WEBVTT

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

2
00:00:00.400 --> 00:00:02.320
welcome back to Papers with Backtest podcast.

3
00:00:02.360 --> 00:00:02.700
Today,

4
00:00:03.260 --> 00:00:06.100
we dive into another algo trading research paper.

5
00:00:06.240 --> 00:00:06.600
We are.

6
00:00:06.660 --> 00:00:08.200
And this one looks at something pretty fundamental,

7
00:00:08.240 --> 00:00:08.400
right?

8
00:00:08.420 --> 00:00:09.160
Trading volume.

9
00:00:09.320 --> 00:00:09.760
Exactly.

10
00:00:09.900 --> 00:00:19.760
We're going to unpack the information hidden within unusual trading volume in the stock market based on the study abnormal volume effect in the stock market.

11
00:00:19.880 --> 00:00:20.980
It's a really interesting question,

12
00:00:21.040 --> 00:00:21.360
isn't it?

13
00:00:21.640 --> 00:00:23.380
We tend to focus so much on price,

14
00:00:24.060 --> 00:00:26.900
but what can those big shifts in trading activity actually,

15
00:00:27.440 --> 00:00:27.680
you know,

16
00:00:27.820 --> 00:00:28.180
tell us?

17
00:00:28.508 --> 00:00:28.688
Yeah.

18
00:00:28.788 --> 00:00:29.209
And for you,

19
00:00:29.309 --> 00:00:31.270
if you're thinking about systematic trading,

20
00:00:31.350 --> 00:00:36.755
understanding if these volume spikes are maybe a genuine early warning sign,

21
00:00:37.415 --> 00:00:37.535
well,

22
00:00:37.555 --> 00:00:38.676
that could be a powerful tool.

23
00:00:38.816 --> 00:00:39.176
Definitely.

24
00:00:39.737 --> 00:00:42.299
This specific paper looked at the Italian equity market.

25
00:00:42.579 --> 00:00:42.739
Right.

26
00:00:42.740 --> 00:00:46.082
A specific time frame to 1997 to 2003.

27
00:00:46.842 --> 00:00:47.803
So our goal today,

28
00:00:47.963 --> 00:00:57.150
our mission is to figure out if the significant kind of unexpected jumps in trading volume can give us a peek into future stock price movements.

29
00:00:57.624 --> 00:00:58.225
and crucially,

30
00:00:58.445 --> 00:01:01.947
if there's a way to potentially build actual trading strategies around them.

31
00:01:02.348 --> 00:01:02.548
Okay,

32
00:01:02.608 --> 00:01:03.168
let's dig in.

33
00:01:03.649 --> 00:01:05.730
The paper talks about abnormal volume.

34
00:01:07.212 --> 00:01:09.153
What exactly did that mean in their analysis?

35
00:01:09.213 --> 00:01:10.134
How did they define it?

36
00:01:10.975 --> 00:01:11.195
Well,

37
00:01:11.935 --> 00:01:13.396
they had a very specific definition.

38
00:01:13.416 --> 00:01:15.558
They weren't just looking at high volume days in general.

39
00:01:16.259 --> 00:01:16.619
Instead,

40
00:01:16.639 --> 00:01:20.642
they defined abnormal as trading volume that deviated significantly.

41
00:01:21.723 --> 00:01:26.987
The threshold was more than 2.33 standard deviations from a stock's average trading volume.

42
00:01:27.496 --> 00:01:29.638
over the previous 66 trading days.

43
00:01:29.858 --> 00:01:31.219
66 days.

44
00:01:31.379 --> 00:01:31.619
Okay,

45
00:01:31.699 --> 00:01:33.481
so quite a look back period.

46
00:01:33.541 --> 00:01:33.661
Yeah,

47
00:01:33.961 --> 00:01:35.462
roughly three months of trading.

48
00:01:36.203 --> 00:01:36.863
Think of it like this.

