WEBVTT

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

2
00:00:00.300 --> 00:00:02.281
welcome back to Papers with Backtest podcast.

3
00:00:02.320 --> 00:00:02.561
Today,

4
00:00:02.820 --> 00:00:05.141
we dive into another Algo trading research paper.

5
00:00:05.383 --> 00:00:05.641
Yes,

6
00:00:05.742 --> 00:00:08.461
today we're looking at lottery-related anomalies,

7
00:00:08.961 --> 00:00:11.125
the role of reference-dependent preferences.

8
00:00:11.203 --> 00:00:11.844
It's by Ahn,

9
00:00:12.359 --> 00:00:12.703
Wang,

10
00:00:13.000 --> 00:00:13.281
Wang,

11
00:00:13.539 --> 00:00:13.844
and Yu,

12
00:00:14.406 --> 00:00:15.922
originally penned back in

13
00:00:16.469 --> 00:00:18.328
November 2015.

14
00:00:18.329 --> 00:00:20.188
And this one tackles a really interesting question,

15
00:00:20.219 --> 00:00:20.625
doesn't it,

16
00:00:21.031 --> 00:00:25.688
about why these sort of lottery ticket stocks tend to do poorly.

17
00:00:25.844 --> 00:00:26.359
Exactly.

18
00:00:26.422 --> 00:00:26.781
Stocks with

19
00:00:26.848 --> 00:00:28.670
With that small chance of a massive payoff,

20
00:00:28.671 --> 00:00:28.891
you know,

21
00:00:28.930 --> 00:00:29.651
high skewness,

22
00:00:30.250 --> 00:00:32.254
why do they generally underperform?

23
00:00:32.453 --> 00:00:33.953
And maybe more importantly,

24
00:00:34.035 --> 00:00:36.895
does that change based on whether investors have,

25
00:00:37.356 --> 00:00:37.496
say,

26
00:00:37.598 --> 00:00:39.137
recently made money or lost money?

27
00:00:39.317 --> 00:00:39.504
Right.

28
00:00:39.520 --> 00:00:40.840
The whole reference point idea.

29
00:00:40.841 --> 00:00:42.379
So these lottery features.

30
00:00:42.426 --> 00:00:42.567
Yeah.

31
00:00:42.926 --> 00:00:46.114
How did they actually pin down what makes a stock lottery like?

32
00:00:46.426 --> 00:00:46.567
Well,

33
00:00:46.582 --> 00:00:47.895
they used a few different measures.

34
00:00:48.051 --> 00:00:52.473
Things like the maximum return a stock had on any single day in the past month.

35
00:00:52.598 --> 00:00:53.770
That's one clear sign.

36
00:00:53.942 --> 00:00:54.114
Okay.

37
00:00:54.395 --> 00:00:55.098
Or they looked at the...

38
00:00:55.240 --> 00:00:57.522
predicted probability of a huge future jump.

39
00:00:57.843 --> 00:00:59.825
They call it the predicted jackpot probability.

40
00:01:00.063 --> 00:01:01.985
Also expected idiosyncratic skewness,

41
00:01:02.067 --> 00:01:03.348
which is a bit more technical.

42
00:01:03.888 --> 00:01:06.052
And even things like failure probability or,

43
00:01:06.552 --> 00:01:06.794
you know,

44
00:01:06.809 --> 00:01:08.192
the chance of outright bankruptcy.

45
00:01:08.973 --> 00:01:11.919
Basically different ways to capture that long shot potential.

46
00:01:12.294 --> 00:01:12.591
Gotcha.

47
00:01:12.981 --> 00:01:13.778
And the big takeaway,

48
00:01:13.934 --> 00:01:15.013
the headline finding.

49
00:01:15.153 --> 00:01:15.388
Yeah.

50
00:01:15.497 --> 00:01:16.544
The key thing is this,

51
00:01:16.888 --> 00:01:18.059
that underperformance,

52
00:01:18.106 --> 00:01:19.122
the lottery anomaly,

53
00:01:19.263 --> 00:01:21.684
it's really strong when investors have experienced losses.

54
00:01:21.950 --> 00:01:22.184
But.

55
00:01:22.403 --> 00:01:22.591
What?

56
00:01:22.731 --> 00:01:24.309
But if investors have been winning.

57
00:01:24.840 --> 00:01:26.122
If they've seen gains,

58
00:01:26.362 --> 00:01:30.544
that underperformance gets much weaker or it can actually flip entirely.

