Sports & Gaming · Statistics · Descriptive Statistics
Basketball Pythagorean Win Calculator
Estimates a basketball team's expected win percentage from points scored and points allowed using the Pythagorean expectation formula.
Calculator
Formula
W% is the expected win percentage. PF is points scored (points for) per game. PA is points allowed (points against) per game. The exponent 13.91 is the empirically fitted value for the NBA, derived by Daryl Morey.
Source: Daryl Morey (Houston Rockets / MIT Sloan Sports Analytics), NBA Pythagorean exponent = 13.91; popularized in Basketball Reference team statistics.
How it works
The formula raises a team's points scored (PF) and points allowed (PA) to the power of 13.91, then divides PF^13.91 by the sum (PF^13.91 + PA^13.91). The resulting decimal is the team's Pythagorean win percentage — the fraction of games they are statistically expected to win given their scoring efficiency. Multiplying that percentage by the number of games played yields projected wins and losses.
The exponent 13.91 was empirically calibrated for the NBA by Daryl Morey (later of the Houston Rockets) and is now the standard value used on Basketball Reference. A higher exponent makes the formula more sensitive to large scoring margins, which suits basketball's higher-scoring environment compared to baseball (where Bill James originally used an exponent of 2) or football (where 2.37 is common).
Teams whose actual win total significantly exceeds their Pythagorean wins are often said to be 'lucky' — winning many close games — while teams below their expectation may have been unlucky or mismanaged their roster in clutch situations. Analysts use the gap between real and Pythagorean wins to forecast regression toward the mean in future games.
Worked example
Suppose a team scores 115.4 points per game and allows 111.2 points per game over an 82-game season.
Step 1 — Raise each to the power 13.91:
115.4^13.91 ≈ 7.803 × 10^28
111.2^13.91 ≈ 4.821 × 10^28
Step 2 — Calculate win percentage:
W% = 7.803 × 10^28 / (7.803 × 10^28 + 4.821 × 10^28) = 7.803 / 12.624 ≈ 0.6184 (61.8%)
Step 3 — Project over 82 games:
Expected Wins = 0.6184 × 82 ≈ 50.7 wins
Expected Losses = 82 − 50.7 ≈ 31.3 losses
If this team's actual record is 47-35, they are under-performing their Pythagorean projection by about 3.7 wins, suggesting they have been losing close games and may improve as the season progresses.
Limitations & notes
The Pythagorean formula uses season-averaged points per game, so it treats all games equally regardless of opponent strength, pace of play, or home/away splits. A team that systematically wins blowouts but loses close games will show a high Pythagorean expectation even if their actual record is modest. The exponent 13.91 is optimized across the full NBA historical sample and may be slightly off for any individual team or era. The calculator also does not account for injury-depleted rosters, back-to-back scheduling, or playoff-style adjustments. For playoff predictions, pace-adjusted offensive and defensive ratings (points per 100 possessions) provide a more granular estimate than raw per-game scoring averages.
Frequently asked questions
Why is the exponent 13.91 used for basketball instead of 2 like in baseball?
Bill James originally used an exponent of 2 for baseball, but basketball has much higher scores and narrower win-percentage variance, meaning point differentials need to be amplified more to distinguish team quality. Daryl Morey found that 13.91 minimizes prediction error across NBA seasons. Higher-scoring sports generally require larger exponents for the formula to be well-calibrated.
What does it mean if a team's actual wins are much higher than their Pythagorean wins?
It suggests the team has been winning a disproportionate number of close games — sometimes called being 'clutch' or sometimes just lucky. Research shows that the ability to win close games is largely non-repeatable year to year, so teams in this position tend to regress toward their Pythagorean expectation as the season continues.
Can I use per-game averages or should I use season totals?
You can use either, because the exponentiation cancels the per-game scaling. If both PF and PA are per-game averages or both are season totals, the resulting win percentage is identical. Per-game averages are more convenient and allow mid-season projections before 82 games are played.
How accurate is the Pythagorean expectation compared to actual wins?
Over a full 82-game NBA season, the Pythagorean formula typically predicts actual wins within 2–3 games for most teams. Teams that deviate by more than 5 wins from their projection in the first half of a season very commonly regress toward the prediction in the second half, making this a useful tool for betting and roster decisions.
Is there a different Pythagorean exponent for college basketball or the WNBA?
Yes. The optimal exponent depends on the league's scoring environment. College basketball, which often features lower scores and different pace, tends to use exponents in the range of 10–11. The WNBA uses lower per-game scores than the NBA, so a smaller exponent is more appropriate. The 13.91 value should be used specifically for NBA analysis; recalibrating the exponent for other leagues improves accuracy.
Does the Pythagorean formula account for strength of schedule?
No. It uses raw points scored and allowed regardless of opponent quality. A team that padded its point differential against weak opponents will have an inflated Pythagorean win percentage. For a strength-of-schedule-adjusted estimate, analysts use opponent-adjusted offensive and defensive efficiency ratings instead of raw per-game scoring.
Last updated: 2025-01-30 · Formula verified against primary sources.