Every playoff probability we publish is scored against what actually happened. This page reports the calibration metrics — Brier score, log loss, and expected calibration error — for the NHL, NBA, and MLB models currently serving odds on the site, recomputed with each nightly simulation run.
A playoff probability is only useful if it is calibrated — if the teams we give a 30% chance actually make the postseason about 30% of the time. Most sites publish odds without ever publishing how well those odds have held up. We score every run against real outcomes and publish the result, including when it is unflattering.
The metrics above are recomputed on every nightly simulation and stored alongside the run that produced them, so the figures on this page always correspond to the same model version currently serving odds across the site.
We publish the model's Brier score, log loss, and expected calibration error for every league on this page, recomputed with each nightly simulation. Rather than claiming an accuracy percentage, we report calibration: whether teams given an X% chance actually qualify about X% of the time.
Brier score runs from 0 (perfect) to 1 (perfectly wrong), with 0.25 being the score of an uninformative 50/50 forecast. Any score meaningfully below 0.25 carries real predictive signal, and well-calibrated sports probability models typically land well under that mark.
Yes. All figures on this page are free to cite with attribution to "In The Playoffs" linked to intheplayoffs.com. See our press and media kit page for citation guidance.