59
00:01:30.825 --> 00:01:32.848
The lottery stocks might even outperform.

60
00:01:33.325 --> 00:01:33.567
Wow.

61
00:01:33.692 --> 00:01:33.809
OK,

62
00:01:34.052 --> 00:01:35.153
so it's state dependent.

63
00:01:35.208 --> 00:01:36.934
It really matters what happened before.

64
00:01:37.216 --> 00:01:37.716
Precisely.

65
00:01:37.856 --> 00:01:39.091
It's not a constant effect.

66
00:01:39.395 --> 00:01:50.950
So our mission today for this deep dive is to really unpack the trading rules and the back tests behind this state dependent anomaly and understand how this ties into reference dependent preferences.

67
00:01:51.013 --> 00:01:51.184
Right.

68
00:01:51.200 --> 00:01:52.653
How are recent gains or losses?

69
00:01:53.016 --> 00:01:53.817
color our choices.

70
00:01:53.957 --> 00:01:54.378
Exactly.

71
00:01:54.437 --> 00:01:57.339
Let's start with how they actually measured those prior gains or losses.

72
00:01:57.882 --> 00:02:01.284
They used a concept called capital gains overhang or CGO.

73
00:02:01.487 --> 00:02:01.964
CGO.

74
00:02:02.026 --> 00:02:03.667
And they didn't just use one method,

75
00:02:03.729 --> 00:02:08.175
they used two distinct approaches to try and capture this feeling of being ahead or behind.

76
00:02:08.331 --> 00:02:08.573
All right,

77
00:02:08.574 --> 00:02:08.933
what were they?

78
00:02:09.128 --> 00:02:10.800
So the first one is ECOGH.

79
00:02:11.128 --> 00:02:12.894
That's based on work by Grimblatt and Hahn.

80
00:02:13.597 --> 00:02:19.769
The idea is to compare the current price to a sort of weighted average of past prices.

81
00:02:20.003 --> 00:02:20.565
Wait and buy.

82
00:02:20.628 --> 00:02:21.019
Bird over.

83
00:02:21.412 --> 00:02:22.613
how much the stock is traded,

84
00:02:23.054 --> 00:02:26.575
it's meant to proxy the average price investors actually bought in at.

85
00:02:26.876 --> 00:02:28.438
They look back up to five years,

86
00:02:28.560 --> 00:02:29.661
provided they have enough data,

87
00:02:29.802 --> 00:02:31.599
like 150 weeks minimum.

88
00:02:31.802 --> 00:02:35.903
And that gives you a sense of whether the current price is above or below where people likely bought in.

89
00:02:36.067 --> 00:02:36.544
Exactly.

90
00:02:37.130 --> 00:02:39.591
High CGO means likely sitting on gains.

91
00:02:40.138 --> 00:02:40.903
Low CGO,

92
00:02:41.528 --> 00:02:42.356
likely losses.

93
00:02:42.997 --> 00:02:46.716
And their data for this ran from 65 to 2014,

94
00:02:46.717 --> 00:02:47.591
so a long period.

95
00:02:47.685 --> 00:02:47.794
OK,

96
00:02:47.903 --> 00:02:49.106
that's CGOGH.

97
00:02:49.153 --> 00:02:50.060
What was the second measure?

98
00:02:50.116 --> 00:02:52.940
The second one was CGOFR from Fred Zini's research.

99
00:02:53.299 --> 00:02:54.241
This takes a different angle.

100
00:02:54.319 --> 00:02:56.041
It looks at actual mutual fund holdings.

101
00:02:56.143 --> 00:02:56.260
Ah,

102
00:02:56.643 --> 00:02:59.205
using fund managers as a proxy for the average investor.

103
00:02:59.307 --> 00:02:59.643
Kind of,

104
00:02:59.666 --> 00:02:59.862
yeah.

105
00:03:00.369 --> 00:03:06.112
The assumption is that the aggregate buying behavior of mutual funds gives you a clue about the reference prices for many investors.

106
00:03:06.455 --> 00:03:07.721
This data was a bit more recent,

107
00:03:07.830 --> 00:03:10.237
April 80 to October 2014.

108
00:03:10.238 --> 00:03:10.596
Interesting.

109
00:03:11.018 --> 00:03:13.362
Two different ways to gauge the market's gain-loss state.

110
00:03:14.018 --> 00:03:18.362
So how do they connect this CGO measure to the performance of those lottery stocks?

111
00:03:18.512 --> 00:03:18.692
Right.

