Research
Corrections log
When a published number, classification or headline turns out to be wrong, we change it and say so. Every such change is listed here with its date, alongside the dated note that stays on the study page itself.
This log covers corrections to work that was already public. Problems caught before a study publishes are more common than the handful below — a coefficient about to be borrowed from another sport instead of measured on its own record, a sample that drifts across an era, a band too thin to quote. Those are fixed in the method so the study goes live correct, and they are not individually logged here. So this page is not the complete list of issues we have found in our own work; it is the complete list of ones that ever reached a reader.
Numbers on every study also refresh as the weather archive and closing-odds backfills top up. Routine refreshes are not corrections and are not listed here.
30 October 2026
Highest-Elevation College Football Stadiums
What was wrong: Air Force’s Falcon Stadium was listed at 6,726 feet, from a coarser stadium-database value, and Wyoming’s gap over it was given as 494 feet.
What we corrected: Changed to 6,621 feet, the Air Force Academy’s published figure, which a terrain lookup at the stadium (2,020 m, about 6,627 feet) supports. The gap to Wyoming is now 599 feet. The ranking order is unchanged.
Read the Highest-Elevation College Football Stadiums study29 October 2026
Highest-Elevation College Football Stadiums
What was wrong: Wyoming’s War Memorial Stadium was listed at 7,205 feet, from a coarser stadium-database value, and its gap over Air Force’s Falcon Stadium was given as 479 feet.
What we corrected: Changed to 7,220 feet, the University of Wyoming’s official figure, which a USGS 1-metre reading at the stadium (about 7,219 feet) confirms. The gap over Falcon Stadium is now 494 feet. The ranking order is unchanged.
Read the Highest-Elevation College Football Stadiums study23 October 2026 (second update)
PGA Tour Wind Wave Study
What was wrong: A one-by-one check of every course location found 18 points that sat beside the course rather than on it — up to about 6 km away (for example TPC Louisiana, Grand Reserve, Sherwood, TPC Twin Cities and Sea Island, whose point was the nearby airport). Weather for those weeks came from the slightly wrong spot. The page reported a 95% range of 0.71–1.14 strokes per 10 km/h.
What we corrected: Every course now uses its mapped location and weather for the affected weeks was re-downloaded; every number on the page was re-run. The headline is unchanged at 0.93 strokes per 10 km/h (about 1.50 per 10 mph). The 95% range is now 0.72–1.14 (0.715–1.141), two wind bands moved by one event-round each, and every conclusion is unchanged.
Read the PGA Tour Wind Wave Study study23 October 2026
PGA Tour Wind Wave Study
What was wrong: Two Japan courses (Narashino Country Club, ZOZO Championship, 4 events; Yokohama Country Club, Baycurrent Classic, 1 event) were stored with the wrong longitude, so their weather came from a point in the Pacific Ocean. The page reported 0.94 strokes per 10 km/h (about 1.51 per 10 mph), range 0.72–1.14.
What we corrected: Both courses now use their map locations and their weather was re-downloaded; every number on the page was re-run. The headline is 0.93 strokes per 10 km/h (about 1.50 per 10 mph), range 0.71–1.14. Sample sizes, the cut result and every conclusion are unchanged. The methodology review covered the earlier run.
Read the PGA Tour Wind Wave Study study6 October 2026
MLB Wind Blowing Out vs the Betting Line
What was wrong: Several ballpark directions (home plate to centre field) were rounded or wrong, so some games were sorted into the wrong wind group. The largest errors were Target Field, Oracle Park and loanDepot park. The page reported wind blowing in at 10–14 mph as finishing 0.44 runs below the closing total (t = -2.10, 364 games), labelled directional.
What we corrected: Every park now uses MLB's official field direction (Citi Field and Target Field checked against aerial and layout sources), and every game was re-sorted. That result is now 0.36 runs below the close (t = -1.69, 359 games), which does not clear our bar. No wind group now shows a gap against the closing total.
