How we calculate UFC ratings

Standard Elo: every UFC fighter starts at 1500. After each bout the winner gains points, the loser loses an equal amount, and the swing scales with the pre-fight gap — favourites gain less for expected wins and lose more for upsets. K-factor is 32, multiplied by 1.25 for title fights, 1.15 for finishes, and 1.5 for fighters with fewer than 5 UFC appearances (so ratings find their level faster). The cumulative Global Elo over a career is what the Honest Elo history chart on each fighter page plots.

Expected score:
  E_A = 1 / (1 + 10^((R_B − R_A) / 400))

Rating update:
  R'_A = R_A + K_eff × (S_A − E_A)

where S_A is 1 for a win, 0 for a loss, 0.5 for a draw, and
K_eff = 32 × title_mult × finish_mult × newcomer_mult.

A retired champion shouldn't top the rankings forever. After 12 months without a UFC bout, a fighter's Global Elo drifts toward the division mean at 15% of the above-mean excess per year, capped at 25% total. The decay is mild by design — it doesn't punish brief layoffs, only prolonged ones. For multi-division fighters, the division mean is taken from the fighter's most recent UFC bout's division. Catchweight bouts fall back to the most recent non-catchweight division on file.

Fighters with fewer than 4 UFC fights have their Global Elo pulled toward the division mean, weighted by n_fights / 4. A 1-fight newcomer keeps only 25% of their raw swing; at 4 fights the rating stands on its own. Below 6 fights the fighter is also flagged with a Provisional chip on their card.

We re-simulate each fighter's last 8 fights 400 times by resampling with replacement and recomputing the Global Elo update chain. The standard deviation of the final ratings is our σ. Fighters with fewer than 8 fights get σ inflated by √(8/n) to reflect the thin sample.

This is an uncertainty proxy — not a full Bayesian posterior. It measures how sensitive a fighter's rating is to which subset of recent fights we credit, not how MMA history would have played out under counterfactual matchmaking. Order matters in real Elo update chains, and bootstrap resampling can produce sequences that wouldn't occur naturally; we accept that tradeoff because the proxy ranks fighters in roughly the right order of reliability.

The headline Elo rating on every leaderboard is Global Elo minus 1.3 × σ. A provisional star with Global Elo 1700 and σ 120 (Elo rating ≈ 1544) ranks below a proven contender with Global Elo 1680 and σ 40 (Elo rating ≈ 1628) — and that's the point. Our published rank is always by Elo rating, never by Global Elo.

Elo rating = Global Elo − 1.3 × σ