Theorem 3: The Late-Game Timeout¶

Claim¶

When trailing by up to three points (or tied), is calling a timeout better than playing on with 20--50 seconds remaining?

How We Measure It¶

We filter the historical play-by-play log for situations where:

• Either the home or away team has possession
• That team is trailing by up to 3 points, or tied
• Between 20 and 50 seconds remain

We group possessions by:

• Timeout: The team stops play with a timeout call.
• Play On: The team's first action is a shot or turnover.

Other first actions are excluded. We calculate the possessing team's historical win percentage and save observation counts and pointwise 95% uncertainty intervals for each group. Each interval uses the wider limits from a game-cluster bootstrap and Wilson/Newcombe finite-sample bounds. Games are resampled as blocks so repeated clock values and overtime periods from one game remain together. This is descriptive: timeout availability, team quality, and why a coach stopped play are not controlled for.

Results¶

Conclusion¶

Calling timeout had the higher observed win rate at 11 of 16 comparable clock points. The 95% interval excluded zero in favor of timeout at 0 points. This association does not show that the timeout itself caused the difference.