49
00:01:37.884 --> 00:01:39.606
We're not just looking for generally busy days,

50
00:01:39.846 --> 00:01:44.730
but for days where the trading activity is just so far outside the norm for that specific stock.

51
00:01:44.870 --> 00:01:45.170
Right,

52
00:01:45.430 --> 00:01:47.492
suggesting something unusual might be happening.

53
00:01:47.732 --> 00:01:48.132
Exactly.

54
00:01:48.212 --> 00:01:50.554
So it's about a real change in the usual pattern,

55
00:01:50.814 --> 00:01:52.876
not just consistently high activity.

56
00:01:53.416 --> 00:01:54.017
That makes sense.

57
00:01:54.057 --> 00:01:54.417
It does.

58
00:01:54.837 --> 00:01:56.539
Because a stock that always trades heavily

59
00:01:57.032 --> 00:01:58.673
Having another high volume day.

60
00:01:59.034 --> 00:01:59.174
Well,

61
00:01:59.514 --> 00:02:00.355
maybe that's not news.

62
00:02:00.575 --> 00:02:02.857
But a quiet stock suddenly seeing a huge spike,

63
00:02:02.917 --> 00:02:03.537
that's different.

64
00:02:03.557 --> 00:02:04.518
That's when your ears perk up,

65
00:02:04.578 --> 00:02:04.758
yeah.

66
00:02:05.359 --> 00:02:06.660
And they use a specific metric,

67
00:02:06.780 --> 00:02:08.201
normalized abnormal volume,

68
00:02:08.341 --> 00:02:08.541
NAV,

69
00:02:08.741 --> 00:02:09.342
to measure this.

70
00:02:09.722 --> 00:02:09.922
NAV.

71
00:02:10.483 --> 00:02:10.703
Okay,

72
00:02:10.763 --> 00:02:12.824
and that helped compare across different stocks.

73
00:02:13.065 --> 00:02:13.545
Precisely.

74
00:02:13.965 --> 00:02:18.269
It let them compare the unusualness of volume spikes across different companies.

75
00:02:18.709 --> 00:02:21.531
Even if one normally trades millions of shares and another only,

76
00:02:22.012 --> 00:02:22.232
you know,

77
00:02:22.233 --> 00:02:22.912
a few thousand,

78
00:02:23.213 --> 00:02:24.554
it standardized the deviation.

79
00:02:24.714 --> 00:02:25.014
Got it.

80
00:02:25.320 --> 00:02:29.443
And what's fascinating here is that by focusing on these really significant deviations,

81
00:02:30.184 --> 00:02:35.588
the study aimed to isolate trading activity that was likely driven by some sort of new information.

82
00:02:36.289 --> 00:02:36.409
OK,

83
00:02:36.549 --> 00:02:39.751
so they pinpointed these moments of abnormal volume using NAV.

84
00:02:40.492 --> 00:02:41.813
What did they discover then?

85
00:02:41.893 --> 00:02:45.976
How did stock prices behave around these events?

86
00:02:46.116 --> 00:02:46.256
Well,

87
00:02:46.257 --> 00:02:47.958
the core finding was quite remarkable,

88
00:02:48.018 --> 00:02:48.298
actually.

89
00:02:48.716 --> 00:02:54.080
They observed strong positive excess returns clustered around these days of extreme trading volumes.

90
00:02:54.160 --> 00:02:55.601
Strong positive returns.

91
00:02:55.702 --> 00:02:55.882
Yeah.

92
00:02:55.922 --> 00:02:59.124
When the NAV went over that 2.33 standard deviation mark,

93
00:02:59.565 --> 00:03:02.907
they saw an average abnormal return of 2.48%

94
00:03:03.067 --> 00:03:05.009
on the very day of the abnormal volume.

95
00:03:05.129 --> 00:03:06.390
2.48%.

96
00:03:06.410 --> 00:03:07.190
That's significant.

97
00:03:07.231 --> 00:03:07.511
It is.

98
00:03:07.512 --> 00:03:08.351
And an additional

99
00:03:09.072 --> 00:03:10.333
1.02% the following day.

