Who will you pick in Ranked?

Using M2 recommendations strategically

Best team predicted win rate — potential under current conditions. The highest predicted win rate achievable with this candidate under the current model, mode, map, fixed allies and supported pool. If teammates can agree on and form that complete team, use this value to choose the best team.

Teammate mean — a uniform average when teammates are undecided. An equally weighted average of predictions for every supported complete team containing the fixed allies and this candidate. Actual pick rates and preferences are not weighted. For early picks without coordination, it is a secondary criterion for expected performance under equally likely remaining combinations.

The candidate with the highest best-team prediction is not always the best early pick. If that potential depends on a particular teammate, the completed team's prediction can fall sharply when that teammate is not chosen. Unequal real teammate choices can also change the expected-value ranking.

The mean does not measure the floor. A higher mean does not guarantee a higher minimum or less variation between combinations. Assessing the floor requires separate distribution information such as the minimum or a lower quantile; the current mean is not presented as a floor or safety score.

Illustrative predictions: candidate X has three completions at 90%, 80% and 10%, averaging 60% with a 10% minimum. Candidate Y has 59%, 59% and 59%, averaging 59% with a 59% minimum. X has the higher mean but the lower floor. These are not actual candidate results.