Early vs. late game champions

Patch 4.20

This week I'm going to start looking at a new topic, game durations. To keep you on your toes, I'm going to put the best plot first this time.

game durations

This week's data set is big, but not huge; over five thousand matches played by challenger and master elo (mmr) players in NA. It encompasses every solo queue match played by approximately the top 1000 players since the release of patch 4.20 a couple weeks ago.

Now that is a beautiful histogram. The median game appears to be just over 30 minutes, and the big spike at 20 minutes must be due to surrenders. Comically, you can't tell from the data api whether or not a game ended in surrender, so further data on that will have to wait.

The plot is almost predictable enough to be boring, but it is nice to see what you expect from time to time. Ha!


Things get more interesting (and significantly more statistically dicey) when we split this dataset up by champion.

Fiddlesticks game duration results

This is a plot of the durations of games won and lost by teams with Fiddlesticks.

He seems to peak a hair before 30 minutes in wins, and suffers a bit if the games go on longer. I have had my eye on the patch 4.20 jungle last week and the week before, and fiddle has been doing well.

This is not too interesting on its own, here are similar plots for a few champions traditionally thought of as 'late game' champions.

Nasus game duration results

Nasus wins most often in the 30 to 40 minute range. The bump in his wins at 20 minutes is interesting. Are people more likely to surrender to a Nasus? The steep drop-off in his losses after 40 minutes is also consistent with him being a late game monster, I think.

Tristana game duration results

Tristana has a steep peak in wins around 30-35 minutes. I haven't run the numbers, but I would hazard a guess that that's around the time of hitting level 18. Her attack range passive makes a huge difference, and I think could contribute to games ending in victory at that point. Her losses are more flat, with some data points in the 25-30 minute range.

Vayne game duration results

Vayne actually looks quite similar to Tristana. Though she does seem to have a propensity for being in shorter games. However, she doesn't lose much after 40 minutes either, similar to Nasus.

Kog'Maw game duration results

Kog'Maw actually wins most of his games early (at least at this mmr/elo). The sample size here is a bit small, sadly, but his losses fall off sharply after 40 minutes also.

The preseason changes feel a bit unfavorable to Kog'Maw, to me at least. As a glass-cannon with no real escapes, I think he'll always fare better in competitive play. I look forward to seeing how he does in the season 5 LCS.


Next lets look at some champions that are considered 'early game' champions.

Lee Sin game duration results

Riven game duration results

Lee Sin and Riven look extremely similar from this vantage, and at least somewhat different from the late game champions above. Their win plots are flatter, and taller in the 20-30 minute range when compared with both the late game champions. They also have many more wins in the 20-30 minute range when compared with the first plot above, with all the games.

Their reputation as early game champions may be deserved.

Shyvana game duration results

Shyvana is also known as an early game champion. This sample size here is pitiful, it would be fun to have access to more data and see these plots fill out. Take this one with a grain of salt.


And here is the plot for Fizz.

Fizz game duration results

The shapes are reminiscent of Lee Sin and Riven. Do you think of Fizz as an early game champion? I guess I kind of do.

With this data for every champion, and some clustering analysis, one could create a nice little categorization of early vs. late game champions. Maybe an article for the future.

Is your favorite champion not listed? Let me know and I can add them.

There's obviously a lot more to say about game durations, I am sure that I will be revisiting this topic more in the future.


Easy histograms, easy life. Enough #LeagueMath for one day. Did I flub the analysis here? Let me know.

No IEM San Jose spoilers here, stay cool and close; I'll be back with more league + math fun fun fun here at LeagueMath.com.


No histogram binning algorithms were harmed in the making of these plots. Big ups to Mathematica, all you blind monks, and Nasus the pimp-cane wielding dog of top lane.

Peace.