100
00:03:10.493 --> 00:03:10.753
Wow.

101
00:03:10.873 --> 00:03:12.034
And these are excess returns,

102
00:03:12.094 --> 00:03:12.274
right?

103
00:03:12.495 --> 00:03:13.095
Mark and Justin.

104
00:03:13.288 --> 00:03:13.668
Exactly.

105
00:03:13.768 --> 00:03:16.350
Above and beyond what the overall stock market did on those days.

106
00:03:16.751 --> 00:03:18.792
So it's genuine outperformance.

107
00:03:19.013 --> 00:03:21.274
So what did they think was driving this?

108
00:03:21.334 --> 00:03:22.155
Was it simply,

109
00:03:22.976 --> 00:03:23.236
you know,

110
00:03:23.237 --> 00:03:26.518
the mechanics of lots of trading pushing the price up temporarily?

111
00:03:26.779 --> 00:03:27.960
That's a logical first thought.

112
00:03:28.000 --> 00:03:28.140
Yeah.

113
00:03:28.360 --> 00:03:33.024
And the researchers did consider if increased liquidity was the main factor.

114
00:03:33.664 --> 00:03:39.909
But they noted that prices didn't tend to fall back or reverse in the days immediately after the abnormal volume.

115
00:03:40.309 --> 00:03:40.629
Ah,

116
00:03:40.990 --> 00:03:41.110
OK.

117
00:03:41.530 --> 00:03:42.311
No quick reversal.

118
00:03:42.631 --> 00:03:42.771
Right.

119
00:03:42.784 --> 00:03:42.904
Right.

120
00:03:42.905 --> 00:03:47.888
That lack of immediate reversal suggested it wasn't just temporary price pressure from the trading itself.

121
00:03:48.088 --> 00:03:48.869
So what else could it be?

122
00:03:49.429 --> 00:03:51.311
They lean towards the idea of price momentum.

123
00:03:51.471 --> 00:03:51.711
You know,

124
00:03:51.712 --> 00:03:54.893
the tendency for prices to keep moving in their current direction,

125
00:03:54.974 --> 00:03:55.914
at least in the short run.

126
00:03:55.934 --> 00:03:56.054
OK,

127
00:03:56.295 --> 00:03:56.795
momentum.

128
00:03:57.255 --> 00:04:01.639
Which suggests that the abnormal volume might indeed be acting as a signal,

129
00:04:02.159 --> 00:04:05.362
a signal of some underlying information pushing the price,

130
00:04:05.822 --> 00:04:08.004
perhaps information that hasn't become public yet.

131
00:04:08.104 --> 00:04:08.224
Now,

132
00:04:08.264 --> 00:04:10.466
that's the interesting part for anyone looking for an edge,

133
00:04:10.686 --> 00:04:10.906
right?

134
00:04:11.066 --> 00:04:12.187
Undisclosed information.

135
00:04:12.628 --> 00:04:13.028
Exactly.

136
00:04:13.428 --> 00:04:13.908
So naturally,

137
00:04:13.928 --> 00:04:14.668
they dug into that.

138
00:04:14.948 --> 00:04:19.788
They looked at whether these volume spikes were just happening when major news announcements hit the market.

139
00:04:19.908 --> 00:04:20.808
Which you'd expect,

140
00:04:20.868 --> 00:04:21.248
wouldn't you?

141
00:04:21.528 --> 00:04:22.148
News breaks,

142
00:04:22.208 --> 00:04:23.028
volume spikes.

143
00:04:23.068 --> 00:04:23.548
You would.

144
00:04:24.148 --> 00:04:24.448
And yeah,

145
00:04:24.468 --> 00:04:29.768
the study did confirm some abnormal volume events did coincide with news releases.

146
00:04:29.968 --> 00:04:30.708
No surprise there.

147
00:04:31.188 --> 00:04:32.928
And what happened then with news?

