:02.693
A regression.

76
00:02:02.771 --> 00:02:02.896
Yeah.

77
00:02:03.037 --> 00:02:04.497
They regress the price against time,

78
00:02:04.693 --> 00:02:04.896
tin,

79
00:02:05.013 --> 00:02:05.552
and times square,

80
00:02:05.615 --> 00:02:05.794
too.

81
00:02:06.357 --> 00:02:09.896
The equation is basically price alpha plus beta plus gamma.

82
00:02:09.943 --> 00:02:10.068
Ah,

83
00:02:10.583 --> 00:02:10.865
okay.

84
00:02:11.255 --> 00:02:12.583
And I'm guessing that gamma,

85
00:02:12.865 --> 00:02:14.677
that's the crucial part for acceleration.

86
00:02:14.708 --> 00:02:15.161
You got it.

87
00:02:15.318 --> 00:02:16.693
Gamma tells you about the curvature,

88
00:02:16.724 --> 00:02:17.443
the convexity.

89
00:02:17.880 --> 00:02:21.458
Positive gamma means the price curve is bending upwards faster and faster,

90
00:02:21.490 --> 00:02:22.130
accelerating.

91
00:02:22.528 --> 00:02:23.828
Like a car speeding up.

92
00:02:24.069 --> 00:02:24.828
And negative gamma.

93
00:02:25.088 --> 00:02:25.928
Decelerating.

94
00:02:26.268 --> 00:02:27.869
The rate of change is slowing down.

95
00:02:28.287 --> 00:02:30.811
Could be an increase slowing or a decrease slowing.

96
00:02:31.108 --> 00:02:33.069
So within the winners and losers,

97
00:02:33.287 --> 00:02:34.030
they sort again,

98
00:02:34.209 --> 00:02:36.412
but this time using this gamma value.

99
00:02:36.490 --> 00:02:36.975
Precisely.

100
00:02:37.170 --> 00:02:39.209
Into another five quintiles based on gamma.

101
00:02:39.694 --> 00:02:41.225
So you get winners accelerating the most,

102
00:02:41.287 --> 00:02:42.725
winners decelerating the most,

103
00:02:42.850 --> 00:02:44.678
losers accelerating downwards fastest,

104
00:02:45.100 --> 00:02:47.053
and losers whose fall is slowing down.

105
00:02:47.256 --> 00:02:49.287
That gives them a lot of combinations to play with.

106
00:02:49.376 --> 00:02:51.599
They design nine strategies based on this.

107
00:02:51.698 --> 00:02:52.380
Nine in total,

108
00:02:52.439 --> 00:02:52.599
yeah.

109
00:02:53.040 --> 00:02:55.161
Strategy one is just your plain vanilla momentum,

110
00:02:55.220 --> 00:02:55.642
buy winner,

111
00:02:55.704 --> 00:02:56.321
sell losers.

112
00:02:56.845 --> 00:02:59.626
But the others mix in this acceleration factor.

113
00:03:00.024 --> 00:03:02.485
Let's focus on the key ones that really highlight their idea.

114
00:03:03.048 --> 00:03:03.751
Strategy six,

115
00:03:03.907 --> 00:03:05.134
the acceleration strategy,

116
00:03:05.657 --> 00:03:06.626
and strategy seven,

117
00:03:06.689 --> 00:03:08.079
the deceleration strategy.

118
00:03:08.189 --> 00:03:09.298
How does strategy six work?

119
00:03:09.689 --> 00:03:12.157
Strategy six is pure acceleration focus.

120
00:03:13.017 --> 00:03:13.923
They buy the winners,

121
00:03:14.173 --> 00:03:16.064
showing the most acceleration top 20%

122
00:03:16.642 --> 00:03:18.126
gamma within the winners.

123
00:03:18.808 --> 00:03:19.809
And they short the losers,

124
00:03:20.129 --> 00:03:23.310
showing the most acceleration downwards bottom 20%

125
00:03:23.393 --> 00:03:25.033
gamma among the losers.

126
00:03:25.115 --> 00:03:27.756
So betting on those visually strong trends to keep going strong.

127
00:03:27.873 --> 00:03:28.256
Exactly.