Read the MLB Wind Blowing Out vs the Betting Line study6 October 2026
MLB Ballpark Wind Sensitivity
What was wrong: Same ballpark-direction error. The page said 3 of 18 measurable parks showed a real out-vs-in gap and quoted Wrigley Field at t = 6.59 (105 out vs 174 in games) and +1.66 runs vs the close (t = 3.12).
What we corrected: With corrected directions, 4 of 18 parks clear the bar: PNC Park joins Wrigley Field, Fenway Park and Busch Stadium. Wrigley is unchanged in substance: about +4.0 runs out vs in (t = 6.62, 106 vs 176 games) and +1.62 runs vs the close (t = 3.02, still directional).
Read the MLB Ballpark Wind Sensitivity study6 October 2026
MLB Wind Direction and Scoring
What was wrong: Same ballpark-direction error. The page quoted wind blowing in at −0.25 home runs and −0.41 runs vs calm, and wind blowing out at +0.18 home runs and +0.11 runs (t = 0.91).
What we corrected: Now −0.24 home runs and −0.34 runs for wind blowing in, and +0.16 home runs and +0.08 runs (t = 0.71) for wind blowing out. The conclusions are unchanged.
Read the MLB Wind Direction and Scoring study25 September 2026
NFL Dome Scoring
What was wrong: The fixed-dome-only split (n=204, t=1.97) was described as sitting at the edge of significance using a flat 1.96 cutoff. Our statistical reviewer pointed out the exact cutoff depends on sample size; for 203 degrees of freedom it is also about 1.97.
What we corrected: The fixed-dome split is now labelled directional only. The headline all-indoor result and the retractable-only split clear the exact cutoff and are unchanged. We now use exact critical t values for every comparison.
Read the NFL Dome Scoring study24 September 2026
NFL Climate Shock
What was wrong: The page called the Under result (15 of 21 graded games, 71.4%, 6.8 points below the closing total) a "strong structural signal" and "not a small-sample fluke". 21 games is below our 40-game minimum.
What we corrected: The result is now labelled directional only throughout, with the sample size stated. No numbers changed.
Read the NFL Climate Shock study24 September 2026
NFL Turnovers in Bad Weather
What was wrong: The summary said "turnovers rise clearly" when rain and cold arrive together. Only total fumbles clear our bar; overall turnovers (t=1.58) do not. The era-drift figures were quoted as 3.49 to 2.31 per game.
What we corrected: The summary now names total fumbles as the real effect and says overall turnovers do not clear the bar. Era drift is now quoted from the current data as about 3.6 to 2.2 per game.
Read the NFL Turnovers in Bad Weather study24 September 2026
MLB Temperature & Scoring
What was wrong: The summary said cold games "still go under the line" and that the market "does not fully price in the cold". No cold band clears |t| >= 1.96 against the closing total (t between -1.74 and -1.87).
What we corrected: The closing-line result is now labelled directional only. The runs and home-run findings, which are real, are unchanged.
Read the MLB Temperature & Scoring study24 September 2026
MLB Wind Direction & Scoring
What was wrong: The page said closing totals were not attached to this sample, which is no longer true.
What we corrected: The page now points to the follow-up study that tests wind against the closing total. No numbers changed.
Read the MLB Wind Direction & Scoring study24 September 2026
College Football Temperature & Scoring
What was wrong: A side note quoted freezing-game figures split by 2016–2020 and 2021–2025 that we could not reproduce from the current data.
What we corrected: The unverifiable figures were removed; the note now says only that neither half has enough freezing games to settle the trend alone. The headline result is unchanged.
Read the College Football Temperature & Scoring study24 September 2026
AFL Temperature & Scoring
What was wrong: A robustness note quoted two exact unadjusted band values (−9.8 and +11.9) that are not in the published data.
What we corrected: The note now states the check result (each band moved by under a point) without the unpublished figures. The headline result is unchanged.