148
00:04:33.248 --> 00:04:34.288
Pretty much what you'd expect.

149
00:04:34.948 --> 00:04:35.688
Positive news,

150
00:04:35.868 --> 00:04:37.308
generally positive market reaction.

151
00:04:37.908 --> 00:04:38.428
Negative news,

152
00:04:38.748 --> 00:04:39.428
negative reaction.

153
00:04:39.608 --> 00:04:39.768
Okay.

154
00:04:40.196 --> 00:04:40.656
Interestingly,

155
00:04:40.716 --> 00:04:44.700
they also found some signs of underreaction,

156
00:04:45.260 --> 00:04:46.261
especially to bad news.

157
00:04:46.781 --> 00:04:50.404
Prices sometimes continue to drift down even after the negative announcement.

158
00:04:50.985 --> 00:04:52.966
Price continuation even with news.

159
00:04:53.006 --> 00:04:53.166
Yeah.

160
00:04:53.587 --> 00:04:54.687
But here's the key thing,

161
00:04:54.688 --> 00:04:56.068
the really crucial takeaway.

162
00:04:56.209 --> 00:04:56.549
Go on.

163
00:04:56.809 --> 00:04:59.131
The majority of these abnormal volume events,

164
00:04:59.431 --> 00:05:00.132
the larger portion,

165
00:05:00.152 --> 00:05:03.054
did not coincide with any major news being released at the same time.

166
00:05:03.274 --> 00:05:03.594
Really?

167
00:05:03.774 --> 00:05:04.455
The majority?

168
00:05:04.615 --> 00:05:04.895
Yes.

169
00:05:05.436 --> 00:05:07.678
So these were unexplained volume surges.

170
00:05:08.175 --> 00:05:12.058
And they still observe those significant excess returns around them.

171
00:05:12.358 --> 00:05:12.619
Wow.

172
00:05:12.879 --> 00:05:12.999
OK,

173
00:05:13.099 --> 00:05:13.659
that's big.

174
00:05:14.140 --> 00:05:17.582
That strongly suggests the volume itself carries information,

175
00:05:17.622 --> 00:05:17.983
doesn't it?

176
00:05:18.443 --> 00:05:20.285
Even without a public news catalyst.

177
00:05:20.325 --> 00:05:21.305
That's the interpretation.

178
00:05:21.365 --> 00:05:21.525
Yeah.

179
00:05:21.826 --> 00:05:23.307
It opens up the possibility that,

180
00:05:23.547 --> 00:05:23.787
you know,

181
00:05:24.148 --> 00:05:30.312
informed traders might be acting on non-public information and their trading is what's causing these volume patterns.

182
00:05:30.513 --> 00:05:30.633
OK,

183
00:05:30.833 --> 00:05:32.194
so if it's not always news,

184
00:05:32.274 --> 00:05:34.315
maybe it relates to the type of company involved.

185
00:05:34.799 --> 00:05:38.842
Did they look into whether certain firm characteristics made a difference?

186
00:05:38.962 --> 00:05:39.303
They did.

187
00:05:39.663 --> 00:05:46.428
They investigated if features of the company influenced this relationship between abnormal volume and the excess returns.

188
00:05:46.969 --> 00:05:47.589
What popped out?

189
00:05:48.190 --> 00:05:50.071
Ownership and control seemed quite relevant.

190
00:05:50.471 --> 00:05:51.052
For instance,

191
00:05:51.452 --> 00:05:56.716
higher levels of control by the ultimate shareholder were linked to larger excess returns,

192
00:05:56.756 --> 00:05:57.997
both positive and negative,

193
00:05:58.057 --> 00:05:59.839
around these abnormal volume events.

194
00:06:00.099 --> 00:06:02.881
So concentrated control amplified the effect.

195
00:06:02.961 --> 00:06:03.541
It appears so.

196
00:06:04.259 --> 00:06:05.540
And family-controlled firms,

197
00:06:05.820 --> 00:06:11.225
they showed a more pronounced negative price drop when abnormal volumes seemed linked to negative information.

