The Gambler's Fallacy: Why Past Results Do Not Predict Future Outcomes

A football team has lost three matches in a row. Every team has a natural winning percentage. This team usually wins about 45% of matches. They've been below that rate lately, so they must be "due" for a win. You place a bet backing them.

This reasoning feels intuitive. It's also wrong.

This is the gambler's fallacy: the belief that past results influence the probability of future outcomes in sequences of independent events. Past results don't influence future results. Each match is independent. The fact that a team lost three in a row tells you something about their current state, but it doesn't change the probability distribution for their next match.

Independent Versus Dependent Events

The gambler's fallacy comes from confusing independent events with dependent events.

In a dependent event system, past results do influence future results. If you have a bag containing 10 red balls and 10 blue balls, and you draw a red ball, the contents of the bag have changed. There are now 9 red and 10 blue. The probability of the next draw being red has decreased from 50% to 47%. This is a dependent system.

In an independent event system, past results don't influence future results. A coin flip is independent. The coin has no memory. It doesn't know it landed on heads five times in a row. The probability of the next flip being tails is still 50%, regardless of history.

Sports matches are closer to independent events than dependent events. Yes, team composition, injuries, and morale can change, which adds some dependence. But the vast majority of the randomness in match outcomes comes from in-game variance, not from predictable deterministic factors.

When you assume that past results predict future results, you're treating an independent system as if it's dependent.

Why Streaks Feel Significant

Humans are pattern-recognition machines. We evolved to spot patterns because spotting patterns (like predator patterns) improved survival. But this pattern-recognition ability is overly sensitive. We see patterns in random data.

A team winning five in a row is within normal variance for a typical team. Variance means that even a team that has a 50% win rate will sometimes win five in a row by pure chance. This doesn't mean the team has become a 100% winning team. It means they've experienced a streak, which happens sometimes due to randomness.

Your brain, however, sees the streak and concludes something has changed. You think, "This team is in form now." Or conversely, "This losing streak means they're in decline."

Both conclusions might be partially true. Form does exist. But the team's true winning percentage hasn't changed as dramatically as the streak suggests. They're likely slightly better or slightly worse than they were before the streak, but not as much as the streak indicates.

Regression to the Mean

Regression to the mean is the principle that extreme outcomes (good or bad) tend to move back toward the average over time.

A team with a true 45% win rate experiences a fortunate run of luck and wins five in a row. Their observed win rate is temporarily 100%. But their true win rate hasn't changed. Over the next 100 matches, you'd expect them to win about 45 again, not 100. The streak regresses back toward the mean.

This doesn't happen because the team is "due" for losses (gambler's fallacy). It happens because the streak was partly luck, and luck is random. Random luck can't sustain indefinitely.

The confusion is that regression to the mean looks like the gambler's fallacy is correct. A team wins five in a row and then loses the next three. It seems like the loss was "due." But what actually happened is that the streak was partly luck, and when luck randomises again, it looks like the streak reversed.

How This Affects Betting

The gambler's fallacy creates specific betting errors:

Betting that streaks continue. You think a team in a winning streak is more likely to keep winning because they're "in form." You overestimate the probability of continuation.

Betting that streaks reverse. You think a team in a losing streak is more likely to win next because they're "due." You overestimate the probability of reversal.

Assuming short-term variance signals long-term change. A team's performance in five matches is noisy. It's influenced heavily by luck. You shouldn't dramatically change your assessment of the team's true strength based on five matches.

Missing the real drivers of change. If a team's true win rate has changed (injuries, tactical changes, managerial change), that's real. But this isn't the gambler's fallacy. That's genuine change in underlying conditions.

The solution is to base your betting on:

1. Underlying team strength (model-based)
2. Changes in conditions that actually affect outcomes (injuries, management, form of specific players)
3. Not on recent results alone

Recent results are a data point, but they're noisy. A small sample of results might look like a trend when it's just variance.

Distinguishing Real Patterns from False Ones

The challenge is distinguishing genuine patterns (team has improved because they've improved underlying factors) from random streaks (team just happened to get lucky a few times).

