Path scoring stays slot-aware.
A third-item read only compares against games that reached a third item. The site stops rewarding late purchases just because those games were already winning.

Gold Diff is a League build lab for people who want more than win-rate bait. It reads item order, rune pages, shards, boots, and spell pairs through contextual MWPA so the site can tell you what actually held up, not just what showed up in already-winning games.
Win rate bait is not a build path.
An item can post a 58% win rate and still be a trap because it only gets bought in games that are already won. Gold Diff filters that noise by comparing every build choice against a contextual baseline, not against the global average.
A third-item read only compares against games that reached a third item. The site stops rewarding late purchases just because those games were already winning.
Primary tree, secondary path, and shards are all sampled together so the rune read still reflects real pages people actually ran.
Meaningless edge roles are filtered out, but champions with real off-role usage still surface when the sample actually supports it.
Don't let Boris cook.
One page to test item order, full rune pages, shards, boots, and spells, before Boris talks you into something cursed. The composite signal tells you whether your loadout outperformed the champion's contextual baseline, or whether you just got talked into a pretty tooltip.
All build components stack into one directional read.
Item order is scored at each checkpoint, not in aggregate.
Rune scoring uses real page combinations, not isolated picks.
Each champion page lets you assemble a full loadout with item path, rune page, shards, boots, and spells, then read the composite signal before you lock in.
Start evaluating builds
before Boris talks you into something cursed.