Metrics
Valorant Clutch Stats: Real Win Odds and CLOE
What counts as a 1vX, what the odds actually are in each state, and why clutches-over-expected is the only fair way to score them.
A clutch is a 1vX: the moment your team drops to exactly one player alive while at least one enemy remains, the survivor is in a clutch. Riptide detects that moment from the kill feed, records the manpower state (1v1 through 1v5) plus the side and plant state, and scores the attempt by whether the round was won. Being the lone survivor is enough to count - you do not have to win it for it to be an attempt.
Pooled across every state, established competitive players win ~17% of their clutch attempts. That single number hides the real story, because a 1v1 and a 1v4 are completely different bets.
Win odds by state
| State | Expected win odds |
|---|---|
| 1v1 | ~61% |
| 1v2 | ~20% |
| 1v3 | ~5% |
| 1v4 | ~1.5% |
| 1v5 | under 1% |
Base expected win rates from Riptide's clutch model, fit to measured lone-survivor outcomes in the live corpus.
These come from the engine's xClutch model: a logistic model over manpower, side, and plant state, whose base rates are the measured win rates of lone survivors in the corpus. Two things stand out. A 1v1 is a favorable fight - the survivor wins more often than not. And everything from 1v3 down is a long shot: losing a 1v4 is the expected outcome about 98 times in 100.
Context moves the odds
Side and plant state shift each cell. A 1v1 as the attacker after the plant is ~73% - you only have to deny the defuse. The same 1v1 as the defender post-plant drops to ~50%, because you must win the fight and still defuse in time. Both offsets are fitted from measured cells in the corpus.
CLOE: clutches over expected
Raw clutch win rate is misleading because it depends on the mix of situations you were handed. A player fed mostly 1v1s should win a lot of them; a player who keeps ending up in 1v3s should not. CLOE fixes this by subtracting what the model expected from what you actually won:
CLOE = clutch wins - sum(xClutch)xClutch is the modeled win probability of each attempt. CLOE above 0 means you beat the odds; 0 is exactly expected.
The arithmetic makes the point: winning a single 1v3 adds about +0.95 clutches over expected, while losing a 1v1 costs about -0.61. A converted 1v4 says more about your clutch skill than a run of comfortable 1v1 wins, and CLOE is the stat that knows the difference.
Multi-player disadvantages
Riptide also tracks the states just above a clutch: rounds where your team was outnumbered but you were not alone (2v3, 3v5, and so on). These never count as clutches - they are scored separately as disadvantage conversions, against their own measured base rates:
| State | Conversion odds |
|---|---|
| 4v5 | ~48% |
| 3v4 | ~43% |
| 2v3 | ~26% |
| 3v5 | ~22% |
| 2v4 | ~10% |
| 2v5 | ~3% |
Empirical per-state conversion base rates from the corpus (side-blind).
Two useful reads: a 4v5 is nearly a coin flip, so consistently losing them is a mid-round problem, not bad luck. And a 2v5 converts about 3 times in 100 - which is why saving weapons in that state is usually the higher-value play.
How clutch feeds Impact
The clutch axis of the Impact rating is CLOE per attempt, shrunk toward zero for small samples so a lucky weekend cannot masquerade as clutch skill:
clutch axis z = (CLOE / (attempts + 6)) / 0.04 weight in Impact: 0.156 pseudo-attempts shrink small samples toward expected; 0.04 is the between-player spread of CLOE per attempt.
How to read your clutch numbers
- Low attempts mean noise. A handful of clutches tells you almost nothing - that is exactly why the rating shrinks small samples toward expected.
- A high raw win rate might just be a friendly mix. Lots of 1v1s inflate it. CLOE is the fair read because it prices every attempt by its difficulty.
- Do not chase 1v5 heroics. The model expects under 1%. Saving your rifle is often the play that wins the next round.
- Watch the 4v5s. At a ~48% base rate, dropping most of them points at trading and mid-round calls - fix those with the trade fundamentals.
Keep reading
Clutch is one of six Impact axes - see the full Impact breakdown for how they combine, or the stats glossary for every definition in one place.