Jacob E. ChandranRésumé PDF ↗︎
← All insights

Stock sentiment

Where today’s tone sits in its own history

The RLTrade box plots compare each stock’s latest news or tweet tone with its own 30-day distribution. Select a stock to see the context behind the dot.

Latest Oct 15, 7:45 PM EDT

The RLTrade sentiment box plots

Each row compares the latest tone reading with that stock’s own last 30 days. Blue is news; orange is tweets. The dot shows the latest reading, and the number on the right is its approximate percentile.

Stock / ETFTone scorePercentile
−1Negative0Neutral+1Positive
QQQQQQ: 194 active 15-minute bars; 5th −0.305, Q1 −0.001, median +0.067, Q3 +0.283, 95th +0.475. Latest +0.063 at 2026-10-15T23:45:00+00:00. Percentile approximate.51stAAPLAAPL: 398 active 15-minute bars; 5th −0.240, Q1 −0.077, median +0.041, Q3 +0.176, 95th +0.445. Latest +0.038 at 2026-10-15T23:45:00+00:00. Percentile approximate.45thAMDAMD: 259 active 15-minute bars; 5th −0.447, Q1 −0.005, median +0.287, Q3 +0.475, 95th +0.627. Latest +0.312 at 2026-10-15T23:45:00+00:00. Percentile approximate.55thAMZNAMZN: 517 active 15-minute bars; 5th −0.023, Q1 +0.091, median +0.233, Q3 +0.354, 95th +0.512. Latest +0.153 at 2026-10-15T23:45:00+00:00. Percentile approximate.41stGOOGLGOOGL: 591 active 15-minute bars; 5th −0.115, Q1 +0.074, median +0.225, Q3 +0.362, 95th +0.543. Latest −0.075 at 2026-10-15T23:45:00+00:00. Percentile approximate.9thMETAMETA: 528 active 15-minute bars; 5th −0.165, Q1 +0.070, median +0.264, Q3 +0.386, 95th +0.565. Latest +0.184 at 2026-10-15T23:45:00+00:00. Percentile approximate.39thMSFTMSFT: 540 active 15-minute bars; 5th −0.053, Q1 +0.141, median +0.266, Q3 +0.357, 95th +0.604. Latest +0.352 at 2026-10-15T23:45:00+00:00. Percentile approximate.77thMUMU: 378 active 15-minute bars; 5th −0.354, Q1 −0.005, median +0.178, Q3 +0.339, 95th +0.571. Latest +0.114 at 2026-10-15T23:45:00+00:00. Percentile approximate.43rdNVDANVDA: 791 active 15-minute bars; 5th +0.010, Q1 +0.210, median +0.323, Q3 +0.436, 95th +0.546. Latest +0.321 at 2026-10-15T23:45:00+00:00. Percentile approximate.52ndTSLATSLA: 352 active 15-minute bars; 5th −0.577, Q1 −0.371, median −0.097, Q3 +0.097, 95th +0.301. Latest −0.184 at 2026-10-15T23:45:00+00:00. Percentile approximate.45th

Box: middle 50% · line: median · whiskers: 5th–95th · dot: latest

Each row is a stock or ETF; the horizontal scale is its sentiment score. Percentile compares the latest reading with that same ticker’s available 15-minute readings from the past 30 days. For example, the 15th percentile means higher than roughly 15% of those readings. A positive score can still rank low if its usual tone is more positive.

Lowest 10%Middle 80%Highest 10%of this ticker’s tone history

Boxes span the middle 50% of active 15-minute readings; whiskers run from the 5th to 95th percentiles. Dots in the bottom or top decile are highlighted. Dashed boxes mark early histories with fewer than 30 readings. Charts appear with at least five active readings; shorter histories show their count.

Universe: QQQ and the nine stocks named in the published study. The latest source timestamp is shown above; these are tone readings, not price predictions.

Do the outcomes actually separate?

Each box summarizes stock-day returns above QQQ, grouped by sentiment relative to that stock’s own prior 60 days. Q1 is the lowest sentiment fifth; Q5 is the highest.

