Real or Random
Two game-by-game streaks, one real and one rolled with the same talent. Pick the real one. If nobody can, the hot hand looks a lot like luck.
Builds on Answers for Coach
- Coach questions
- 3
- Asked on air
- 31×
- Top question
- Q-01
Coach questions it could answer
The playbook →- Why do teams beat or miss their expected record?
Deal one-run game sequences from lucky and unlucky teams without names. The one-run study found that record is mostly coin flips; if players can't tell the good teams from the lucky ones, that's the same finding from the gut side.
- Are team hot and cold streaks more than coin-flip noise?
Put team win-loss runs in the deck, including the 2026 Red Sox's 31–5. Answers for Coach found that run unlikely as pure luck; this asks whether people can see it, too.
- Do hitter hot and cold streaks predict anything at all?
Show a hitter's real game-by-game log next to copies rolled from his season averages. If players can't pick the real one more often than chance, his hot and cold stretches look like what randomness produces anyway.
Data it needs
The data lake →1 of 2 on hand; the rest arrives in phase 3, Every play.
Could use later: Best stretches
The question
Everyone sees streaks. The hard part is knowing whether a streak means anything, because pure chance produces streaks too, and more of them than people expect. Can you tell a real hot streak from a random one?
How it would work
- Each round shows two sequences side by side: a real game log (hits and outs, or wins and losses) and one simulated from the same player’s or team’s overall rate. You pick the real one.
- Names and dates are hidden, so it’s the pattern alone.
- Guessing scores exactly one in two. Beating that consistently means real streaks have a shape randomness doesn’t.
First version
- 40 pairs: 20 hitters over 30-game stretches, 20 teams over 36-game stretches, both chosen without peeking at how streaky they look.
- A results page with accuracy, and which real sequences fooled people most.
The data
MLB Stats API results. Team game results for 1901–2026 are already cached by Answers for Coach; hitter game logs come from the same API.
What it needs
- Fair simulations. Copies need the same overall rate and, for hitters, the same plate appearances per game, so the only difference is order.
- Honest selection. Picking famous streaks would stack the deck, so sequences should be drawn at random from qualified seasons.
Status
An idea on the launchpad. Nothing is built yet.