The 9-3 curse, measured
Every VALORANT player has watched a 9-3 lead evaporate. We track win probability from every score state across millions of ranked games. Here is the live tool — and a check of whether 9-3 really is special.
- A team up 9-3 at halftime wins — of the time (— games). The "curse" is a 1-in-12 event that feels like more because it hurts.
- 9-3 is not special: every halftime lead converts a few points less often than a pure coin-flip model predicts (9-3: —, 8-4: —, 10-2: —). The side swap and momentum swings cost every leader about the same.
- Use the tool below to check any score on any map, side and rank bucket.
Live win probability from any score
Pick a map, which side you started on, and a rank bucket. Click any cell in the heatmap or use the score controls. Overtime is broken out separately so you can see how much of a comeback runs through 12-12.
Win Probability Heatmap
Your score (rows) vs their score (columns). Green = favored, red = unfavored. Click a cell to inspect it.
Is 9-3 actually cursed?
"Cursed" only means something compared to an expectation. The cleanest expectation is a coin-flip model: if every remaining round were a 50/50, how often would the team up 9-3 reach 13 first? We compare the observed conversion rate against that model for every halftime lead.
Every lead — 7-5 through 11-1 — converts a few points below the coin-flip model, and 9-3 sits right in that pattern rather than standing out from it. Two things pull every leader down: the side swap at halftime (the half you dominated is over, and the other team now gets the side that suited you), and the fact that a lopsided first half often reflects one hot side rather than a large skill gap. None of that is unique to 9-3.
What makes 9-3 feel cursed is base rates. Far more games pass through 9-3 than through 10-2 or 11-1, so even a 1-in-12 collapse rate produces a steady stream of painful memories, and a blown 9-3 is far more memorable than a routine 13-5.
Method
- Source. Round-by-round score trajectories from competitive matches collected by OTPStats via the official Riot API, all six regions. Every score state a game passes through counts once from each team's perspective.
- Win probability is the share of games that reached a state and were eventually won by that team, overtime included. Regulation and overtime outcomes are also reported separately in the tool.
- Map, side and rank splits use the map played, the side the team started on, and the average rank of the lobby. Sparse buckets fall back to the all-ranks numbers.
- Coin-flip model: each remaining round is an independent 50/50, first to 13 wins, and 12-12 is settled by a 50/50 overtime. It ignores skill gaps and side strength, which is exactly why the gap to it is informative.
/#curse now land here.