112
00:03:18.792 --> 00:03:20.094
This is the core of the back test.

113
00:03:20.674 --> 00:03:22.397
They used independent double sorts.

114
00:03:22.916 --> 00:03:24.037
It sounds complicated,

115
00:03:24.299 --> 00:03:25.537
but the idea is straightforward.

116
00:03:25.698 --> 00:03:25.838
Okay,

117
00:03:25.858 --> 00:03:26.557
break it down for us.

118
00:03:26.916 --> 00:03:27.744
At the end of every month,

119
00:03:27.799 --> 00:03:28.784
they'd take all the stocks.

120
00:03:29.002 --> 00:03:29.346
First,

121
00:03:29.502 --> 00:03:30.948
sort them into five buckets,

122
00:03:31.307 --> 00:03:31.948
quintiles,

123
00:03:32.487 --> 00:03:33.510
based on their CGO,

124
00:03:33.909 --> 00:03:37.166
from low CGO losses to high CGO gains.

125
00:03:37.369 --> 00:03:37.916
Step one,

126
00:03:38.416 --> 00:03:39.229
sort by CGO.

127
00:03:39.479 --> 00:03:39.994
Step two,

128
00:03:40.369 --> 00:03:45.463
independently sort all those same stocks into five other buckets based on one of the lottery measures,

129
00:03:45.526 --> 00:03:45.698
say...

130
00:03:46.104 --> 00:03:46.665
MaxRet,

131
00:03:46.705 --> 00:03:48.787
the maximum daily return from the last month.

132
00:03:48.886 --> 00:03:49.447
So you end up with,

133
00:03:49.507 --> 00:03:49.648
what,

134
00:03:49.707 --> 00:03:50.988
25 portfolios?

135
00:03:51.488 --> 00:03:51.730
Yeah,

136
00:03:51.789 --> 00:03:52.230
exactly.

137
00:03:52.250 --> 00:03:53.250
A five by five grid.

138
00:03:53.390 --> 00:03:57.836
Each cell represents stocks with a certain level of CGO and a certain level of lotteriness.

139
00:03:58.273 --> 00:04:00.757
Then they track the returns of these portfolios for the next month.

140
00:04:00.836 --> 00:04:05.781
And this lets you see how lottery stocks perform specifically within the loss group versus the gain group?

141
00:04:05.937 --> 00:04:06.468
Precisely.

142
00:04:06.687 --> 00:04:07.390
And the results,

143
00:04:07.422 --> 00:04:10.031
especially using MaxRet and the CGOGH measure,

144
00:04:10.140 --> 00:04:11.187
were quite striking.

145
00:04:11.547 --> 00:04:11.765
All right,

146
00:04:11.766 --> 00:04:12.187
let's hear them.

147
00:04:12.765 --> 00:04:14.531
What happened in the low CGO group?

148
00:04:14.924 --> 00:04:16.746
The stocks where investors were likely down.

149
00:04:17.227 --> 00:04:17.446
Okay,

150
00:04:17.506 --> 00:04:20.008
so within that bottom CGO quintile,

151
00:04:20.270 --> 00:04:22.153
the stocks that were most lottery-like,

152
00:04:22.211 --> 00:04:23.852
the top quintile for MaxRet,

153
00:04:24.336 --> 00:04:27.774
they seriously underperformed the least lottery-like stocks,

154
00:04:27.977 --> 00:04:29.680
the bottom quintile for MaxRet.

155
00:04:29.899 --> 00:04:30.938
How much underperformance?

156
00:04:31.094 --> 00:04:33.266
We're talking 1.38%

157
00:04:33.344 --> 00:04:35.000
per month in excess returns.

158
00:04:35.266 --> 00:04:36.141
It's a big difference,

159
00:04:36.250 --> 00:04:38.219
and it was statistically very significant,

160
00:04:38.469 --> 00:04:40.844
T-stat of negative 5.35.

161
00:04:40.845 --> 00:04:41.110
Wow.

162
00:04:41.766 --> 00:04:44.297
So chasing those high flyers after you've already lost money.

163
00:04:45.053 --> 00:04:46.935
seemed to backfire badly according to this.

164
00:04:47.136 --> 00:04:48.456
That's what the data suggests.

165
00:04:48.636 --> 00:04:48.837
Now,

166
00:04:48.995 --> 00:04:53.144
contrast that with the high CGO group stocks where investors were likely sitting on gains.

