The first picture, the bar graph, is sorcus’s game analytics - it is showing how many people left within what time.
In that server, there are 6 people who still haven’t left after more than half an hour straight. I feel like players leaving after literally a few seconds who haven’t loaded the game should not count toward the second picture, the website stats, as it is misleading and skews averages.
[Edit: another ninja request would be to not count a visit if the player only visits for literally 2 seconds. Not sure how hard that would be to code.]
Edit2: Stravant’s idea is way better. Instead of just having averages on the site, make it a histogram like in the first picture, that way we can see the bounce rate too
An average playtime is actually a pretty bad statistic to use, which is why the analyser gives you a histogram. All parts of the histogram are actually quite important (How much fast dropoff there is, how long the typical session is, and how many outliers there are with very long playtimes).
I don’t think it’s worth worrying about an average playtime, if anything the website statistics should be formatted to give out data as a histogram.
You can’t just ignore players who leave your game after a few seconds. It’s a metric known as “bounce rate” and can be really useful. If a lot of players are leaving your game a few seconds after joining, there might be an issue that’s causing them to be unable to connect.