AI Research ARMARM_news

ARM extreme daily moves vs news sentiment intensity

699
Total trading days

Contrary to the hypothesis that ARM’s steepest daily swings occur when news goes quiet, the data points the other way. Over 699 trading days, the average absolute return on the 7 high-sentiment days was 4.60% versus 3.32% on low-sentiment days. High-sentiment days also produced a larger share of the market’s extreme top‑10% moves.

But that picture rests on just a handful of observations. With a p‑value of 0.41 and a correlation of 0.08, the difference is too weak to be called a pattern. Below, we break down the numbers—bin comparisons, the full t‑test, and the scatter—so you can judge the noise for yourself.

The research question

For ARM over the past ~3 years, do its largest daily price swings happen on days with extreme news sentiment, or do the biggest moves cluster on quiet days? Thesis: The biggest moves land on low-sentiment days, implying that positioning and liquidity, not fresh news, drive the stock's extreme swings.

How this was measured

Daily absolute close-to-close returns from ARM minute bars. News sentiment intensity defined as the daily mean of absolute ticker_sentiment_score from ARM_news headlines; days without any news receive intensity = 0 (quietest days). All trading days are binned into terciles of intensity (Low/Mid/High). Mean absolute return and share of top‑10% extreme moves are compared across bins, supplemented by a Welch t‑test (Low vs High) and Pearson correlation.

The key numbers

Total trading days
699
ARM data since IPO, all sessions with price data
Low‑sentiment days
612
includes days with zero news
High‑sentiment days
7
Mean |return| – Low Sentiment
3.3222%
Mean |return| – High Sentiment
4.5991%
Mean diff (Low – High)
-1.2769%
Welch t‑statistic
-0.887
positive → larger moves on low‑sentiment days
Welch p‑value
0.4084
p=0.4084 ≥ 0.05 → not significant
Pearson r (|ret| vs |sent|)
0.080
linear correlation
Top‑10% days on Low Sentiment
8.9869%
share of low‑sentiment days that are extreme
Top‑10% days on High Sentiment
14.2857%

Reading the numbers

Over 699 trading days, high-sentiment days (7 total) had a higher mean absolute return (4.60%) than low-sentiment days (3.32%), but the difference is not statistically significant (p=0.41). However, nearly 1 in 7 high-sentiment days landed in the top 10% of returns, versus about 1 in 11 low-sentiment days — suggesting extreme moves are somewhat more likely on news-heavy days.

The charts

ARM absolute daily return by sentiment intensity tercile
What this chart says

This box plot shows the spread of ARM's daily absolute returns across three sentiment groups. The low-sentiment group (612 days) has a wide range, with a maximum swing of 27.7%, but its average return of 3.32% is the smallest. The high-sentiment group (only 7 days) has a narrower range but a higher average of 4.60%. This indicates that while a few extreme outliers occur on quiet news days, the typical daily move is actually larger when sentiment is elevated.

Mean absolute daily return by sentiment tercile
What this chart says

The bar chart directly compares the mean absolute return across sentiment terciles. The tallest bar is for high-sentiment days at 4.60%, followed by mid at 3.97% and low at 3.32%. This pattern contradicts the thesis that the biggest moves cluster on low-sentiment days—instead, the average price swing is larger on days with stronger news sentiment.

Summary by sentiment tercile

SentimentNMean |return|Median |return|Std |return|
Low Sentiment6120.03320.02320.0337
Mid Sentiment800.03970.02810.0371
High Sentiment70.0460.03130.0379

Extreme‑day share (top 10% |return|) by sentiment tercile

SentimentExtreme DaysTotal DaysShare
Low Sentiment556120.0899
Mid Sentiment14800.175
High Sentiment170.1429

The takeaway

The data does not support the thesis that ARM's biggest daily moves happen on quiet (low-sentiment) days. In fact, the average absolute return on high-sentiment days (4.60%) was larger than on low-sentiment days (3.32%), and the share of extreme top-10% moves was also higher — 14.3% vs 9.0%. But there were only 7 high-sentiment days in total, so the Welch t-test p-value of 0.41 tells us the difference is easily explainable by chance. The Pearson correlation between return size and sentiment intensity is a weak 0.08, barely above noise. Overall, this is an inconclusive result: the direction runs opposite to the hypothesis but the evidence is too thin to call it a real signal. The practical takeaway is that for ARM's short post-IPO history, extreme price swings don't reliably cluster on news-quiet days — if anything, they might be slightly more common when sentiment is intense, but you wouldn't bet on it.

The fine print