Does damage win games?
tl;dr Yes... but we still must play as a team.
The post-game lobby offers some nice statistics that are both fun and helpful for analysis. Lately and in my experience, most players (and commentators!) are hung up on who did the most damage. Of course, doing damage is fun, especially to champions, but does it really decide the game?
That question is a bit hard to answer, but by bravely diving into the pool of data with clenched fists we can potentially emerge with some insights. This week I used results from 1,418,962 recent (patches 5.3-5.10) ranked NA solo-queue matches including players of all skill levels. This is broad enough that the set of matches from each league (bronze up to challenger) supports some interesting results.
The first thing a carry looks at after the game is, 'Did I do the most damage to champions?' – and the conclusions range from 'Yes, that is why we won.' to 'My team sucks, I did the most damage to champs and we still lost.'
What can be said?
This first table shows, for each league, what percentage of the time the team with the player that did the most damage to champions won the game:
|League||Player Win% (damage to champs)|
Two things to notice here. First, these numbers are pretty big. About 80% conditional probability of winning is high. This is comparable to other conditions that win games, like being the first team to take an inhibitor. So, yes, being the player who did the most damage to champions will help you win.
The second thing to notice is perhaps more interesting though. These numbers are monotonically increasing with player skill. Better players win more often when they do the most damage to champions (even when their opponents are also better). This could have many implications about the possibility of carrying games, and how that varies with the likelihood of players making mistakes (lower ranked players make more mistakes). Fun things to think about, dear reader.
This second table is similar, but shows what percentage of games the team with the player that did the most total damage, as opposed to only damage to champions, won:
|League||Player Win% (damage)|
These numbers are similar, but again, there are two things to notice right off the bat (was that a baseball metaphor?). First, these numbers are higher than those in the first table. That means having the player who did the most total damage on your team correlates slightly better with winning than having the player who did the most damage to champions on your team. Farming is important, kids.
The second thing to notice is that these numbers also increase with player skill. The fewer mistakes you and those around you (both teammates and opponents) are making, the better these stats predict victory.
League of Legends is a team game (such a controversial statement). I would like to now look at similar tables, but at the team level, instead of the individual player level.
This table shows, per-league, what percentage of games are won by the team that did more damage to champions:
|League||Team Win% (damage to champs)|
Happily, these numbers are even larger than those in the previous two tables. Maybe League is a team game after all. I jest. And, again, these numbers are increasing as we head down the rows, and up the skill chain. It is reassuring when data affirms suspicion, but the sheer magnitude of these numbers is potentially surprising. Separate from destroying Nexuses and forcing forfeit, I do not know of many things that condition a 90% win rate.
For completeness, here is a table that shows what percentage of games are won by the team that did the most total damage (not just damage to champs):
|League||Team Win% (damage)|
Higher yet, astounding. Also interesting to compare the values here to those for single players damaging champs, especially in lower leagues. This could be construed, by a sufficiently motivated or deranged individual, as evidence that farming minions as a team trumps solo kills in lower elos (MMR).
I love data.
Before I call it quits for today, I want to look at the same data sliced one more way.
Here is a table that shows, per-league and per-role, what percentage of the time that role does the most damage to champions in a game:
Here, 'Bot' means a carry-style champion played in the bottom lane with a traditional support duo laner. These numbers in each row do not quite add up to 100% for a couple of reasons. Occasionally bottom-lane 'support' champions end up doing the most damage. I am looking at you Luden's Echo rush Nidalee/Vel'Koz. Also, some fraction of games are not played in a way that the classifier can identify player's roles. I am looking at you double ranged AD top troll comp. Actually, I really like non-meta picks, but they do complicate analyses such as this one.
To me, the striking features of this table are that as top-laners skill up, they do significantly less damage to champions, and that bronze marksmen stink. Besides that, the gradients as we move up/down this table do not say much. It is interesting to see how the damage to champions is split up by role, though.
This final table is similar to the last one, except it shows total damage, not just damage to champs:
The jungle and bot numbers have gone up, and the mid numbers have gone down. This makes sense, since jungle and bot are the roles that do the most damage to creeps. Again, top-laners are doing much less damage at higher elos (MMR). This could reflect a higher incidence of tank champion picks at high skill levels (a topic for a future article?), or other strategic shifts in the game as player skill improves.
Skilled marksmen do a larger share of their team's damage, which could be something for lower-ranked players to think about. The jungle numbers are a bit all-over-the-place. It may be worth mentioning here that since there are so many fewer master/challenger level players, my dataset has many fewer master/challenger level games. The data in the bottom two rows of the tables in this post will be noisier.
The end of each post is like the deep end of a pool. "Ok, we have made it to the bottom... Now what?" Well, we have no choice but to return to the surface; to face the day. Here are some suggestions:
- Go look again at how much better it is for your team to do damage than it is for you to do damage to champions.
- Follow me on twitter and blast me for what you perceive to be my lack of statistical rigor.
- Tread the index and dive in again for more insights.
- Try the rss feed on for size; it is your best bet for knowing immediately when the next post goes up.
League and math are like chips and salsa; except you can run out of chips and have salsa left over, or run out of salsa and have chips left over. The league never runs out and the math never runs out here at LeagueMath.com.
No feelings of insta-lock bronze marksmen mains were harmed in the making of this article.