198
00:06:11.445 --> 00:06:11.905
Interesting.

199
00:06:12.145 --> 00:06:15.248
What about things like pyramidal structures or non-voting shares,

200
00:06:15.428 --> 00:06:17.289
ways to separate ownership from control?

201
00:06:17.549 --> 00:06:17.690
Yeah,

202
00:06:17.710 --> 00:06:18.390
they looked at that too,

203
00:06:18.410 --> 00:06:20.692
but the results there weren't as clear-cut.

204
00:06:20.892 --> 00:06:24.395
They didn't find a statistically significant effect for that separation,

205
00:06:24.675 --> 00:06:25.616
at least in this dataset.

206
00:06:25.916 --> 00:06:26.156
Okay.

207
00:06:26.697 --> 00:06:28.398
Any other company features that mattered,

208
00:06:28.518 --> 00:06:29.999
like size or valuation?

209
00:06:30.500 --> 00:06:30.740
Yes.

210
00:06:31.219 --> 00:06:35.382
Smaller firms tended to show larger excess returns around these events.

211
00:06:35.482 --> 00:06:36.403
That kind of makes sense.

212
00:06:36.543 --> 00:06:38.465
Maybe information has a bigger impact there.

213
00:06:38.525 --> 00:06:38.965
Possibly,

214
00:06:39.125 --> 00:06:39.285
yeah.

215
00:06:39.826 --> 00:06:40.086
Also,

216
00:06:40.466 --> 00:06:43.148
firms with higher market-to-book ratios,

217
00:06:43.188 --> 00:06:44.189
sort of growth stocks,

218
00:06:44.630 --> 00:06:47.192
you might say they also exhibited larger excess returns.

219
00:06:47.292 --> 00:06:47.472
Okay.

220
00:06:47.932 --> 00:06:48.713
And liquidity.

221
00:06:49.033 --> 00:06:52.896
Less liquid stocks showed a stronger price impact from abnormal volume.

222
00:06:53.717 --> 00:06:54.017
Again,

223
00:06:54.077 --> 00:06:54.978
perhaps intuitive,

224
00:06:55.778 --> 00:06:59.661
a big surge in trading might move the needle more if the stock doesn't usually trade much.

225
00:06:59.841 --> 00:07:00.122
Right.

226
00:07:00.479 --> 00:07:01.160
Anything else?

227
00:07:01.320 --> 00:07:02.421
Just one more they highlighted.

228
00:07:02.741 --> 00:07:06.724
Higher dividend yield was actually negatively associated with these excess returns,

229
00:07:07.024 --> 00:07:10.127
though they caution that might be partly due to industry effects,

230
00:07:10.507 --> 00:07:10.747
you know,

231
00:07:10.767 --> 00:07:12.589
certain types of companies paying higher dividends.

232
00:07:12.869 --> 00:07:13.189
Got it.

233
00:07:13.769 --> 00:07:19.274
So company characteristics definitely seem to play a role in how informative these volume spikes are.

234
00:07:19.334 --> 00:07:19.914
Seems that way,

235
00:07:19.954 --> 00:07:20.195
yes.

236
00:07:20.575 --> 00:07:20.695
OK,

237
00:07:20.775 --> 00:07:22.116
so let's get to the practical side.

238
00:07:22.737 --> 00:07:23.437
For you listening,

239
00:07:23.597 --> 00:07:24.498
thinking about trading,

240
00:07:25.499 --> 00:07:27.740
did the paper actually test?

241
00:07:28.407 --> 00:07:31.069
If you could create a profitable strategy from this,

242
00:07:31.710 --> 00:07:32.971
what did the backtesting show?

243
00:07:33.651 --> 00:07:34.492
They did explore that,

244
00:07:34.552 --> 00:07:34.692
yeah.

245
00:07:34.912 --> 00:07:37.774
They looked at a pretty basic trading strategy concept.

