Following on my analysis of the average LoL game, here’s some scatterplots and histograms of various statistics for games. Specifically, for each game I’m looking at match length (0-3600 seconds is typical), best ping (10-150ms is typical), median ping (10-150ms is typical), and packets lost / minute (0-1 is typical, but long tail).
A huge caveat for this data: this is not a valid statistical sample of all LoL players. This is a random sample of Logs of Lags users, almost certainly biased towards people with network problems. I don’t intend to publish this data widely because it’s not a fair sample.
Honestly, the scatter plot is not the best treatment here; too many dots, not enough correlation. The histograms at top and right examining single variables are probably more interesting. Some notes:
- Median ping and best ping are pretty tightly correlated. No big surprise there; they are nearly identical on a connection with low jitter and no packet loss. Not sure about the diagonal lines above the x=y line, maybe an artifact of how the game synthesizes a fake ping number in the presence of packet loss?
- People’s pings are an interesting bimodal distribution, roughly peaking at 50ms and 110ms. I can’t account for the relative unpopularity of 80ms. Well if this were all North American data I’d say it was west coast vs. east coast, but I get a fair amount of European uploads too.
- Packet loss / minute has a lovely long tail distribution. FWIW, I call anything over 5 a “bad link”; that’s ~30% of my logs, perhaps I should move that number up.
- I was hoping to see evidence of correlation between packet loss and best ping, on the theory longer connections would be worse. I don’t see any evidence of that in the picture, it’s just a mess of dots. Need a better analysis tool. Same problem with packet loss vs. match length
- Similarly, I can’t draw any conclusion from median ping vs. match length. That previous graph I did suggests the average ping for players does not go up during the game.