Same dayNext dayBox: middle 50% · whiskers: 10th–90th percentiles

Sentiment quintiles: same-day and next-day excess-return distributions, 25th–75th percentile boxes and 10th–90th percentile whiskers−5000+500Q1Same day, sentiment Q1: n=603; 10th -296.5, 25th -149.9, median -30.1, 75th 64.1, 90th 186.3 basis points; mean -47.4 basis points above QQQNext day, sentiment Q1: n=602; 10th -239.3, 25th -109.1, median 7.2, 75th 121.3, 90th 274.2 basis points; mean 15.9 basis points above QQQQ2Same day, sentiment Q2: n=603; 10th -217.8, 25th -101.0, median 1.4, 75th 110.6, 90th 233.9 basis points; mean 4.6 basis points above QQQNext day, sentiment Q2: n=601; 10th -215.0, 25th -115.3, median -14.3, 75th 104.7, 90th 274.6 basis points; mean 5.8 basis points above QQQQ3Same day, sentiment Q3: n=603; 10th -225.3, 25th -109.7, median -14.8, 75th 102.5, 90th 252.6 basis points; mean -2.3 basis points above QQQNext day, sentiment Q3: n=601; 10th -229.9, 25th -111.0, median 2.6, 75th 132.9, 90th 297.3 basis points; mean 19.3 basis points above QQQQ4Same day, sentiment Q4: n=603; 10th -197.4, 25th -93.4, median 0.3, 75th 120.6, 90th 297.1 basis points; mean 22.9 basis points above QQQNext day, sentiment Q4: n=601; 10th -232.1, 25th -99.6, median -9.5, 75th 100.1, 90th 239.8 basis points; mean -2.7 basis points above QQQQ5Same day, sentiment Q5: n=603; 10th -189.4, 25th -77.8, median 27.0, 75th 167.9, 90th 427.5 basis points; mean 85.7 basis points above QQQNext day, sentiment Q5: n=601; 10th -202.3, 25th -100.9, median -13.1, 75th 119.9, 90th 290.9 basis points; mean 25.0 basis points above QQQReturns above QQQ · basis points · Q1 negative → Q5 positive

The same-day average rises from −47.4 to +85.7 basis points. The next-day averages have no consistent ordering. The boxes make the overlapping outcomes visible.

Counts, means and quartiles
Window / fifthnMean bp10th25thMedian75th90th
Same day Q1603-47.4-296.5-149.9-30.164.1186.3
Same day Q26034.6-217.8-101.01.4110.6233.9
Same day Q3603-2.3-225.3-109.7-14.8102.5252.6
Same day Q460322.9-197.4-93.40.3120.6297.1
Same day Q560385.7-189.4-77.827.0167.9427.5
Next day Q160215.9-239.3-109.17.2121.3274.2
Next day Q26015.8-215.0-115.3-14.3104.7274.6
Next day Q360119.3-229.9-111.02.6132.9297.3
Next day Q4601-2.7-232.1-99.6-9.5100.1239.8
Next day Q560125.0-202.3-100.9-13.1119.9290.9

Which readings move with price, or lead it?

Six readings, three horizons. Cells show average daily rank correlation and its t-statistic. Same-day association overlaps the reading; it cannot establish a usable advance signal.

Six news readings tested at three horizons: rank correlation and t-statisticSame dayNext dayNext 5 daysSentiment levelSentiment level ret_same: correlation 0.126, t 6.15, 346 periods+0.126 · t +6.2Sentiment level ret_next1: correlation -0.009, t -0.47, 345 periods−0.009 · t −0.5Sentiment level ret_next5: correlation -0.012, t -0.43, 341 periods−0.012 · t −0.4vs own 60 daysvs own 60 days ret_same: correlation 0.135, t 6.91, 335 periods+0.135 · t +6.9vs own 60 days ret_next1: correlation -0.009, t -0.42, 334 periods−0.009 · t −0.4vs own 60 days ret_next5: correlation -0.011, t -0.42, 330 periods−0.011 · t −0.41-day change1-day change ret_same: correlation 0.123, t 6.20, 344 periods+0.123 · t +6.21-day change ret_next1: correlation 0.001, t 0.06, 343 periods+0.001 · t +0.11-day change ret_next5: correlation -0.008, t -0.51, 339 periods−0.008 · t −0.5News attentionNews attention ret_same: correlation 0.031, t 1.61, 340 periods+0.031 · t +1.6News attention ret_next1: correlation -0.022, t -1.06, 339 periods−0.022 · t −1.1News attention ret_next5: correlation -0.019, t -0.84, 335 periods−0.019 · t −0.8DisagreementDisagreement ret_same: correlation -0.107, t -5.04, 345 periods−0.107 · t −5.0Disagreement ret_next1: correlation 0.010, t 0.45, 344 periods+0.010 · t +0.5Disagreement ret_next5: correlation -0.038, t -1.65, 340 periods−0.038 · t −1.6TrendTrend ret_same: correlation 0.125, t 6.06, 346 periods+0.125 · t +6.1Trend ret_next1: correlation 0.004, t 0.21, 345 periods+0.004 · t +0.2Trend ret_next5: correlation -0.021, t -0.79, 341 periods−0.021 · t −0.8

The study uses |t| ≥ 3 with agreeing sample halves as its threshold for a clear result. Multiple testing and repeated observations matter; these cells are not independent discoveries.