246
00:07:38.015 --> 00:07:38.635
How did it work?

247
00:07:39.095 --> 00:07:40.096
It was quite intuitive,

248
00:07:40.176 --> 00:07:40.477
actually.

249
00:07:40.517 --> 00:07:43.459
You see a day with abnormal volume,

250
00:07:44.039 --> 00:07:44.259
right,

251
00:07:44.600 --> 00:07:46.621
using that NAV 2.33 signal.

252
00:07:46.821 --> 00:07:47.082
Okay.

253
00:07:47.362 --> 00:07:49.323
Then you look at the price movement on that same day,

254
00:07:49.403 --> 00:07:50.024
the event day.

255
00:07:50.324 --> 00:07:51.745
And use that as a direction signal.

256
00:07:51.865 --> 00:07:52.266
Exactly.

257
00:07:52.306 --> 00:07:53.667
The hypothetical strategy was,

258
00:07:54.307 --> 00:07:55.068
by the next day,

259
00:07:55.188 --> 00:07:57.490
if the price had increased on the abnormal volume day,

260
00:07:57.679 --> 00:07:58.920
and sell or go short,

261
00:07:59.440 --> 00:08:01.522
the price had decreased on that abnormal volume day.

262
00:08:01.622 --> 00:08:01.802
Okay,

263
00:08:01.842 --> 00:08:02.663
makes logical sense.

264
00:08:02.683 --> 00:08:03.904
You're trying to ride that momentum,

265
00:08:03.964 --> 00:08:07.567
assuming the event day price move reflected the direction of the hidden information.

266
00:08:07.767 --> 00:08:09.708
So did it work in the back test?

267
00:08:10.009 --> 00:08:11.510
The results gave some interesting insights,

268
00:08:11.550 --> 00:08:13.511
but also a healthy dose of reality.

269
00:08:13.712 --> 00:08:14.092
Uh-oh,

270
00:08:14.352 --> 00:08:14.652
how so?

271
00:08:14.953 --> 00:08:15.373
First,

272
00:08:15.573 --> 00:08:16.954
they tested a really simple strategy.

273
00:08:17.334 --> 00:08:19.976
Just buy after any abnormal volume spike,

274
00:08:20.537 --> 00:08:21.958
regardless of the price move that day.

275
00:08:22.098 --> 00:08:23.479
Just based on volume alone?

276
00:08:23.900 --> 00:08:24.080
Yeah.

277
00:08:24.599 --> 00:08:30.659
And that long-only strategy actually showed negative returns once they factored in estimated bid-ask spreads.

278
00:08:31.039 --> 00:08:31.379
Ouch.

279
00:08:31.779 --> 00:08:33.699
So transaction costs killed it.

280
00:08:34.039 --> 00:08:34.419
Pretty much.

281
00:08:35.179 --> 00:08:40.699
Just seeing a volume spike on its own wasn't a reliable buy signal once you account for the cost of trading.

282
00:08:41.379 --> 00:08:41.999
At least not here.

283
00:08:42.259 --> 00:08:42.459
Okay,

284
00:08:42.579 --> 00:08:44.079
so that simple approach is out.

285
00:08:44.499 --> 00:08:46.839
What about the one incorporating the price direction?

286
00:08:47.359 --> 00:08:48.699
Buy after a positive day,

287
00:08:49.019 --> 00:08:50.219
short after a negative day.

288
00:08:50.459 --> 00:08:50.679
Right.

289
00:08:51.079 --> 00:08:51.939
When they tested that,

290
00:08:52.019 --> 00:08:53.019
things looked a bit different.

291
00:08:53.715 --> 00:08:57.898
For the long positions buying after a positive return on the abnormal volume day,

292
00:08:58.319 --> 00:08:59.840
they did observe some profitability.

293
00:09:00.380 --> 00:09:01.841
Even after transaction costs?

294
00:09:01.961 --> 00:09:02.222
Yes,

295
00:09:02.402 --> 00:09:05.204
even after accounting for those estimated bid-ask spreads.

