Tech stats

Uplift answers a single question: does researched this tech help you win more often than chance alone would predict? Win % alone can’t answer it — a tech that shows up in mostly-2-player games naturally has a high win rate, while one that only appears in 6-player games is fighting longer odds. Uplift normalises that out.

The formula. Every time the tech is researched in a finished game, we record a baseline expected win of 1 / N, where N is that game’s player count. Sum those across every pick and you get the Expected wins column. Subtract that from the actual Wins column and you get Uplift = Wins − Expected wins. It’s an absolute count of wins above (or below) the random-chance baseline, not a rate.

Worked example. Suppose Plasma Cannon was researched 400 times across games that averaged 4 players. Each pick had a 1-in-4 random chance of ending up on the winner’s sheet, so the baseline expected wins are 400 × ¼ = 100. If Plasma Cannon actually shows up on the winner’s sheet 130 times, the uplift is +30 — winners researched it about 30% more than chance alone would predict.

How to read it. Positive uplift means the tech over-performs the baseline — winners reach for it more than chance predicts. Negative uplift means the opposite — it gets researched but doesn’t convert to wins at the random rate. Zero is exactly chance.

Small-sample caveat. Uplift is unweighted by sample size. A tech with only a handful of picks can swing to a big positive or negative number on coin-flip luck alone. Cross-check the Picks column before trusting an extreme uplift — and treat anything with under ~20 picks as noisy.