ReadingHorizonCorrelationtDays
Sentiment levelNext day−0.009−0.47345
Sentiment levelNext 5 days−0.012−0.43341
Sentiment levelSame day+0.126+6.15346
vs own 60 daysNext day−0.009−0.42334
vs own 60 daysNext 5 days−0.011−0.42330
vs own 60 daysSame day+0.135+6.91335
1-day changeNext day+0.001+0.06343
1-day changeNext 5 days−0.008−0.51339
1-day changeSame day+0.123+6.20344
News attentionNext day−0.022−1.06339
News attentionNext 5 days−0.019−0.84335
News attentionSame day+0.031+1.61340
DisagreementNext day+0.010+0.45344
DisagreementNext 5 days−0.038−1.65340
DisagreementSame day−0.107−5.04345
TrendNext day+0.004+0.21345
TrendNext 5 days−0.021−0.79341
TrendSame day+0.125+6.06346

A different reading: one-day change in tone

Same dayNext dayBox: middle 50% · whiskers: 10th–90th percentiles

Sentiment quintiles: same-day and next-day excess-return distributions, 25th–75th percentile boxes and 10th–90th percentile whiskers−4000+400Q1Same day, sentiment Q1: n=620; 10th -285.9, 25th -135.9, median -26.3, 75th 86.9, 90th 228.5 basis points; mean -34.4 basis points above QQQNext day, sentiment Q1: n=618; 10th -216.3, 25th -109.7, median 8.2, 75th 141.0, 90th 329.6 basis points; mean 36.7 basis points above QQQQ2Same day, sentiment Q2: n=619; 10th -217.2, 25th -104.3, median -7.5, 75th 88.1, 90th 233.2 basis points; mean -0.1 basis points above QQQNext day, sentiment Q2: n=617; 10th -233.2, 25th -109.2, median -16.1, 75th 101.1, 90th 241.5 basis points; mean -6.0 basis points above QQQQ3Same day, sentiment Q3: n=619; 10th -227.7, 25th -111.9, median -16.5, 75th 98.9, 90th 249.5 basis points; mean 1.4 basis points above QQQNext day, sentiment Q3: n=617; 10th -209.4, 25th -109.3, median -5.7, 75th 103.3, 90th 238.4 basis points; mean 7.4 basis points above QQQQ4Same day, sentiment Q4: n=619; 10th -199.9, 25th -100.4, median 3.0, 75th 125.3, 90th 314.1 basis points; mean 26.6 basis points above QQQNext day, sentiment Q4: n=617; 10th -207.3, 25th -92.0, median -11.9, 75th 101.1, 90th 282.8 basis points; mean 10.4 basis points above QQQQ5Same day, sentiment Q5: n=619; 10th -195.7, 25th -72.1, median 32.7, 75th 167.0, 90th 385.5 basis points; mean 73.2 basis points above QQQNext day, sentiment Q5: n=618; 10th -250.2, 25th -114.6, median 6.5, 75th 128.8, 90th 291.5 basis points; mean 18.0 basis points above QQQReturns above QQQ · basis points · Q1 negative → Q5 positive

Here Q1 means the lowest one-day sentiment change, Q5 the highest. This is a separate grouping of the same stock-day data. A different transformation does not create a reliable next-day forecast.

What this view establishes

  • The tone boxes use the same RLTrade view: 15-minute readings with channel activity, 25th–75th percentile boxes, 5th–95th percentile whiskers, median lines and the latest reading as a dot.
  • Blue is news; orange is tweets. The percentile is an approximate rank within that stock’s own 30-day channel history; it is not a probability of a price move.
  • Source timestamps remain visible. Available histories with at least five readings are shown. Dashed boxes mark fewer than 30 active readings; these early-history percentiles are descriptive and can change quickly.
  • The most negative sentiment fifth averaged −47 bp against QQQ that day; the most positive averaged +86 bp.
  • Boxes show the middle 50% of stock-day outcomes; whiskers are the 10th and 90th percentiles. Both panels use the same return scale.
  • Sentiment is standardized against each stock’s prior 60 days, then pooled observations are split into fifths separately for each return horizon. Returns overlap the reading in the same-day comparison.
  • Next-day results are not ordered by sentiment. These are descriptive research results, not a standalone trading signal.
RLTrade public sentiment-distribution source ↗︎Published study · data, charts & method ↗︎Polygon.io · news/price source attribution ↗︎
Read the original study or full series entry →