15 February 2016 - 04:37

What Champions contribute the most to Wins? Introducing Adjusted Winrate and Wins Added

In professional League of Legends, Champion statistics are a field that faces significant challenges when it comes to drawing conclusions about Champions actual strength.
Dot Esports

In professional League of Legends, Champion statistics are a field that faces significant challenges when it comes to drawing conclusions about Champions actual strength. The sample size of Winrate data we have for each Champion is of questionable size for even a simple system, much less one with thousands of variables like LoL. Even ignoring all the micro variables within each game, simply the variables of 10 different players and 10 different champions each game would require a sample size in the thousands in order to have certainty in a single Champions Winrate being significant. Unfortunately, a thousand professional games may not even be played in a single season, much less for a single Champion, and even further less on a single patch.

My goal with these statistics I'm introducing today is to remove one of the key variables in Champion Winrate in order to strengthen the legitimacy of the limited sample we do have. That variable is team strength. It is easy to look at a champion with a 60% Winrate and think they must be a powerful Champion, but what if the teams that played the Champion average a 59% Winrate? The champion is more likely average than powerful in this case, as they are not actually contributing much to increasing their teams chances of winning.

This method of valuation for team strength certainly isn't without it's own sample size issues, as a teams record 10 games into the season could easily not be indicative of their overall strength. That being said it's the most objective measure of team performance we have, and because the statistic will be taking into account the combined record of every team who plays a given Champion, this sample size issue is mitigated to a significant degree.

Now lets go over the statistics and the results so far this season. All data is from the EU LCS, NA LCS, and LCK through week 5 this season, as these are the regions I have been covering and had the time to log data for. For the an example showing how to calculate these statistics, we will look at Zed.

Adjusted Winrate is simply the difference between a Champions Expected Winrate and their actual Winrate. Expected Winrate is the average Winrate of the teams that have played the Champion, weighted by how many times each has picked the Champion. It is calculated by taking the win % of each team that played the Champion - again weighting for number of times the team picked the champion - and converting it into a decimal. The decimal totals are then added together and divided by the number of Picks total in order to get your Expected Winrate. Keep in mind that we are using individual game Winrate here, NOT BO3 or match Winrate. In the case of Zed:

8 Picks total
1 Pick by Afreeca Freecs - 29% Winrate - 0.29 expected wins
1 Pick by G2 - 80% Winrate - .80 expected wins
1 Pick by Splyce - 30% Winrate - .30 expected wins
3 Picks by UOL - 70% Winrate - 2.10 expected wins
1 Pick by CLG - 60% Winrate - .60 expected wins
1 Pick by IMT - 100% Winrate - 1 expected win

Now we add the expected wins together, divide it by the 8 total Picks, and convert it back to a percentage.

.29 + .8 + .3 + 2.1 + .60 + 1 = 5.09

5.09/8 = .636

63.6 % Expected Winrate

Zed's actual Winrate in these games is 62.5%.

This means his Adjusted Winrate is -1%.

This tells a much different story than if you were to just look at his 62.5% Winrate and assume he's a strong champion. His teams actually have a slightly lower Winrate in games with him than without him, so his results this season have actually been slightly negative rather than very positive as they would appear.

Obviously 8 games is an extremely small sample size and this particular data point is rather meaningless, but you can see the value in removing team record from the equation.

Wins Added makes use of the same formula, but attempts to balance sample size with the strength of the Champions performance. It is simply a champions Adjusted Winrate multiplied by the number of times they have been picked, thus balancing how big an influence the champion has on their teams Winrate with how many times they have been played in order to show how many wins they have added (or subtracted) from their teams over the course of the season.

Now that the explanations are out of the way, here's the results for every champion played across these three regions (EU,NA,KR) this season. (Wins Added won't exactly match what the results would be based on Adjusted Winrate and Picks shown due to Adjusted Winrate rounding decimals that Wins Added does not).

The table is sorted by Wins Added.



ChampionWinrateExpected WinrateAdjusted WinrateWins AddedPicks
Lee Sin51.72%49.24%2.48%0.720355706429
Next Article