An item's win rate mostly shows how the game was going when it was bought. The highest win rate item is a little better than the popular one. Not as much as a stats page says.
We took 9.2 million ranked solo games on patch 16.18 and followed 229,717 of them minute by minute, 8.4 million finished items in all. We repeated everything on patch 16.19.
How can you tell if an item's win rate is lying?
Compare it with how likely its buyers were to win anyway.
The minute you finish an item, your gold, kills, towers and objectives already say how likely you are to win. If an item's buyers win more often than that, the item is really better. If they win exactly that often, its win rate is just the game state.
The chance comes from both teams' gold, kills, towers and objectives at that minute.
Why do later items have higher win rates?
Because you only finish a fifth or sixth item in games you're already winning. Win rate climbs with every item you finish.
The two lines are the same. So an item you usually buy fourth will show a higher win rate than one you buy first, whatever it does.
First items come at minute 11 and win 50.1% of games. Sixth items come at minute 36 and win 57.9%.
Is the highest win rate build better than the most popular one?
Barely. It looks several points better on a stats page. It's really 1 point better.
The later the item, the bigger the gap on a stats page. That's just its buyers being further ahead.
Across 1,472 champion, role and slot combinations. Patch 16.19 gives the same 1 point. It isn't better players either. Both items' buyers win their other games about as often.
Which high win rate builds are really better?
A few high win rate items really are better. Others only look better because their buyers were already ahead.
Patch 16.18. How many points more the highest win rate item wins than the most popular one, on a stats page and really. The first three are really better. The last three only look better.
Thresh players who rush Boots of Swiftness win 52.9% of games, and Locket players 49.7%. But the Swiftness buyers were already 3.9 points more likely to win.
Which items' win rates lie the most?
Most item win rates are just the game state. Mejai's Soulstealer and Guardian Angel have the best win rates and really are better. Umbral Glaive has the worst and is perfectly average.
Patch 16.18. Really better compares buyers with players on the same champion and role who had the same chance of winning.
Umbral Glaive's champions lose more than their chance says, whatever they build. The item itself is average.
97% of the 109 most bought items are within 1.5 points of average.
How much should you trust item win rates?
Build the highest win rate item. It really is a little better. Just shrink what the stats page tells you. A 5 point lead is really about 1 point.
The later the item, the more you should shrink it. A big win rate gap on your fourth, fifth or sixth item is mostly the game already being won.
1 point is about one extra win every 100 games. A few picks are worth more, like Immortal Shieldbow first on Yone or Infinity Edge second on Tristana.
