First Day Stats

I figured it'd be interesting to share the first day's stats for TR. Keep in mind that this is with one server of each gamemode, stats are dumped at 5 minute intervals, and that the wait time is the average of the last 5 games (so if there's not constant games, the wait time stat will increase indefinitely). We'll look at addressing some of the more obvious defects asap, but it's quite interesting to see how the numbers turn out.

In both line graphs, grid lines are 1 hour (so we're still shy of one day here).


Total Players and Games
(the drop-off is where the date ticked over on the server)
Starting simple, this is the total number of players recorded and total number of games played. You can see the system has handled almost 300 games in under a day, with about 150 players participating.


Players vs Wait
Here we see how the instantaneous players varied over time, along with the average wait in minutes. The relationship is pretty clear here, but we'll take a closer look soon.
Without actual UTC times (which were unfortunately missing from the data set this time around) we cant say exactly what regions people are most likely playing in, however from about 10 to about 14 hours after the system went up, there was a peak period. Interestingly, there were no points recorded where only one player had been active in five minutes - it was either zero or two. This implies that people are generally jumping on with a friend, or else just missing each other as both disconnect seeing an empty lobby. We'll have to do more research into the exact reasons (and make sure it's not just a bug in the stats recording!).


Players vs Wait - Scatter
Here we plot the wait time per game data against the player count directly - and an obvious trend emerges. As player count goes up, wait times become less varied, and decrease to a reasonably stable two to three minutes. This appears to be the current maximum throughput of the system, and coincides with the observed average game time of a Run game. With a single game server servicing the lobby, this makes sense. As player counts increase, we will need to add more game servers to decrease the wait per-player.


Players vs Wait Factor - Scatter
This is the players versus the total waited time - rather than the time between games, the cumulative time spent waiting by all players. We can see that the optimal player count is currently around 5-15 in terms of minimising both the wait factor and the per-game wait time. This basically confirms what we saw above - we'll need more infrastructure to service more players.

I hope this has been interesting to some of you. We'll be continuing to monitor stats and might do more public analysis like this if there's interest!

Take care!

 

Definitely interested in seeing more stat breakdowns like this. It's nice seeing some hard numbers and comparing them to the less real numbers in my head.

 

You can see more here https://forum.thd.vg/threads/preview-weekend-overall-stats.24727/