167
00:04:53.378 --> 00:04:53.605
Okay,

168
00:04:53.722 --> 00:04:54.441
what happened there?

169
00:04:54.683 --> 00:04:56.566
Did the lottery stocks still underperform?

170
00:04:56.745 --> 00:04:56.863
No,

171
00:04:57.128 --> 00:04:57.988
this is where it flips.

172
00:04:58.308 --> 00:04:59.824
In the top CGO quintile,

173
00:05:00.105 --> 00:05:03.652
the high max-ret stocks actually outperformed the low max-ret stocks.

174
00:05:03.730 --> 00:05:04.839
Outperformed by how much?

175
00:05:04.917 --> 00:05:06.292
By 0.54%

176
00:05:06.527 --> 00:05:06.949
per month,

177
00:05:07.230 --> 00:05:08.636
also statistically significant,

178
00:05:08.777 --> 00:05:10.620
with a T-stat of 2.30.

179
00:05:10.621 --> 00:05:11.386
That's fascinating.

180
00:05:11.417 --> 00:05:12.699
So the same type of stock

181
00:05:13.013 --> 00:05:16.537
behaves completely differently depending on whether the average holder is likely up or down.

182
00:05:16.976 --> 00:05:17.660
Exactly.

183
00:05:17.718 --> 00:05:20.320
It really highlights the state dependent nature.

184
00:05:20.519 --> 00:05:22.925
What if you just ignored CGO altogether?

185
00:05:23.245 --> 00:05:26.886
Just looked at high max rate versus low max rate across all stocks?

186
00:05:26.964 --> 00:05:27.144
Well,

187
00:05:27.347 --> 00:05:28.925
the effect largely disappears.

188
00:05:29.105 --> 00:05:30.277
The unconditional spread,

189
00:05:30.433 --> 00:05:32.370
just P5 minus P1 for max rate,

190
00:05:32.699 --> 00:05:34.167
was only 0.24%

191
00:05:34.480 --> 00:05:34.855
per month.

192
00:05:35.037 --> 00:05:37.019
and that wasn't statistically significant.

193
00:05:37.020 --> 00:05:39.861
So you need that CGO context to see the strong anomaly.

194
00:05:39.862 --> 00:05:40.704
That seems crucial,

195
00:05:40.783 --> 00:05:41.064
yes.

196
00:05:41.204 --> 00:05:44.204
And they quantified this further with a difference-in-differences calculation.

197
00:05:44.564 --> 00:05:44.783
Right,

198
00:05:44.829 --> 00:05:46.009
comparing the lottery spread,

199
00:05:46.189 --> 00:05:46.650
P5-P1,

200
00:05:46.947 --> 00:05:49.494
in the high CGO group versus the low CGO group.

201
00:05:49.548 --> 00:05:49.829
Yes.

202
00:05:50.353 --> 00:05:53.329
And that difference was a hefty 1.92%

203
00:05:53.330 --> 00:05:55.579
per month using MaxRet and CGOGH.

204
00:05:56.158 --> 00:05:57.642
Very statistically significant.

205
00:05:57.876 --> 00:05:59.126
T-stat of 7.50.

206
00:05:59.361 --> 00:06:04.564
It really drives home that the lottery effect itself changes dramatically based on prior gains or losses.

207
00:06:04.897 --> 00:06:06.980
That's pretty compelling evidence for the state dependency.

208
00:06:07.480 --> 00:06:10.204
Did this pattern hold up when they used the other lottery measures,

209
00:06:10.482 --> 00:06:12.524
like skewness or bankruptcy risk?

210
00:06:12.806 --> 00:06:13.204
Generally,

211
00:06:13.282 --> 00:06:13.564
yes.

212
00:06:13.704 --> 00:06:16.368
They found similar state-dependent patterns across the board,

213
00:06:16.407 --> 00:06:18.251
predicted jackpot probability,

214
00:06:18.610 --> 00:06:19.048
skewness,

215
00:06:19.228 --> 00:06:20.134
failure probability,

216
00:06:20.290 --> 00:06:21.071
bankruptcy risk.

217
00:06:21.431 --> 00:06:22.837
The magnitudes varied a bit,

218
00:06:22.853 --> 00:06:23.118
you know.

219
00:06:23.415 --> 00:06:25.118
But the core story was consistent.

220
00:06:25.603 --> 00:06:31.446
The lottery anomaly was much stronger following losses and weaker or even reversed following gains.

