## The Recency Problem
Recency bias is the tendency to weight recent events more heavily than older events, even when the older events are statistically more representative. In betting, this produces predictable errors.
## Classic Recency Errors
**The form team overreaction:**
A team wins 4 matches in a row. Their next-match odds shorten significantly. The market — and most bettors — extrapolate the winning run as evidence of superiority. Statistical analysis shows: winning runs of 4 in a sport like football are common even for average teams due to schedule variance and goalkeeper performance regression.
**The bad run underreaction:**
A strong team loses 3 consecutive matches. Bettors avoid them; the market drifts. But the underlying xG data shows they are performing well — they have been unlucky. The recency bias creates value for the analytically rigorous bettor.
**The star player game:**
A striker scores in 3 of 4 matches. His anytime scorer price falls. Recency extrapolation ignores his underlying xG rate (the best predictor) in favour of recent results.
## The Narrative Trap
Media coverage amplifies recency bias. When a team is "on fire," every outlet publishes this narrative. This amplified narrative affects public betting behaviour and moves markets independently of actual probability change. The narrative and the underlying probability can diverge significantly.
## The Correction Mechanism
Maintain a rolling model that weights historical data appropriately:
- Last 5 matches: 30% weight
- Matches 6–15: 40% weight
- Matches 16–25: 20% weight
- Pre-season/historical baseline: 10% weight
This prevents any short-term run from dominating the probability estimate.
Create a free account to track your progress and save bookmarks.
