How the tier list score works
The score is win impact, presence and mastery, mixed at 0.5, 0.25 and 0.25
Every number on this page comes out of the same response the tier list is built from. When the server retunes a weight or a floor, this page says the new number the next time it builds. Nothing here is typed in by hand.
484 champion and role pairs clear the 30 game floor on patch 16.14. 308 do not, and get no score.
+0.5 wins per 100 games(range -0.6 to 1.5) 6,176 games Pooled 24% of this is borrowed from similar championsthe range covers zero, so this could go either way
The score runs from 0 to 100 and it is not a win rate. A win rate answers "did this win", which over a few hundred games is mostly a question about luck. The score answers three things at once: does this pair beat the average pick in its role, are we sure it is really being picked there, and does practice on it pay.
The first part is the one that carries the weight, and it is the one that is pulled hardest toward the role average when the games are thin. That pull is what stops a champion with two lucky weeks from sitting at the top of a list.
These numbers come from 108,616 ranked games we recorded on EUW, Korea and NA, found by following Diamond players and the people they play with. That is not every rank and it is not every region. Everything on this page is patch 16.14. Counted up to .
What goes into the score?
Three parts, fixed weights, both read from the response.
| part | weight | what it measures |
|---|---|---|
| win impact | 0.5 | How the pair does against the average pick in its role on this patch, pulled toward that average by how few games it stands on. |
| presence confidence | 0.25 | How sure we can be that the champion is being picked in that role at all, rather than turning up in a handful of games. |
| mastery | 0.25 | How much better the pair does for players with practice on it than for players on their first games with it this patch. |
Each part is scored out of 100 on its own, then the three are mixed at those weights. A champion can therefore be strong and still score in the middle, if it is barely picked in that role or if practice on it does nothing measurable.
Why is win impact not the win rate?
The part of this page that does the real work.
A raw list against a pulled one
What a raw sorted list does
A win rate answers "did this win", which over a few hundred games is mostly a question about luck. Sorting a raw list picks up the largest luck along with the largest effect, which is the trap the methodology hub opens with. That is how a 60% win rate champion ends up sitting on 40 games at the top of a list.
What the pull does instead
Win impact is how far a pair beats the average pick in its role on the same patch, counted in wins per 100 games. Before it is ranked against anything, it is pulled toward that role average, with the pull worth 2,000 pseudo games.
The arithmetic is simple enough to do in your head. A pair with 30 games gets almost all of its number from the role average, because 2,000 pseudo games swamp 30 real ones. A pair with twenty thousand games keeps almost all of its own number. In between, the share borrowed from the role is the pull strength divided by the pull strength plus the games.
Pulling first costs a real champion nothing: a genuine edge over thousands of games survives the pull almost untouched.
When a number is mostly borrowed, the estimate says Pooled rather than Measured, and the chip on screen tells you which one you are reading.
The mastery part, and what it is not
Measured from the same store, with its own floor.
For every pair we split the recorded games in two: players on their first games with the champion this patch, and players with more than a few. Each side needs 30 recorded games of its own, which is a separate floor from the one for scoring at all. The mastery part of the score is how much better the practiced side does.
When either side is too thin, the split reports nothing and the mastery part takes a neutral 50 out of 100. It does not guess, and it does not quietly drop the pair off the list either. You can see which pairs have a real split, because the ones that do print the games on each side.
This is an association over recorded games, not a promise about your next game. A champion where practice pays is a champion that punishes a first attempt, which is worth knowing before you pick it blind into ranked. It does not mean the champion will reward your practice specifically.
The tiers
Cutoffs on the 0 to 100 score, read from the response.
| tier | score at least |
|---|---|
| S+ | 90 |
| S | 78 |
| A | 64 |
| B | 48 |
| C | 32 |
| D | 0 |
When do you refuse to score a champion?
308 champion and role pairs on patch 16.14 get no number at all.
A pair needs 30 recorded games on the patch before we put a score on it. Under that it is listed with the games it has so far and no number, no tier and no rank. A number worked out from nine games is luck with a decimal point on it.
Those pairs are not missing and they are not hidden. They score as soon as the games arrive, which for a champion people actually play is a day or two into a patch.
| role | scored | seen |
|---|---|---|
| top | 124 | 171 |
| jungle | 77 | 131 |
| mid | 112 | 168 |
| bot | 79 | 155 |
| support | 92 | 167 |
Worked example: Lee Sin jungle on patch 16.14
The pair with the most recorded games on this patch, picked by the data rather than by us.
Read it in the order the score is built. Win impact against the average jungle pick:
+0.5 wins per 100 games (range -0.6 to 1.5) 6,176 games Pooled
The three parts scored 55 for win impact, 86 for presence and 55 for mastery, out of 100 each. Mixed at 0.5, 0.25 and 0.25, that is a score of 63 and tier B. You can do that arithmetic yourself from the four numbers in this paragraph.
The mastery split behind it: 5,692 games from players on their first games with the champion this patch, and 484 from players with more.
Against patch 16.13 the score moved from 58 to 63, measured over 16,780 games on the older patch.
How to check us
-
Open https://api.ko.lol/v1/tierlist?patch=16.14. The formula block at the end of the response carries the weights, the pull strength, the games floor and the tier cutoffs. Compare them with the numbers on this page.
-
Recompute a row. Each entry ships its games, its wins, its three part scores and its final score, so the mixing arithmetic is yours to redo.
-
Work out a pull. The share of a row borrowed from the role average is the pull strength divided by the pull strength plus that row's games. Compare it against the shrinkage figure the response prints for the same row.
-
Look at any other site's tier list for the same patch and count what sits next to each row: the games it stands on, and how sure they are.
What this score does not tell you
It does not tell you what to pick. It tells you what the recorded games say about a champion and role pair, in a high rank pool, on one patch. Your own games with a champion beat any list on this site, and a pair three points lower that you have played a hundred times is the better pick.
It also does not carry a matchup. Counters and lane matchups are not on this site yet, because the server does not measure them yet.
ko.lol for desktop
Your champion pool, your matchups and your patch notes, on your desktop while you play.
Free. For Windows.