What it does
You sign in with Steam, pick a recent match and read its report: the win probability curve, the moments that weighed the most and what each player contributed in that match.
See a sample reportWho is behind it
Deadlock Dispatch is made by one person, Ridel Saavedra Flores, from Santa Cruz de Tenerife, Spain. It is a personal, independent project: there is no company behind it and it has no connection with Valve.
Trained on replays of the game
The curve does not come from a formula. It is a model trained on almost 4,000 replays of real ranked matches, close to a terabyte of them: from every moment of each one it learned who went on to win.
- ~4,000
- replays of ranked matches, all from one patch (3,969 exactly)
- ~1 TB
- of compressed replays (950 GB)
- 1.1 pt
- average gap between what it predicts and what happens, over 592 matches it never saw
What this paper does not print
ScoresRankingsMVPStreaksAveragesAdvice
No player names or identifiers of anyone either: of the twelve players in a match, only the hero appears.
How to read the figures
The cost is how much less your team wins, on average, after that moment: association, not cause. It is not a player score.
Contact
For any question or problem, write to support@deadlockdispatch.pro