296
00:09:06.085 --> 00:09:11.729
And this potential profitability was more noticeable when they focused on larger price movements on that event day.

297
00:09:12.390 --> 00:09:14.852
They looked specifically at thresholds like a 3%

298
00:09:15.092 --> 00:09:16.012
or even a 5%

299
00:09:16.133 --> 00:09:18.214
price increase on the abnormal volume day.

300
00:09:18.374 --> 00:09:22.257
So abnormal volume plus a significant positive price move on that day.

301
00:09:22.915 --> 00:09:26.375
Buying the next day showed some potential edge for longs.

302
00:09:26.535 --> 00:09:28.915
That's what the backtest suggested for this period in marketing,

303
00:09:28.935 --> 00:09:29.155
yes.

304
00:09:29.555 --> 00:09:29.975
Interesting.

305
00:09:30.335 --> 00:09:31.375
What about the other side?

306
00:09:31.835 --> 00:09:32.435
Short selling.

307
00:09:32.775 --> 00:09:35.595
Selling after abnormal volume and a negative price move.

308
00:09:35.695 --> 00:09:35.835
Ah,

309
00:09:36.215 --> 00:09:36.395
well,

310
00:09:36.475 --> 00:09:37.835
that's where it didn't pan out so well.

311
00:09:38.475 --> 00:09:40.395
The short selling strategies they tested,

312
00:09:41.515 --> 00:09:42.595
based on that same principle,

313
00:09:43.195 --> 00:09:45.975
they ended up resulting in losses after transaction costs.

314
00:09:46.055 --> 00:09:47.455
Losses even for the short side?

315
00:09:47.775 --> 00:09:48.495
Why might that be?

316
00:09:49.115 --> 00:09:50.155
Could be several reasons.

317
00:09:50.575 --> 00:09:52.395
Short selling often has higher costs.

318
00:09:52.623 --> 00:09:52.843
you know,

319
00:09:52.844 --> 00:09:53.843
borrowing fees and such.

320
00:09:54.143 --> 00:09:54.383
Plus,

321
00:09:54.523 --> 00:09:56.703
there are often more restrictions or risks involved.

322
00:09:56.783 --> 00:09:57.063
Right.

323
00:09:57.503 --> 00:10:00.223
And maybe just market dynamics during that period in Italy,

324
00:10:00.663 --> 00:10:03.963
perhaps an upward bias or specific short-selling rules.

325
00:10:04.623 --> 00:10:07.183
The paper didn't pinpoint one exact reason,

326
00:10:07.603 --> 00:10:08.583
but the result was clear.

327
00:10:09.404 --> 00:10:13.087
Shorting based on the signal lost money after costs in their test.

328
00:10:13.307 --> 00:10:13.487
Right.

329
00:10:13.507 --> 00:10:15.569
Those market specific factors are always lurking,

330
00:10:15.609 --> 00:10:15.929
aren't they?

331
00:10:16.370 --> 00:10:20.373
So while the paper found these really intriguing patterns around abnormal volume.

332
00:10:20.393 --> 00:10:20.513
Yeah,

333
00:10:20.514 --> 00:10:22.114
the information content seems to be there.

334
00:10:22.274 --> 00:10:26.337
But turning it into a consistently profitable real world system,

335
00:10:26.477 --> 00:10:28.939
especially considering both long and short sides,

336
00:10:29.320 --> 00:10:32.602
sounds like it's much tougher because of those unavoidable trading costs.

337
00:10:32.962 --> 00:10:34.644
That's a really crucial point to understand.

338
00:10:34.724 --> 00:10:35.184
Absolutely.

339
00:10:35.805 --> 00:10:38.527
The theoretical signals might look promising on paper.

340
00:10:39.112 --> 00:10:41.574
But the practicalities of actually trading,

341
00:10:41.974 --> 00:10:42.955
especially the costs,

342
00:10:43.355 --> 00:10:43.756
friction,

343
00:10:45.157 --> 00:10:47.559
they can really change the bottom line significantly.