128
00:03:28.334 --> 00:03:29.498
Riding that visual momentum,

129
00:03:29.537 --> 00:03:29.920
you could say.

130
00:03:30.021 --> 00:03:30.217
Okay.

131
00:03:30.482 --> 00:03:31.381
And strategy seven,

132
00:03:31.404 --> 00:03:32.560
the deceleration one,

133
00:03:32.701 --> 00:03:33.685
is that the opposite bet?

134
00:03:33.779 --> 00:03:34.185
Pretty much.

135
00:03:34.240 --> 00:03:36.592
It's betting against those accelerating trends continuing.

136
00:03:37.045 --> 00:03:41.264
They buy winners whose price gains are slowing down bottom 20%

137
00:03:41.514 --> 00:03:42.451
gamma in winners.

138
00:03:42.951 --> 00:03:46.748
And they sell losers whose price drops are also slowing top 20%

139
00:03:46.779 --> 00:03:47.639
gamma in losers.

140
00:03:47.788 --> 00:03:49.749
Kind of a contrarian play within momentum,

141
00:03:49.849 --> 00:03:52.572
fading those visually strong moves that seem to be losing steam.

142
00:03:52.833 --> 00:03:53.615
You could see it that way,

143
00:03:53.673 --> 00:03:53.794
yeah.

144
00:03:53.795 --> 00:03:54.173
All right.

145
00:03:54.236 --> 00:03:54.955
The moment of truth.

146
00:03:55.474 --> 00:03:56.236
The back tests.

147
00:03:56.478 --> 00:03:58.361
What happened when they ran these strategies?

148
00:03:58.697 --> 00:03:59.783
Let's use their main example.

149
00:04:00.142 --> 00:04:00.962
Rank on 12 months,

150
00:04:01.244 --> 00:04:02.103
hold for six months.

151
00:04:02.541 --> 00:04:02.666
J12,

152
00:04:03.080 --> 00:04:03.205
K6.

153
00:04:03.837 --> 00:04:05.072
The results were pretty striking.

154
00:04:05.181 --> 00:04:05.728
Plain momentum,

155
00:04:05.853 --> 00:04:06.525
strategy one,

156
00:04:06.822 --> 00:04:07.244
did okay.

157
00:04:07.666 --> 00:04:09.587
83.22 basis points per month.

158
00:04:09.806 --> 00:04:11.978
That's roughly 10.5%

159
00:04:12.119 --> 00:04:12.416
a year.

160
00:04:12.837 --> 00:04:13.197
Not bad.

161
00:04:13.369 --> 00:04:13.556
Okay.

162
00:04:13.572 --> 00:04:14.181
The benchmark.

163
00:04:14.228 --> 00:04:16.181
How did the acceleration strategy stack up?

164
00:04:16.524 --> 00:04:17.145
Strategy six,

165
00:04:17.165 --> 00:04:18.026
the acceleration one,

166
00:04:18.086 --> 00:04:18.626
blew it away.

167
00:04:19.266 --> 00:04:22.171
132.46 basis points per month.

168
00:04:22.229 --> 00:04:22.491
Wow.

169
00:04:22.549 --> 00:04:22.749
That's,

170
00:04:22.952 --> 00:04:23.690
let me calculate,

171
00:04:23.952 --> 00:04:24.850
over 17%

172
00:04:24.975 --> 00:04:25.413
annually.

173
00:04:25.569 --> 00:04:25.733
Yeah.

174
00:04:25.991 --> 00:04:27.178
A really significant jump.

175
00:04:27.491 --> 00:04:28.272
Almost 7%

176
00:04:28.358 --> 00:04:30.741
more per year than standard momentum in this test.

177
00:04:30.975 --> 00:04:31.936
And strategy seven,

178
00:04:32.382 --> 00:04:33.475
the deceleration one,

179
00:04:33.819 --> 00:04:34.757
betting against the curve.

180
00:04:35.022 --> 00:04:35.350
That one?

181
00:04:35.522 --> 00:04:35.678
Well,

182
00:04:35.679 --> 00:04:36.491
it had the lowest return,

183
00:04:36.569 --> 00:04:38.944
only 46.80 basis points per month.

