What regression to the mean is
Regression to the mean — Extreme results tend to be followed by ordinary ones, because the extreme was partly noise and noise does not persist.
Any observed result is a mix of true ability and random variation. The variation is by definition not repeatable, so the next observation is closer to the ability alone.
It applies to the measurement, not to the entity. A team that genuinely improved does not regress; its new level is its level.
The problem arrives next. It is constantly invoked as a force pulling results back, which is the gambler's fallacy wearing a statistical term.
What to do with it. Ask whether an extreme result came from a change in the team or from noise, and only expect regression from the second.
The archive
What the published price did in US sports
11,262 games in US sports that are over and carried a price when we wrote them up, between November 24, 2025 and November 6, 2026. Each count sits beside what it is out of, and the share is the one divided by the other.
| What | How often | Out of | Share |
|---|---|---|---|
| The shorter price won | 7,058 | 10,936 | 64.5% |
| The favourite covered the handicap | 3,373 | 7,003 | 48.2% |
| The total went over | 5,518 | 11,236 | 49.1% |
| The home side won | 6,361 | 11,262 | 56.5% |
How they ended
Over 16,021 finished games in US sports whose result line says how it ended — the feed's own one-line verdict, not a reading of the play.
| What | How often | Out of | Share |
|---|---|---|---|
| Went past regulation | 1,212 | 16,021 | 7.6% |
| Settled on a shootout | 161 | 16,021 | 1.0% |
| Called before the end | 9 | 16,021 | 0.1% |
286 games landed exactly on the number — the handicap or the total — so they are in the counts they are out of and in neither result.
The prices counted here are DraftKings', as they stood when we last wrote each game up — one snapshot per game, near kick-off for almost all of them, and never a closing line. So this says what the published line did, not how it moved on the way there.
A season that happened. Nothing here predicts the next game, and none of it is advice.
Counted November 6, 2026.
Today's free pick, for your email
Team, bet, price and the hash it was sealed under, on this page the moment you send. Lose or win, it stays public.
or by email
Public record of the main pick, losses included: 115 won, 44 lost. See every pick
One free pick a day by email until you unsubscribe, in one click. 21+.