Recency bias in sports betting: why last week isn't everything

A running back rushes for 180 yards on Sunday. A guard drops 40 on Tuesday. A team you trusted gets blown out by 30. By the next game, that one result feels like the whole story.

That feeling is recency bias. It's the habit of giving the latest information more weight than it deserves when you guess what happens next. In betting, it shows up as chasing whoever just looked great and fading whoever just looked bad.

Where it comes from

Psychologists Amos Tversky and Daniel Kahneman described a related shortcut in 1973 and called it the availability heuristic. People judge how likely something is by how easily examples come to mind. Recent games are the easiest to recall. You watched them, saw the highlights and read the takes. Game 14 from last season is gone. Last Sunday is right there.

Memory researchers also describe a recency effect: people tend to recall the last items in a list best. Put those together and you get a simple problem. The games you remember most clearly are not a fair sample of the games that matter.

One game is a small slice

Here's how little a hot stretch can change a real average.

Say a player scored 22 points a game over his first 57 games. Then he goes for 31 a game over his next 3. That's 93 points in the hot stretch. His average over all 60 games is now 22.45.

Three loud games moved his 60-game average by less than half a point. If your gut now prices him like a 30-point scorer, your gut is working from 3 games and ignoring 57.

That doesn't mean the last 3 games are worthless. Maybe his minutes went up. Maybe a teammate got hurt and his role grew. Those are real reasons. "He looked unstoppable" is not one.

Luck makes streaks look like signals

Short runs are noisy, even when nothing has changed.

Take a team that is truly a coin flip against the spread, 50% to cover every week. The chance it covers at least 4 of 5 games by luck alone is 6 in 32, or 18.75%. In a 32-team league, that works out to about 6 teams doing it over any 5-week stretch, with no real change in quality.

Your own record works the same way. If you're a true 50% bettor, the typical swing in your win rate over 20 bets is about 11 percentage points. A 12-8 run is 60%, which sits inside that normal range. Over 500 bets, the typical swing shrinks to about 2 points. Small samples shout. Big samples whisper the truth.

This pull back toward normal is called regression to the mean. Extreme results tend to be followed by less extreme ones, because part of what made them extreme was luck.

The market sees last week too

Sportsbooks watched the same game you did. So did everyone else betting it. By the time you look at the next line, the blowout is usually already in the number. That's what line movement reflects: new information and new money.

So the useful question isn't "did they look good?" It's "is this price wrong?"

Research on whether markets overreact to streaks is mixed. Economist Colin Camerer studied NBA point spreads in the 1980s and found that teams on winning streaks were slightly overvalued. He concluded the error was too small to be profitably exploited. A later study of NBA seasons from 2001 to 2013 found the opposite pattern. Smart researchers looking at different years reached different answers. That's a good reason not to build a bet around a streak alone.

Where recency bias hits hardest

• Player props: One big game gets replayed all week. Check role, minutes and matchup before you pay for the last box score.
• Early-season games: After two or three weeks, every team has a tiny sample. Weekly results swing opinions far more than they should.
• Your own results: A hot week makes you want to bet bigger. A cold week makes you want to quit a process that might be fine. Neither tells you much on its own.
• Teams coming off blowouts: A 30-point win or loss is one game. Ask what caused it and whether that cause will still be there next week.

How to keep it in check

• Zoom out first. Before you look at last week, look at the full season or the last 10 games. Then decide if the latest game adds real information.
• Name the change. If you think a team or player is different now, write down why. An injury, a lineup change or a new role counts. A feeling doesn't.
• Compare the price, not the vibe. At standard -110 pricing, you need to win 52.4% of the time to break even. If a line has already moved because of last week, ask whether you still clear that bar at the new number.
• Grade bets over hundreds, not dozens. Track your record and your closing line value over a big sample before you change your approach.
• Keep stakes flat. Bet sizes shouldn't rise after a heater or shrink after a skid. Your unit stays your unit.

Last week matters. It just isn't the only week. The bettors who last are the ones who put each new game in its proper place, as one more data point in a long record.

The DARIMARKET board ranks every play by confidence and shows the price it needs to beat.

If it stops being fun, call or text 1-800-MY-RESET (1-800-697-3738). 21+.