184
00:04:39.600 --> 00:04:43.116
Suggests that fading those visually strong trends wasn't the winning play here.

185
00:04:43.444 --> 00:04:43.866
Interesting.

186
00:04:43.928 --> 00:04:46.038
So betting on acceleration paid off much better.

187
00:04:46.432 --> 00:04:47.333
But what about risk?

188
00:04:47.793 --> 00:04:49.997
Higher returns often mean higher volatility.

189
00:04:50.594 --> 00:04:51.977
How did the Sharpe ratios look?

190
00:04:52.079 --> 00:04:52.579
Good question.

191
00:04:52.739 --> 00:04:55.602
The acceleration strategy also won on risk-adjusted terms.

192
00:04:55.899 --> 00:05:00.665
Sharpe ratio of 0.260 for strategy 6 versus 0.204 for plain momentum.

193
00:05:00.969 --> 00:05:03.461
So better returns and better risk-adjusted returns.

194
00:05:03.711 --> 00:05:04.415
That's compelling.

195
00:05:04.508 --> 00:05:09.071
It definitely suggests the acceleration factor adds something meaningful beyond just higher raw returns.

196
00:05:09.290 --> 00:05:09.461
Now,

197
00:05:09.462 --> 00:05:11.149
is this just a lucky period or set up?

198
00:05:11.243 --> 00:05:12.274
Did they check if it holds up?

199
00:05:12.415 --> 00:05:13.274
Robustness checks?

200
00:05:13.352 --> 00:05:13.665
They did.

201
00:05:13.788 --> 00:05:14.248
quite a few.

202
00:05:14.388 --> 00:05:19.490
The findings were consistent across different subperiods within that long 1962 to 2011 time frame.

203
00:05:19.869 --> 00:05:21.451
They held up across different exchanges too.

204
00:05:21.826 --> 00:05:22.053
Okay.

205
00:05:22.646 --> 00:05:26.334
And they checked using midpoints of badass quotes,

206
00:05:27.428 --> 00:05:27.670
you know,

207
00:05:27.709 --> 00:05:29.998
to make sure it wasn't just bid-ask bounce noise.

208
00:05:30.545 --> 00:05:33.295
They also removed January returns of the January effect,

209
00:05:33.654 --> 00:05:35.014
didn't change the main conclusion.

210
00:05:35.639 --> 00:05:38.561
And results were robust to different ranking and holding periods as well.

211
00:05:38.873 --> 00:05:39.920
Sounds pretty solid then.

212
00:05:40.311 --> 00:05:41.076
Not just a fluke.

213
00:05:41.420 --> 00:05:43.389
The outline mentions mean difference tests.

214
00:05:43.716 --> 00:05:45.577
confirming the outperformance was real.

215
00:05:45.779 --> 00:05:45.958
Yeah.

216
00:05:46.218 --> 00:05:46.640
Table 3,

217
00:05:46.699 --> 00:05:47.060
Panel B,

218
00:05:47.640 --> 00:05:51.585
they formally tested the difference between Strategy 6 and Strategy 1 returns.

219
00:05:52.120 --> 00:05:52.628
And yes,

220
00:05:52.722 --> 00:05:56.691
the acceleration strategy's outperformance was statistically significant.

221
00:05:56.831 --> 00:05:56.949
OK.

222
00:05:57.652 --> 00:06:00.995
What if you looked back even further than 12 months to rank them,

223
00:06:01.261 --> 00:06:02.339
say 24 months?

224
00:06:02.417 --> 00:06:03.308
Did that change anything?

225
00:06:03.495 --> 00:06:03.870
Actually,

226
00:06:03.980 --> 00:06:11.464
Appendix 1 shows that using longer ranking periods like J24 often led to even higher returns for the acceleration strategy.

227
00:06:11.636 --> 00:06:11.886
Real?

228
00:06:11.949 --> 00:06:12.464
And get this.

229
00:06:12.648 --> 00:06:14.190
For some of those longer ranking periods,

230
00:06:14.390 --> 00:06:18.894
the sharp ratio for the acceleration strategy was almost double that of plain momentum.

231
00:06:19.015 --> 00:06:19.394
Double.