Read the AFL Temperature & Scoring study23 September 2026
College Football Rain & Totals
What was wrong: The page described heavy rain as finishing under the closing total without explaining why the market might leave the gap open. Readers could infer the books simply "miss" rain, which overstates the claim.
What we corrected: Copy was sharpened around the forecast-lag hypothesis: heavy-rain totals appear to move too slowly once a reliable pre-kickoff forecast is in place, while light rain is priced correctly. The page now explicitly says this mechanism is the leading explanation but has not yet been independently validated with open-to-close line moves. No numbers changed.
Read the College Football Rain & Totals study23 September 2026
College Football Wind, Scoring & Kicking
What was wrong: The study only tested crosswind and reported game totals as showing "no material deviation". The wind-speed scoring question was answered from an older 5,167-game weather record that predates our current 40-game minimum, per-band t statistic and season-adjustment standards.
What we corrected: The study was rebuilt on the full 7,072-game closing-total sample. Every wind band is now season-adjusted against a calm (<10 mph) control with a reported t statistic and a dry-games-only control column: 10-15 mph costs 1.3 points (t=-2.28, n=1,170) and 15-20 mph costs 4.0 points (t=-3.34, n=254). The 20 mph+ band (n=38) is below the minimum and is marked directional only. No band beats the closing total by our own standard, and the page now says so instead of describing totals as unaffected. The college lean model wind coefficients were updated to the re-measured figures.
Read the College Football Wind, Scoring & Kicking study13 September 2026
NFL Wind & Totals
What was wrong: The 20+ mph "gale" band (n=40) was presented alongside the moderate and strong bands as though it carried the same weight, and the filtered strong-wind result was described without noting how close it sits to our significance threshold.
What we corrected: The gale band is now labelled directional evidence only — n=40 merely clears our pre-existing minimum sample and is not a reliable standalone finding — and it is excluded from headlines, summary statistics and any pricing or lean use. The filtered strong-wind wording now states plainly that the scoring loss remains after excluding wet and colder games but is reduced and sits near the threshold for statistical significance.
Read the NFL Wind & Totals study11 September 2026
College Football Rain & Totals
What was wrong: The season table showed an empty 2024 row and only 6 games for 2025, which read as if almost no games were played in those seasons.
What we corrected: Those rows now say "No closing totals captured yet — odds backfill pending", and the limitations section explains that the gap is incomplete historical odds coverage, not missing games. The closing-line backfill has since filled most of it and the tables update automatically as it completes.
Read the College Football Rain & Totals study3 September 2026
College Football Wind & Kicking
What was wrong: The headline claimed strong crosswind cost about 5.3 points of field-goal accuracy. That figure came from distance bands holding only 8–26 attempts each — well below our own 40-attempt minimum.
What we corrected: The headline was recalculated using only bands that clear the minimum sample. At sample sizes we trust, crosswind does not measurably move college field-goal accuracy, and the study now says so. The original number is described on the study page in a dated correction note rather than quietly removed.
Read the College Football Wind & Kicking study2 September 2026
MLB Wind Direction & Scoring
What was wrong: Coors Field was stored with a home-plate-to-centre-field bearing of 35°, which disagreed with published orientation figures. Since wind is classified as blowing out, in or across relative to that bearing, a wrong bearing misclassifies every windy game at the park.
What we corrected: The bearing was corrected to 0° (north) before the analysis was run, matching the published figures in Andrew Clem’s stadium tables and The Shadium. All 31 park orientations were verified against published figures and aerial imagery as a pre-analysis step, and that check is now part of the method rather than something done after seeing results.
Read the MLB Wind Direction & Scoring study29 August 2026
NFL Climate Shock
What was wrong: The Raiders were classified by their current venue, so games played outdoors in Oakland through 2019 would have been counted as dome-team road games — inflating the pool of "climate shock" matchups.
What we corrected: Venue type is now applied per season: Oakland games through 2019 count as outdoor, Las Vegas games from 2020 count as enclosed. The qualifying-game count and every band was recomputed on the corrected classification.
Read the NFL Climate Shock study