Preview Weekend Overall Stats

Hey everyone, since there was some interest in the stats breakdown, I'll do another on the weekend overall. I'm aware it's not over everywhere yet, but as it's Monday here I'm calling it a wrap and pulled the data out for analysis.

As a side note, I also pulled the chat logs, and a cursory glance shows that most of the complaints people have are overlapping (chat, run maps/ease of making a run, and EU hosting being the main ones), with a few being divisive (various class interactions).

The takeaways if you cant be bothered looking over the graphs are:

• We'll likely stick up two more servers as soon as we can through the week, to allow multiple games of each mode and reduce congestion.
• Overall take-up has been modest but encouraging.
• Most of the community seems to be US based.
• Run mode looks to be more popular.

Player Activity:


Overall population - total players and total games
We can see an instance of a single server restart here. The last day was much more active. Overall average is around 150 players per day and 300 games per day.


Population peaks - Player activity scattered over time
Here we can see where the players are, to some extent, or at least what times are busiest. The server time is EU, so the main peak corresponds largely with the USA afternoon, evening, and night time. This is the only time where the player counts are actually consistent, with Pacific and EU "peak gaming" times containing a lot of noise. It looks like the majority of players at this point are based in the USA, so we'll definitely be running the next preview on US servers.

Wait Times:
These are more or less the same as yesterday, so I'll do combined analysis at the bottom.


Wait times vs player count


Wait factor vs player count

Here we can see that our day 1 suspicions are more or less confirmed, and we need more servers to support an instantaneous population more than 15 or so players. This corresponds pretty directly with the game server capacity, but is good to confirm. If the wait factor started going up before that, or levelled off earlier, that would imply that the games were not being properly filled by players, and slots were being wasted on bots.

We can see that run is, overall, more popular, because the wait time correlates strongly with the average game time of a run game, not a skirmish game. This is good news for us, because we can keep more players busy with a smaller number of run servers than skirmish servers - but it also means we need to focus on new run maps that address people's concerns as soon as we can.

Hope this has been interesting for you, until next time!
Max

 

Nope, not yet. We'll try to get more stats infrastructure together before the next preview week to start looking at things like classes picked, kills per class, kills and deaths overall, etc. This time around, these numbers are all we have as we were more concerned with when and how many people would be on, and what the wait time stuff would be like (if the backend stuff didn't crash!).

 

Does the player count show unique visitors or is it just cumulative? Because I know there have been days this weekend where I've logged in 5-8+ times looking for games over the day, as well as the 5-6 times I relogged after crashing KAG-style. The most people I've seen from a skewed EU time-perspective-thing has been maybe 15-ish.

 

It's based on names seen. You could grief the stats by logging with 10 different names, but A) it'll use usernames once we have a supporter version -> logins, and B) I didn't feel like exposing IP addresses to the scripting layer (and you can change them too ofc).

We're working on maps to help with this for the next test (or maybe the end of this one) - they're a bit longer, with a flat zone behind spawns that can help with defending back-capping, and generally have fewer separate paths to encourage conflict. We're not likely to up the round timer as it's a powerful motivating factor for aggressive play.

 

Yep, known but not going to be fixed - it's luck of the ping who gets what class, as the receiving server just has the names of both and what classes they're meant to have. Whoever connects first is Name and whoever connects second ends up as Name2. They may or may not have had the same ordering in the lobby.

Logins will fix the issue, except in the case of logging in twice with the same account.

 

Oh, yeah, also, in Run, there's a bug where people can share classes, you may have noticed it, I don't know.

When you get out of the truck, your class selection is remembered. If everyone on a team gets in the truck one at a time and chooses the same class, then everyone can be that class when you all get in together at the end.
It also works when people don't switch from assault.

Firing an engineer who is also firing their rocket confuses the fuck out of them.

 

Yep, its kinda related (kinda not) - but its the reason you see quite a few teams with multiple assaults. It's on the list already :)