221
00:06:31.587 --> 00:06:31.759
Okay,

222
00:06:31.790 --> 00:06:33.103
but could this just be risk?

223
00:06:33.737 --> 00:06:35.940
Maybe these lottery stocks are just way riskier,

224
00:06:35.959 --> 00:06:37.020
especially after losses,

225
00:06:37.461 --> 00:06:38.864
and the returns reflect that.

226
00:06:39.160 --> 00:06:41.203
Did they check against standard risk models?

227
00:06:41.383 --> 00:06:42.106
Good question.

228
00:06:42.207 --> 00:06:42.746
They did.

229
00:06:43.184 --> 00:06:45.668
They looked at Fama French three-factor alphas.

230
00:06:45.950 --> 00:06:47.434
These alphas adjust for market,

231
00:06:47.575 --> 00:06:47.989
size,

232
00:06:48.114 --> 00:06:48.871
and value risk.

233
00:06:48.950 --> 00:06:50.106
Did the anomaly survive?

234
00:06:50.340 --> 00:06:50.746
It did.

235
00:06:51.215 --> 00:06:54.903
The patterns in the alphas were very similar to the raw excess returns.

236
00:06:55.121 --> 00:06:55.653
For example,

237
00:06:55.684 --> 00:06:57.465
with MaxRed and CGOGH,

238
00:06:57.903 --> 00:07:01.762
the alpha spread in the low CGO group was Mattis 1.76%.

239
00:07:01.887 --> 00:07:02.200
percent

240
00:07:02.281 --> 00:07:02.641
per month,

241
00:07:02.721 --> 00:07:03.703
still very significant.

242
00:07:03.863 --> 00:07:05.203
And in the high CGO group?

243
00:07:05.285 --> 00:07:06.064
It was positive,

244
00:07:06.285 --> 00:07:07.605
plus 0.35%

245
00:07:07.687 --> 00:07:07.988
per month,

246
00:07:08.066 --> 00:07:08.929
also significant.

247
00:07:08.969 --> 00:07:13.109
So standard risk factors don't seem to explain away this state-dependent effect.

248
00:07:13.250 --> 00:07:14.297
Sounds like they were pretty thorough.

249
00:07:14.298 --> 00:07:15.172
Did they run other checks,

250
00:07:15.211 --> 00:07:16.398
other ways to slice the data?

251
00:07:16.515 --> 00:07:16.812
Oh yeah,

252
00:07:16.875 --> 00:07:18.422
they did a lot of robustness checks.

253
00:07:18.750 --> 00:07:23.687
They used Fama-Macbeth regressions to control for other stock characteristics simultaneously,

254
00:07:23.797 --> 00:07:24.578
things like size,

255
00:07:24.640 --> 00:07:25.281
book-to-market,

256
00:07:25.437 --> 00:07:25.969
turnover,

257
00:07:26.078 --> 00:07:26.640
volatility.

258
00:07:26.812 --> 00:07:28.125
The state-dependent effect held.

259
00:07:28.531 --> 00:07:29.203
It persisted.

260
00:07:29.737 --> 00:07:32.957
They also confirmed the results using that other CGO measure,

261
00:07:33.239 --> 00:07:34.176
CGOFR,

262
00:07:34.297 --> 00:07:35.418
based on mutual funds.

263
00:07:36.258 --> 00:07:37.418
Got similar results,

264
00:07:37.539 --> 00:07:39.621
a difference in differences for max rent,

265
00:07:39.739 --> 00:07:41.520
around 1.88%.

266
00:07:41.996 --> 00:07:44.098
They even calculated a residual CGO,

267
00:07:44.262 --> 00:07:44.762
basically,

268
00:07:45.137 --> 00:07:47.778
the part of CGO not explained by other stock features,

269
00:07:47.856 --> 00:07:48.981
and still found the pattern.

270
00:07:49.778 --> 00:07:52.184
Used equal weighted returns instead of valuated,

271
00:07:52.278 --> 00:07:52.950
results held.

272
00:07:53.621 --> 00:07:56.184
Used conditional sorting instead of independent sorts,

273
00:07:56.559 --> 00:07:57.106
still held.

274
00:07:57.262 --> 00:07:57.606
Wow.

275
00:07:57.840 --> 00:07:58.418
And finally,

276
00:07:58.590 --> 00:07:59.137
those fama-

277
00:07:59.425 --> 00:08:06.725
Beth regressions directly tested the interaction term between lottery proxies and CGO and found it was positive and significant,

278
00:08:06.846 --> 00:08:09.588
directly supporting the state dependent hypothesis.

