BP top-quintile news sentiment with and without Brent tailwind: next-morning gap and open drift
Top-quintile BP news sentiment looks like a simple bullish signal, until you split the tape by what Brent did that same day. Across roughly three years, the combination of strong headlines and a crude tailwind produced a next-morning open gap of just 0.02%, against 0.22% for sentiment-only days. That gap difference is statistically indistinguishable from noise. The pattern only sharpens once the regular session starts: combo days drifted -0.30% from the open, while sentiment-only days gained +0.54%, a swing with about a 2.6% chance of being random.
The full analysis breaks down the event sets, the t-tests, and the daily distributions behind those numbers. Bottom line ahead of the details: for BP, crude does not amplify the opening pop — it flips the post-open tape.
For BP over the past ~3 years, does a top-quintile daily news-sentiment score on a day when Brent crude also rises produce a larger next-morning open gap and a more negative open-to-close drift than a top-quintile sentiment day alone? I expect the combination of bullish headlines and crude tailwind to be front-run into the opening gap, leaving fast-money fade from the open.
How this was measured
Daily BP OHLC was built from NY-regular-session minute bars, and daily news sentiment was computed from BP_news by averaging ticker_sentiment_score per day (falling back to overall score). Days whose daily sentiment met the top-quintile threshold were split by whether the same-day Brent crude return was positive. For each event day, the next-morning open gap is open[t+1]/close[t]-1 and the open-to-close drift is close[t+1]/open[t+1]-1. The two groups are compared with Welch unequal-variance t-tests.
The key numbers
Reading the numbers
Combo days averaged a +0.02% next-morning gap versus +0.22% for sentiment-only days, and a -0.30% open-to-close drift versus +0.54%. Only the drift difference is statistically clear (p=0.026); the gap difference is not (p=0.672).
The charts
This bar chart puts the two outcome metrics side by side for both event types. On the left, the next-morning gap is actually smaller for the combo days (+0.02%) than for sentiment-only days (+0.22%), so adding a Brent rise did not produce a bigger opening pop. On the right, the open-to-close drift flips from positive for sentiment-only (+0.54%) to negative for the combo (-0.30%), which is where the Brent tailwind seems to matter.
The gap distributions overlap heavily: combo days span from -6.37% to +2.75% with a mean of +0.02%, while sentiment-only days span from -1.70% to +2.64% with a mean of +0.22%. The combo group actually has a longer negative tail, the opposite of what a 'larger opening gap' would look like. With such overlapping spreads and a p-value of 0.67, this chart says the combo does not produce a clearly bigger gap.
The open-to-close drift distributions separate more clearly. Combo days average -0.30% and range from -3.29% to +1.94%, while sentiment-only days average +0.54% and range from -1.25% to +2.84%. The combo mass sits lower, matching the idea that bullish headlines plus a Brent tailwind get faded from the open; the p-value of 0.026 says this difference is unlikely to be random noise, though the sample is small at 24 events.
Event-type return summary
| Metric | Group | N | Mean | Median | Std |
|---|---|---|---|---|---|
| Next-morning open gap | Top sentiment + Brent up | 24 | 0.0002 | 0.0042 | 0.0196 |
| Next-morning open gap | Top sentiment only | 20 | 0.0022 | 0.0011 | 0.0117 |
| Open-to-close drift | Top sentiment + Brent up | 24 | -0.003 | 0.0009 | 0.0134 |
| Open-to-close drift | Top sentiment only | 20 | 0.0054 | 0.0053 | 0.0106 |
The takeaway
The short answer is no for the gap, yes for the fade. On BP days with top-quintile news sentiment (threshold ~0.31), when Brent was also up the next open averaged just +0.02% versus +0.22% for sentiment-only days — the -0.20 percentage-point gap difference is squarely in noise territory (p=0.67). The open-to-close action tells a different story: combo days drifted -0.30% from the open while sentiment-only days gained +0.54%, a -0.84 percentage-point swing with p=0.026, or about a 2.6-in-100 chance of being pure luck. So the evidence backs the fade-from-the-open idea, not the front-running-into-the-gap idea. It's a lean, not a lock — just 24 vs 20 event days and a combo median drift near zero, so the mean is driven by a few rough closes. Practical read: for BP, the crude-plus-bullish-headline setup is less about where it opens and more about it giving back gains during the regular session.
The fine print
- Small sample: 24 combo vs 20 sentiment-only days, some overlapping in time; Welch t-tests are approximate here.
- Top-quintile threshold is the in-sample 80th percentile of daily sentiment, not pre-registered or out-of-sample.
- Daily sentiment is an average of all BP headlines with no relevance filter, so mixed-news days can read neutral.
- Brent up/down is same-day and coincident, not a lead; the combo drift average leans on a few negative outliers (median ~0.09%).