In this notebook, we use Alphalens to examine the Brain Sentiment Indicator dataset (BSI). This dataset captures sentiment from news sources.
Start by setting the bundle for this notebook:
from zipline.research import use_bundle
use_bundle("usstock-1d-bundle")
Next, we create a Pipeline that contains the BUZZ_VOLUME_SENTIMENT field using a 1-day lookback. This field quantifies news volume relative to the past. We limit the analysis to the top 1000 stocks by dollar volume.
from zipline.pipeline import Pipeline, brain, master
from zipline.pipeline.factors import AverageDollarVolume
# limit analysis to stocks
universe = master.SecuritiesMaster.SecType.latest.eq("STK")
avg_dollar_volume = AverageDollarVolume(window_length=90)
buzz_volume_sentiment = brain.BSI.slice(1).BUZZ_VOLUME_SENTIMENT.latest
pipeline = Pipeline(
columns={
"buzz_volume_sentiment": buzz_volume_sentiment,
},
initial_universe=universe,
screen=(
avg_dollar_volume.top(1000)
& buzz_volume_sentiment.notnull()
)
)
Next, we run the Pipeline and generate the tear sheet. Stocks with higher than normal news volume (quantile 5) underperform stocks in other quantiles.
import alphalens as al
factor_data = al.from_pipeline(
pipeline,
start_date="2016-08-01", # start of dataset
end_date="2024-03-31",
factor="buzz_volume_sentiment",
periods=[1, 21],
quantiles=5,
segment="Y"
)
| min | max | mean | std | count | avg daily count | count % | |
|---|---|---|---|---|---|---|---|
| Buzz Volume Sentiment Quantile | |||||||
| 1 | -1.689 | 2.335 | -0.488 | 0.328 | 208,426 | 108.2 | 20.1% |
| 2 | -0.889 | 4.351 | -0.133 | 0.416 | 207,218 | 107.6 | 20.0% |
| 3 | -0.686 | 5.654 | 0.198 | 0.560 | 207,228 | 107.6 | 20.0% |
| 4 | -0.535 | 8.723 | 0.688 | 0.779 | 207,197 | 107.6 | 20.0% |
| 5 | -0.373 | 148.870 | 2.458 | 2.703 | 207,959 | 108.0 | 20.0% |
| 1D | 21D | |
|---|---|---|
| Ann. alpha | -0.005 | -0.009 |
| beta | -0.004 | -0.003 |
| Mean Relative Return Top Quantile (bps) | -0.643 | -0.333 |
| Mean Relative Return Bottom Quantile (bps) | 0.046 | 0.265 |
| Mean Spread (bps) | -0.689 | -0.605 |
| 1D | 21D | |
|---|---|---|
| IC Mean | -0.002 | -0.005 |
| IC Std. | 0.063 | 0.060 |
| Risk-Adjusted IC | -0.032 | -0.081 |
| t-stat(IC) | -1.395 | -3.575 |
| p-value(IC) | 0.163 | 0.000 |
| IC Skew | 0.130 | -0.196 |
| IC Kurtosis | 0.492 | 0.467 |
| 1D | 21D | |
|---|---|---|
| Quantile 1 Mean Turnover | 0.684 | 0.745 |
| Quantile 2 Mean Turnover | 0.800 | 0.828 |
| Quantile 3 Mean Turnover | 0.823 | 0.846 |
| Quantile 4 Mean Turnover | 0.823 | 0.851 |
| Quantile 5 Mean Turnover | 0.709 | 0.836 |
| 1D | 21D | |
|---|---|---|
| Mean Factor Rank Autocorrelation | 0.331 | 0.136 |