232
00:06:19.655 --> 00:06:19.894
Wow.

233
00:06:20.015 --> 00:06:22.620
So that visual effect might build up over longer trends.

234
00:06:22.621 --> 00:06:22.722
See,

235
00:06:22.723 --> 00:06:23.276
it's plausible,

236
00:06:23.316 --> 00:06:23.480
yeah.

237
00:06:23.839 --> 00:06:26.120
The effect appears stronger over longer lookbacks.

238
00:06:26.503 --> 00:06:26.620
OK,

239
00:06:26.823 --> 00:06:33.253
but is this acceleration strategy just finding the absolute highest momentum stocks anyway?

240
00:06:33.347 --> 00:06:35.956
Is it just like a fancier way to pick extreme winners?

241
00:06:36.347 --> 00:06:36.925
They checked that,

242
00:06:36.972 --> 00:06:37.175
too.

243
00:06:37.878 --> 00:06:40.409
Looked at the overlap between stocks picked by strategy six.

244
00:06:40.904 --> 00:06:43.807
And a simple extreme momentum strategy like top bottom 4%.

245
00:06:44.247 --> 00:06:44.407
Yeah.

246
00:06:44.507 --> 00:06:45.948
The overlap was surprisingly low,

247
00:06:46.229 --> 00:06:46.909
only about 20,

248
00:06:47.112 --> 00:06:48.112
25%.

249
00:06:48.113 --> 00:06:48.229
Huh.

250
00:06:48.471 --> 00:06:52.753
So it's genuinely picking a different subset of stocks based on that gamma,

251
00:06:52.893 --> 00:06:53.596
that curvature.

252
00:06:53.917 --> 00:06:54.354
Exactly.

253
00:06:54.393 --> 00:06:56.261
It's not just relabeling extreme momentum.

254
00:06:56.339 --> 00:06:59.268
It's capturing something different related to the pattern of returns.

255
00:06:59.346 --> 00:07:02.581
Did they try to control for other known factors like size,

256
00:07:02.690 --> 00:07:03.159
value,

257
00:07:03.471 --> 00:07:03.940
volatility,

258
00:07:03.987 --> 00:07:05.534
maybe the 52 week high effect?

259
00:07:05.535 --> 00:07:07.143
Does acceleration still matter then?

260
00:07:07.487 --> 00:07:07.690
Yes.

261
00:07:07.753 --> 00:07:09.237
They ran FOMM and kept progressions,

262
00:07:09.268 --> 00:07:09.737
table four.

263
00:07:09.753 --> 00:07:09.971
Right.

264
00:07:10.088 --> 00:07:11.589
They threw in controls for size,

265
00:07:11.670 --> 00:07:12.211
book-to-market,

266
00:07:12.271 --> 00:07:12.830
volatility,

267
00:07:12.931 --> 00:07:13.451
nearness to the

268
00:07:13.892 --> 00:07:14.252
52-week high,

269
00:07:14.513 --> 00:07:15.513
all the usual suspects.

270
00:07:15.591 --> 00:07:16.955
And the acceleration dummies.

271
00:07:17.115 --> 00:07:17.935
Still significant,

272
00:07:17.998 --> 00:07:21.060
both the dummies for accelerating winners and accelerating losers held up.

273
00:07:21.623 --> 00:07:26.271
Suggests this visual pattern has predictive power even after accounting for those other effects.

274
00:07:26.568 --> 00:07:27.599
That's quite powerful.

275
00:07:28.099 --> 00:07:30.646
It means the look of the trend matters independently.

276
00:07:31.302 --> 00:07:33.646
What about standard factor models like Fama French?

277
00:07:33.787 --> 00:07:35.630
Can they explain these extra profits?

278
00:07:35.834 --> 00:07:36.787
That's in Table 5.

279
00:07:37.380 --> 00:07:38.099
And the answer is no.

280
00:07:39.276 --> 00:07:44.021
The Fama-French three-factor model couldn't explain the alpha from the acceleration strategy.

281
00:07:45.021 --> 00:07:47.927
Neither could models adding momentum itself or the

282
00:07:48.404 --> 00:07:49.364
52-week high factor.

