EOG next-day return conditional on daily news sentiment quintile
Bullish headlines didn't lead to a next-day pop in EOG. Over 378 trading days, the stock averaged a 0.47% gain the day after its most negative news sentiment days — and a 0.13% loss after its most positive days. That flips the usual momentum story on its head.
The test was straightforward: rank every day by a composite news sentiment score, split into five quintiles, and measure the close-to-close return the next session. The pattern steps down almost in order as sentiment rises, and the top-versus-bottom gap is roughly -0.60% with a t-stat of -1.98. The p-value sits right at 0.049, so this is a lean, not a conclusive verdict.
The full breakdown, including quintile means, methodology, and caveats, is laid out in the analysis below.
When EOG's daily news sentiment score is in its top quintile, does the stock deliver above-average next-day returns? I expect bullish headlines to attract momentum buyers the following day, causing a short-term rally.
How this was measured
EOG daily close-to-close returns and next-day returns from resampled minute bars. Daily news sentiment is the mean of headline-level ticker_sentiment_score (fallback overall_sentiment_score) aggregated to calendar day. Days with both price and sentiment are ranked by sentiment into quintiles (1=lowest, 5=highest). Next-day returns are compared across quintiles, and a Welch two-sample t-test contrasts the top and bottom quintile means to assess if the most bullish days deliver statistically distinct next-day returns.
The key numbers
Reading the numbers
Across 378 days, the highest-sentiment quintile averaged -0.13% next-day return versus +0.47% for the lowest-sentiment quintile. That -0.60% gap is statistically significant (p=0.049), so the expected bullish-headline bounce does not appear—if anything, the pattern runs the other way.
The charts
The box plot spreads out next-day returns for each sentiment quintile, so you can see both the averages and the ranges behind them. The highest-sentiment group (5) has a mean return of -0.13% and reaches as low as -7.46%, while the lowest-sentiment group (1) has a mean of +0.47% and reaches +4.83%. The rightmost box, for the most bullish headlines, is not the one sitting highest on the chart—it is the only one with a negative mean. That visual pattern is the opposite of what a momentum-buyer story would predict.
The bars show the mean next-day return falling from +0.47% in the lowest-sentiment quintile to -0.13% in the highest-sentiment quintile. The last bar is the one to focus on: days with the most bullish sentiment actually average a slightly negative next-day return. The top-minus-bottom gap is -0.60%, and with p=0.049 it is unlikely to be just random noise. So this chart directly answers the question—bullish headlines do not appear to attract the next-day rally that was hypothesized.
Next-day return summary by quintile
| Quintile | Mean Return | Std Return | N days |
|---|---|---|---|
| 1 | 0.0047 | 0.015 | 76 |
| 2 | 0.0013 | 0.0224 | 75 |
| 3 | 0.0012 | 0.0178 | 76 |
| 4 | 0.0004 | 0.0186 | 75 |
| 5 | -0.0013 | 0.0215 | 76 |
The takeaway
No — the data show the opposite of a momentum effect. Over 378 days, EOG's top-quintile news-sentiment days were followed by an average next-day return of -0.13%, while bottom-quintile days averaged +0.47%. That's a gap of roughly -0.60%, with a t-stat of -1.98 and a p-value of 0.049, meaning the chance this is just luck sits right around 5%. The quintile means step down almost monotonically as sentiment rises (0.47%, 0.13%, 0.12%, 0.04%, -0.13%), so the relationship looks inverse, not positive. Because the p-value is borderline and each bucket only has ~76 days, this is a lean, not a slam dunk. Practical takeaway: bullish headlines didn't attract next-day momentum buyers; if anything, they marked a mild fade.
The fine print
- News timestamps may lag publication, so sentiment can reflect headline-driven moves that already happened; this design misses intraday response.
- Quintile cutoffs come from the full sample; a true out-of-sample test could find a weaker version of this result.
- EOG is an energy name; sentiment may be proxying crude oil or natural gas news rather than company-specific headlines.
- The p-value is right at 0.05; a few extreme days could flip the sign, so treat it as suggestive rather than proven.