344
00:10:47.599 --> 00:10:47.719
OK,

345
00:10:47.919 --> 00:10:50.421
so let's try and summarize the key practical takeaways then.

346
00:10:50.721 --> 00:10:54.224
For anyone listening who's thinking about trading rules based on this kind of idea,

347
00:10:54.964 --> 00:10:56.706
what should they keep top of mind?

348
00:10:57.106 --> 00:10:58.327
I think the main thing is,

349
00:10:59.368 --> 00:10:59.688
yes,

350
00:11:00.249 --> 00:11:01.770
abnormal trading volume,

351
00:11:01.830 --> 00:11:03.071
especially in this study,

352
00:11:03.631 --> 00:11:07.174
did seem to contain information and precede short term price moves.

353
00:11:07.214 --> 00:11:07.334
OK,

354
00:11:07.494 --> 00:11:08.375
information is there.

355
00:11:08.816 --> 00:11:16.842
And that information seemed clearer or perhaps more actionable when you combine the volume signal with the direction of the price change on that same high volume day.

356
00:11:16.902 --> 00:11:17.143
Right,

357
00:11:17.144 --> 00:11:18.143
the combined signal.

358
00:11:19.224 --> 00:11:23.147
But the profitability of any strategy built purely on this is heavily,

359
00:11:23.267 --> 00:11:25.149
heavily impacted by transaction costs.

360
00:11:25.209 --> 00:11:25.549
Got it.

361
00:11:25.569 --> 00:11:26.170
Costs are key.

362
00:11:26.530 --> 00:11:34.036
Their backtesting suggested maybe some potential for long positions if you focused on significant positive price moves on those abnormal volume days.

363
00:11:34.316 --> 00:11:36.718
But short selling was a no-go after costs.

364
00:11:37.058 --> 00:11:37.378
Correct.

365
00:11:37.672 --> 00:11:38.513
based on their findings.

366
00:11:38.514 --> 00:11:42.296
So it's definitely not some kind of guaranteed path to easy profits.

367
00:11:42.336 --> 00:11:43.356
No magic bullet then?

368
00:11:43.597 --> 00:11:44.397
Definitely not.

369
00:11:45.058 --> 00:11:48.300
But it does highlight that unusual trading activity,

370
00:11:48.380 --> 00:11:51.643
that volume itself can be more than just noise.

371
00:11:52.584 --> 00:11:54.625
It can carry real information that gets reflected,

372
00:11:54.626 --> 00:11:55.386
at least briefly,

373
00:11:55.806 --> 00:11:56.367
in prices.

374
00:11:56.507 --> 00:11:56.707
Yeah,

375
00:11:56.787 --> 00:12:00.030
it adds another dimension to think about when you're analyzing market behavior,

376
00:12:00.150 --> 00:12:02.171
beyond just looking at price charts.

377
00:12:02.211 --> 00:12:02.792
Precisely.

378
00:12:02.912 --> 00:12:06.495
It underscores that idea that volume isn't just a consequence of trading.

379
00:12:06.956 --> 00:12:09.678
Sometimes it might be an important signal in its own right.

380
00:12:09.979 --> 00:12:10.799
Great insights.

381
00:12:10.800 --> 00:12:13.842
A lot to think about there regarding volume and potential signals.

382
00:12:13.962 --> 00:12:14.842
Definitely food for thought.

383
00:12:15.143 --> 00:12:17.765
Thank you for tuning in to Papers with Backtest podcast.

384
00:12:18.185 --> 00:12:20.667
We hope today's episode gave you useful insights.

385
00:12:21.187 --> 00:12:23.389
Join us next time as we break down more research.

386
00:12:23.849 --> 00:12:25.391
And for more papers and backtests,

387
00:12:25.531 --> 00:12:28.513
find us at https.paperswithbacktest.com.

388
00:12:29.113 --> 00:12:29.714
Happy trading.