283
00:07:49.622 --> 00:07:54.052
So the acceleration profits are distinct from those known factors.

284
00:07:54.091 --> 00:07:54.935
Not only distinct,

285
00:07:55.029 --> 00:07:55.732
but here's the kicker.

286
00:07:56.247 --> 00:08:02.638
They found evidence that the acceleration strategy could actually help explain some of the profits of plane momentum and the

287
00:08:03.185 --> 00:08:04.122
52-week high strategy.

288
00:08:04.138 --> 00:08:04.388
Whoa,

289
00:08:04.497 --> 00:08:04.841
hang on.

290
00:08:05.075 --> 00:08:06.216
It explains them,

291
00:08:06.544 --> 00:08:07.529
not the other way around.

292
00:08:07.716 --> 00:08:08.997
That's what the results suggested.

293
00:08:09.152 --> 00:08:14.959
It points towards this visual acceleration potentially being a more fundamental driver underneath some of those other effects.

294
00:08:14.960 --> 00:08:16.338
That is a really interesting finding.

295
00:08:16.600 --> 00:08:16.760
Okay,

296
00:08:16.779 --> 00:08:17.662
what about longer term?

297
00:08:17.764 --> 00:08:18.842
Minimum eventually reverses,

298
00:08:18.904 --> 00:08:19.084
right?

299
00:08:19.085 --> 00:08:19.725
Did they look at that?

300
00:08:19.881 --> 00:08:20.826
They did in Table 6.

301
00:08:21.061 --> 00:08:22.147
Left out up to 60 months.

302
00:08:22.608 --> 00:08:22.944
And yes,

303
00:08:23.029 --> 00:08:23.787
reversals happen,

304
00:08:23.889 --> 00:08:24.889
typically starting around the

305
00:08:25.326 --> 00:08:25.748
24-month mark.

306
00:08:25.967 --> 00:08:27.858
Consistent with other momentum research,

307
00:08:28.311 --> 00:08:30.389
anything different about the acceleration stocks?

308
00:08:30.717 --> 00:08:30.951
Yes.

309
00:08:31.717 --> 00:08:32.920
The reversals were larger,

310
00:08:33.279 --> 00:08:34.139
and they happened sooner,

311
00:08:34.701 --> 00:08:38.139
for the stocks identified by those acceleration strategies like 5,

312
00:08:38.186 --> 00:08:38.342
6,

313
00:08:38.404 --> 00:08:38.717
and 8.

314
00:08:39.000 --> 00:08:41.161
So a bigger boom followed by a bigger bust,

315
00:08:41.220 --> 00:08:41.421
maybe.

316
00:08:41.901 --> 00:08:42.280
Kind of.

317
00:08:42.401 --> 00:08:43.901
It fits their narrative perfectly.

318
00:08:44.581 --> 00:08:47.983
The visual acceleration leads to a stronger initial overreaction,

319
00:08:48.382 --> 00:08:51.085
which then results in a more pronounced correction later on.

320
00:08:51.460 --> 00:08:51.882
Makes sense.

321
00:08:51.921 --> 00:08:56.601
Did they give any color on what types of companies these accelerating winners and losers tend to be?

322
00:08:57.038 --> 00:08:58.288
A little bit in table seven.

323
00:08:58.648 --> 00:09:02.663
Accelerating winners tended to be larger firms,

324
00:09:02.726 --> 00:09:03.398
less volatile,

325
00:09:03.569 --> 00:09:04.085
more liquid.

326
00:09:04.226 --> 00:09:04.351
OK.

327
00:09:04.851 --> 00:09:06.726
And the losers accelerating downwards.

328
00:09:06.804 --> 00:09:07.913
Those tended to be smaller.

329
00:09:08.188 --> 00:09:09.749
More volatile and less liquid,

330
00:09:10.008 --> 00:09:12.767
especially compared to losers whose fall was decelerating.

331
00:09:12.990 --> 00:09:13.209
Right.

332
00:09:13.490 --> 00:09:14.088
And just quickly,

333
00:09:14.388 --> 00:09:15.470
did it work consistently,

334
00:09:15.670 --> 00:09:17.490
like through recessions and expansions?

