Leverage & Clutch
Not every round matters equally. Leverage Index is how much win probability is on the table before a round starts — how big a swing the round is capable of producing, no matter who ends up winning it.
1.00 is an average TBL round. 2.00 means twice the normal stakes. 0.10 means the round is nearly meaningless. The most important situation possible in TBL is a tied match with one round to go, at 6.63 — more than six times an average round.
LI belongs to the situation, not the fighter. Both fighters in a round face exactly the same leverage.
See the highest-leverage rounds of the season on the Biggest Moments board.
| Situation | LI |
|---|---|
| Tied, 1 round left | 6.62 |
| Tied, 2 rounds left | 4.18 |
| Tied, 4 rounds left | 3.10 |
| Tied, 12 rounds left | 1.86 |
| Tied, opening bell | 1.32 |
| Up 3, 12 rounds left | 1.26 |
| Up 5, 8 rounds left | 0.51 |
| Up 10, 8 rounds left | 0.02 |
| Up 15, 4 rounds left | 0.00 |
Leverage climbs as rounds run out while the score stays close, and collapses toward zero once a lead is out of reach. Up 15 with 4 rounds left, the match is over in all but name — LI 0.00.
TBL matches run in three phases of equal length. Sort every round by phase and something counterintuitive falls out:
| Phase | Rounds | Avg LI | Median LI | Max LI |
|---|---|---|---|---|
| Launch | 438 | 1.21 | 1.32 | 1.57 |
| Middle | 438 | 1.01 | 1.20 | 2.16 |
| Money | 438 | 0.78 | 0.09 | 6.62 |
Money rounds have the lowest average importance of any phase — and by far the highest ceiling. The median Money round has an LI of just 0.09, meaning the match is usually already settled by the time it arrives. But every one of the most important rounds in TBL history is a Money round.
Money rounds aren't consistently big. They're either nothing or everything.
Zoom out and the same pattern holds leaguewide: 22% of all TBL rounds have an LI below 0.10 — roughly one round in 5 is competitively meaningless — while 6.8% carry more than double average importance.
Did a fighter's results come in the moments that mattered? Clutch takes what each round result would have been worth at average importance, and subtracts it from what the fighter actually produced.
Positive Clutch means a fighter's wins landed in bigger spots than their losses. Negative means they piled up round wins in garbage time and lost the ones that counted.
Take Money Powell: 19 rounds, 12 of them won — but an average leverage of 0.41, the lowest of any qualified fighter. His WPA is just +0.04. The same results at average leverage would have been worth +0.59. That's a Clutch of −0.55: he won plenty of rounds, but almost none of them mattered.
Now take Erika Sanchez. Her +1.54 WPA is well clear of the +1.09 the same results would have been worth in ordinary rounds — a Clutch of +0.45. Same league, opposite story.
Both are on the advanced leaderboard, sortable by Clutch and by average leverage.
- LI is descriptive, not evaluative. A high average leverage means a fighter was used in big spots. It says nothing about whether they were good in them.
- Clutch over one season is noisy. In every sport that computes it, clutch performance has proven hard to sustain year over year. Treat it as a description of what happened, not a prediction of what comes next.
- Neither adjusts for opponent quality.
- Disqualification rounds are excluded from both.
Technical Details — the formal spec
Leverage Index. Before a round, the state is a score differential d and rounds remaining r — where r INCLUDES the round about to be fought. Take every possible round outcome, weight each by how often it occurs, and average the absolute win-probability change:rawLI(d, r) = Σ over margins v of [ P(v) · |WP(d + v, r − 1) − WP(d, r)| ]LI(d, r) = rawLI(d, r) / 0.0753028919P(v) is the same per-round margin distribution the win-probability model uses (see How WPA Works) — it is not redefined here.
The normalizer is frozen. 0.0753028919 is the mean rawLI across all 1,314 competitive rounds of 2026. It is deliberately never recomputed per season: if it were, an LI of 1.20 in 2026 and 1.20 in 2027 would mean different things and the stat would stop being comparable across years. LI is exactly symmetric — LI(d, r) = LI(−d, r) — because both teams face identical stakes; the shipped table is asserted against that on every regeneration.
Clutch = WPA − cnWPA. Context-neutral WPA depends only on the round's point margin, derived as (WP(v, 23) − 0.5) × 0.7579047422 — a tied match at the opening bell as the reference state, rescaled to exactly average leverage. Symmetric by construction: winning by 2 is worth exactly the negative of losing by 2.
| Round margin | -4 | -3 | -2 | -1 | 0 | +1 | +2 | +3 | +4 |
|---|---|---|---|---|---|---|---|---|---|
| cnWPA | -0.221511 | -0.174726 | -0.121034 | -0.061960 | +0.000000 | +0.061960 | +0.121034 | +0.174726 | +0.221511 |
Edge cases. Scheduled rounds vary by match (24 for most 2026 matches; matches 7 and 48 were 21), and r is measured against that match's own count. Disqualification rounds are excluded from fighter-level LI and Clutch — they already produce zero WPA, so counting their leverage in the denominator would distort both stats; the round still carries an LI for display on the match page. This affects 8 rounds, so a fighter's LI round count can be lower than their WPA round count. Match 14's post-match round 25 is excluded entirely. Rounds scored 0-0 count normally: they carry real leverage and a small, real WPA. Leaguewide, Σ cnWPA and Σ Clutch are both 0.000.