Here's a practical approach:

Look at underlying metrics, not just results. A team might lose three matches but improve their underlying performance metrics (expected goals created, defensive efficiency, etc.). If underlying metrics are improving, the losses might be bad luck rather than a downward trend.

Use a large sample size. Three matches is too small to draw conclusions about true strength. Twenty matches is better. Fifty is better still. The larger the sample, the more signal you have relative to noise.

Account for opponent quality. A team might have won five in a row, but if the opponents were all weak, the streak is less impressive than if the opponents were all strong. Strength of schedule matters.

Track expected results versus actual results. A team might have lost three matches but generated three high-quality chances in each match and should have won most of them. The losses are bad luck, not a sign of decline.

Ask: what changed? Has the team's roster changed? Has the manager changed? Have they discovered a new tactical approach? If nothing changed, then the streak is likely variance, not a signal of new reality.

The Professional Approach

Sharp bettors use models that adjust for recent results but don't overreact to them. A model might say:

• Team has a base win rate of 45% (from season-long history)
• They've won 4 of their last 5 matches
• This is better than the 45%, suggesting either luck or improvement
• I'll adjust slightly upward, maybe to 48%
• But I won't adjust all the way to 80% (their recent observed rate) because that would be overreacting to variance

This adjustment is based on Bayesian updating: new information (recent results) updates your belief, but you don't overweight the new information relative to the prior belief (season-long history).

Casual bettors often do the opposite. They see the recent streak and conclude the team has become a 80% winning team. This is overweighting recent results.

In Summary

• The gambler's fallacy is the belief that past results predict future results, but football matches are largely independent events where recent streaks reflect luck, not permanent changes
• A team winning five in a row usually hasn't permanently improved from 45% to 100% win rate; the streak is partly luck that regresses toward their true level
• Regression to the mean is what happens when luck randomises again; this looks like the streak is reversing, but it's actually the team returning to their underlying strength
• Sharp bettors use models that adjust slightly for recent results (maybe from 45% to 48% after a 4-in-5 streak) but don't overreact to variance (not jumping to 80%)
• The solution is to base assessments on underlying metrics (expected goals, defensive efficiency) rather than match results alone, and only change assessments when conditions genuinely change
• Real patterns occur when something has actually changed (injuries, tactics, management), not when a team merely experiences variance in results
• Large sample sizes reduce noise: three matches is too small to draw conclusions about true strength; fifty matches provides much more reliable signal

---

Frequently Asked Questions

Q: Doesn't a team's winning streak show they're in good form? A: Form is real, but it's much smaller than streaks suggest. A team might be slightly in form (maybe their win rate increased from 45% to 47%), but the streak makes it look like their win rate is 100%. That's overestimating the form effect.

Q: If past results don't predict future results, why does bookmaker pricing change after matches? A: Bookmakers are updating on new information. A team losing a match reveals something about the team (maybe they're worse than expected, or just unlucky). But bookmakers don't overreact. They update thoughtfully, not dramatically.

Q: Is there any situation where the gambler's fallacy is correct? A: Only when something has actually changed. If a team's manager is sacked or a key player is injured, that's a genuine change, not the fallacy. The fallacy is thinking change has happened when it hasn't, just because of a streak.

Q: How do I know if a betting trend is real or just variance? A: Look at sample size (larger is better), underlying metrics (do they support the trend?), and underlying changes (did something change that would cause the trend?). If the sample is small, metrics are mixed, and nothing changed, it's probably just variance.

Q: Should I ever bet based on streaks? A: Only if the streak is indicating something real about the underlying conditions. A team winning five in a row with strong underlying metrics has probably genuinely improved. A team winning five in a row with poor underlying metrics might just be lucky. Judge the underlying reality, not the streak.

Q: Is regression to the mean the same as the gambler's fallacy? A: No. Regression to the mean is what actually happens. The gambler's fallacy is the incorrect belief about why it happens. The gambler thinks a streak reverses because the opposite is "due." In reality, the streak reverses (or stabilises) because luck randomises again and the team returns toward their true level.