279
00:08:10.205 --> 00:08:11.549
So across different measures,

280
00:08:11.666 --> 00:08:12.385
different methods,

281
00:08:12.471 --> 00:08:13.791
controlling for various factors,

282
00:08:14.244 --> 00:08:15.744
the conclusion seems pretty solid.

283
00:08:15.885 --> 00:08:17.291
It really does seem robust.

284
00:08:17.369 --> 00:08:23.307
The underperformance of lottery stocks seems heavily tied to whether investors are in a loss domain relative to their reference point.

285
00:08:23.432 --> 00:08:25.510
Which brings us back to the author's interpretation.

286
00:08:25.572 --> 00:08:26.947
How do they explain why this happened?

287
00:08:27.253 --> 00:08:30.476
Their explanation leans heavily on behavioral finance concepts,

288
00:08:30.896 --> 00:08:32.156
reference-dependent preferences,

289
00:08:32.199 --> 00:08:33.117
and mental accounting.

290
00:08:33.340 --> 00:08:33.699
Meaning,

291
00:08:34.121 --> 00:08:35.062
when you've lost money,

292
00:08:35.484 --> 00:08:37.281
your reference point has shifted down.

293
00:08:37.726 --> 00:08:39.164
You might become more risk-seeking,

294
00:08:39.281 --> 00:08:41.445
specifically for gambles that offer a chance,

295
00:08:41.547 --> 00:08:42.445
even a small one,

296
00:08:42.469 --> 00:08:43.344
to get back to even.

297
00:08:43.750 --> 00:08:43.984
Ah,

298
00:08:44.250 --> 00:08:45.390
the break-even effect.

299
00:08:45.953 --> 00:08:48.875
So investors might pile into lottery stocks after losses,

300
00:08:49.281 --> 00:08:50.500
hoping for that quick win.

301
00:08:50.656 --> 00:08:51.359
That's the idea.

302
00:08:51.984 --> 00:08:55.094
This extra demand could push prices up temporarily.

303
00:08:55.505 --> 00:08:56.886
above fundamental value,

304
00:08:57.226 --> 00:08:59.369
leading to lower subsequent returns.

305
00:08:59.748 --> 00:09:00.388
And conversely,

306
00:09:00.431 --> 00:09:01.510
if you're sitting on gains.

307
00:09:01.970 --> 00:09:03.111
If you're in the gain domain,

308
00:09:03.131 --> 00:09:04.513
you might be more risk-averse,

309
00:09:04.873 --> 00:09:06.099
especially toward volatile,

310
00:09:06.334 --> 00:09:07.193
skewed stocks.

311
00:09:07.576 --> 00:09:08.717
You want to protect your winnings,

312
00:09:08.779 --> 00:09:09.178
perhaps,

313
00:09:09.179 --> 00:09:12.646
so you might avoid lottery stocks or demand a higher return for holding them.

314
00:09:13.099 --> 00:09:17.990
Which could explain why they don't underperform or even outperform in the high CGO state.

315
00:09:18.053 --> 00:09:18.443
Exactly.

316
00:09:18.506 --> 00:09:20.787
It fits that framework where our decisions aren't

317
00:09:20.921 --> 00:09:22.782
purely rational in the traditional sense,

318
00:09:22.843 --> 00:09:25.325
but influenced by whether we feel like we're winning or losing.

319
00:09:25.747 --> 00:09:30.872
It certainly provides a behavioral lens on why this particular anomaly seems to behave the way it does.

320
00:09:31.489 --> 00:09:35.075
Really fascinating link between market outcomes and investor psychology.

321
00:09:35.458 --> 00:09:35.935
Absolutely.

322
00:09:36.013 --> 00:09:38.279
It suggests these anomalies aren't static things,

323
00:09:38.560 --> 00:09:42.763
but can be quite dynamic depending on the market's recent history and investor sentiment.

324
00:09:43.304 --> 00:09:45.949
Thank you for tuning in to Papers with Backtest podcast.

325
00:09:46.308 --> 00:09:48.533
We hope today's episode gave you useful insights.

326
00:09:49.096 --> 00:09:51.037
Join us next time as we break down more research.

327
00:09:51.455 --> 00:09:52.963
And for more papers and backtests,

328
00:09:53.017 --> 00:09:56.010
find us at https.paperswithbacktest.com.

329
00:09:56.650 --> 00:09:57.189
Happy trading.