335
00:09:17.732 --> 00:09:18.552
Table 8 looked at that.

336
00:09:18.912 --> 00:09:21.795
Performance held up across different subperiods and business cycles.

337
00:09:22.334 --> 00:09:24.677
Profitability was generally higher in expansions,

338
00:09:24.990 --> 00:09:25.974
which isn't too surprising,

339
00:09:26.318 --> 00:09:29.677
but the relative advantage of the acceleration strategy seemed consistent.

340
00:09:29.974 --> 00:09:33.130
So not just an artifact of a particular economic climate?

341
00:09:33.380 --> 00:09:34.115
Doesn't seem like it.

342
00:09:34.927 --> 00:09:37.693
And they also double-checked using midpoint quotes to rule out

343
00:09:37.836 --> 00:09:39.956
bid ask bounce effects in table nine.

344
00:09:40.476 --> 00:09:41.776
The main results held there too.

345
00:09:42.198 --> 00:09:42.698
And finally,

346
00:09:42.757 --> 00:09:45.499
they had a few other checks mentioned on page 38,

347
00:09:45.698 --> 00:09:48.359
excluding certain stocks or tweaking the timing.

348
00:09:48.741 --> 00:09:48.898
Yeah,

349
00:09:48.921 --> 00:09:53.999
they found profits were often even higher if they didn't skip a month between ranking and holding.

350
00:09:54.614 --> 00:09:57.757
Also stronger if they excluded penny stocks or micro caps.

351
00:09:58.398 --> 00:09:59.220
And interestingly,

352
00:09:59.337 --> 00:10:02.181
the effect seemed particularly potent for Nasdaq stocks.

353
00:10:02.646 --> 00:10:02.763
OK,

354
00:10:03.060 --> 00:10:03.841
lots of digging there.

355
00:10:04.302 --> 00:10:05.763
So wrapping this all up,

356
00:10:06.247 --> 00:10:08.951
what's the big takeaway for you from this deep dive?

357
00:10:09.310 --> 00:10:09.568
For me,

358
00:10:09.747 --> 00:10:11.287
it's that the visual pattern,

359
00:10:11.333 --> 00:10:14.615
the acceleration of the price trend seems genuinely important.

360
00:10:14.677 --> 00:10:15.958
It's not just that a stock went up,

361
00:10:16.021 --> 00:10:17.396
but how it went up visually.

362
00:10:18.099 --> 00:10:19.880
This appears to capture investor attention,

363
00:10:20.162 --> 00:10:21.099
lead to overreaction.

364
00:10:21.537 --> 00:10:22.865
And create these measurable...

365
00:10:23.074 --> 00:10:25.116
potentially superior trading opportunity.

366
00:10:25.197 --> 00:10:25.716
Exactly.

367
00:10:25.757 --> 00:10:29.199
The acceleration strategy they outline looks like a more focused,

368
00:10:29.281 --> 00:10:32.804
maybe more powerful way to harness momentum than just the standard approach.

369
00:10:33.460 --> 00:10:35.687
It adds a new dimension based on how trends look.

370
00:10:35.984 --> 00:10:38.827
It's fascinating to think that something like a visual illusion,

371
00:10:38.991 --> 00:10:40.132
how a chart curves,

372
00:10:40.695 --> 00:10:44.882
could be a key driver behind a major market anomaly like momentum.

373
00:10:45.195 --> 00:10:45.773
It really is.

374
00:10:46.163 --> 00:10:50.804
And the authors make a strong case that this kind of visually driven overreaction provides a

375
00:10:50.805 --> 00:10:53.460
a pretty compelling explanation for why momentum works.

376
00:10:53.921 --> 00:10:56.302
Thank you for tuning in to Papers with Backtests podcast.

377
00:10:56.544 --> 00:10:58.704
We hope today's episode gave you useful insights.

378
00:10:59.025 --> 00:11:01.009
Join us next time as we break down more research.

379
00:11:01.290 --> 00:11:02.673
And for more papers and backtests,

380
00:11:02.829 --> 00:11:06.071
find us at https.paperswithbacktests.com.

381
00:11:06.556 --> 00:11:07.157
Happy trading.