Food for Thought: The “Hot Hand” in Cricket – Fallacy or Fact?

Most sports fans have heard of the hot hand fallacy. It’s the belief that if a player has hit a few shots in a row (say, in basketball), they’re more likely to keep scoring. Statisticians tell us that’s an illusion: each shot is independent, and streaks are often just chance.

But at MatchMind, our models suggest the story might be more nuanced.

Over the last three major T20 leagues – CPL, IPL, and BBL – our Head-to-Head (H2H) algorithm has shown a remarkable pattern. In the middle phase of each season (matches 13–24 in CPL/BBL, 20–34 in IPL), our ensemble of models consistently goes on a “hot streak,” calling outcomes correctly in about 78–80% of games. We joke internally that “the algo is hot.”

What’s interesting is that it’s not every individual model, but subsets within the ensemble that “heat up.” Think of the ensemble like a team: sometimes a few players find form, and their contribution lifts the overall performance.

Unlike the classical hot hand fallacy—which says streaks are random—our evidence suggests these model streaks aren’t luck. They arise because the models are learning as the season unfolds:

1. Early season: A burn-in period, where models gather context about teams and player form.

2. Middle season: Stability kicks in—squads are settled, data is rich, and the models converge on the right variables.

3. End of season: Chaos returns—injuries, load management, and playoff dynamics add noise.

So, streaks in this context are measurable, explainable, and repeatable—not random.

Which raises an intriguing question: if our algorithms can genuinely get hot, why not players too? When a batter or bowler goes on a run of form, perhaps it isn’t just chance. Maybe, consciously or subconsciously, they’re attuning to conditions, opponents, and season dynamics—gaining real skill edge in that window.

Of course, the big assumptions here are that players have the bandwidth to store all current and past information, and the ability to effectively “watch” and process every game. Models can do this automatically; for players, it would depend on how much context they can absorb and apply in real time.

We’re not in the basketball game, but imagine if we redefined the “hot hand” in cricket. What if we analysed when players’ streaks occur—early, middle, or late in a season? My hypothesis: the middle phase (when uncertainty is lowest) is where both models and players are truly in sync with the game’s rhythms.

So maybe the hot hand isn’t always a fallacy. Sometimes, it’s a signal—of learning, adaptation, and genuine form.