Build rater

How it works

Everything comes from public match data via deadlock-api.com, for normal-mode matches from October 6, 2026 to 2026-10-07 14:00 UTC. The rating has four parts:

1. Item score. For every item, we compare the hero's win rate when they bought it with other items of the same price that the same hero bought in the same minute of the match. That removes most of the "rich teams buy expensive items" effect. The build's score is the cost-weighted average of its items. Small samples are pulled towards what items with a similar pick rate usually score, so an item bought 80 times can't claim a huge number. Items almost nobody buys on that hero count as −2.16 points, because across all heroes that's how rare picks perform.
2. Grade. S is +1.25 or better, A +0.6, B 0, C −0.6, D below that. For scale: each hero's twelve most-bought items usually land between C and A, and random builds mostly get a D. Each major warning (no Vitality spike, or far less spending on the hero's main damage category than usual) drops the grade one step, two at most. Impossible builds (more than 12 items or 4 actives) get no grade.
3. Checks. Rules from our build theory guide: slots, actives, components, the 4.8k investment spike, and how the build's spending compares with what players of that hero usually buy.
4. Matchups. The hero's own win rate against each enemy, plus how the build's items usually shift when that enemy is on the other team. "+2.0" means about two percentage points more wins than the hero's average. Switch to Lane opponent to count only games where that enemy was in your lane.
5. Item pairs (experimental). Pairs of items in your build that win notably more or less together than each does alone, after adjusting for their combined cost. These are shown as hints and don't change the grade.

What it can't tell you

• It measures association, not cause. Players who buy an item may win for other reasons, such as skill or being ahead.
• Defensive items look worse than they are, because players buy them while losing. Judge them by the matchup section, not the item score.
• Core items can look worse than they are, because everyone buys them, including players who are behind, while players who are ahead can afford greedier picks.
• The grade scores items one at a time. Item pairs are shown separately because, with thousands of possible pairs, some results are luck, and pairs with counter items (such as Inhibitor) mostly reflect the games they're bought in. Buy order isn't modelled yet.
• Matchups look at one enemy at a time and ignore the rest of both teams. Lane matchups have smaller samples.
• All ranks are mixed together, and new heroes have very little data.
• Resist totals only count permanent resist, not active items or bonuses below half health.