PO Server Activity

Antar

Antar

is a Battle Simulator Admin Alumnusis a Top Programmer Alumnusis a Senior Staff Member Alumnusis a Top Community Contributor Alumnusis a Top Tiering Contributor Alumnus
So yesterday, Steamroll asked me if there was any way we could look at how active each tier is on the server at any given time.

Since all the battle logs are time-coded, this was actually a pretty easy thing to look at. So I wrote a few scripts, ran 'em, and here's the results:

graph01.png

(using one-hour bins)

That's kinda cool. You can definitely see quite clearly when the server went down around the 19th.

Next I looked at binning daily instead of hourly, to see if there were week-long trends.
graph02.png

(only showing OU, for simplicity)

No weekly trends that I can see, really--server downtime kinda screws that kind of thing up.

Finally, I decided to look at average hourly activity across the month (as in, what TIMES of the day are best for battling)
graph03l.png


Clearly, 5am EST is a bad time to battle, while the server is hopping when those US kids get off school around 4pm EST

Probably a better way to understand this data is to not look at "battles per hour" but rather "average wait time for a battle:"
graph04.png

(I realize this isn't actually frequency, but inverse frequency. Sorry).

So for OU, you're gonna be waiting around 10 seconds for a battle. For RU/NU, it's like 3 1/2 minutes during peak hours. For LC, well, you're gonna be waiting a while.

What about some of the other tiers?
graph05.png


Three hours for a Clear Skies battle. Five minutes (peak time) for Challenge Cup. Interesting.

Anyway, I hope you all find this interesting. Feel free to request data for other tiers.
 
Thanks Antar!

Anyways I originally had the idea to see how many people were in each tier at a given time, like 178 users in OU at 4pm for example. Turns out that was a total pain in the ass to do and I decided to ask Antar for help on it. These numbers are actually very informative for users so I think it would be a good idea to link to this on PO in the topic (or the joining message).
 
Looking as this, it's weird how the 17th (I assume) was so active. I can understand the 20th/21st (?) because that probably was a result of all the peeps *finally* being able to battle again.

But yeah, this is damn interesting. Maaaaybe you could post stuff like that on a regular matter (in the usage stats threads maybe?) if it isn't too much work? Because this is awesome! Thank you Antar for doing this great stuff! :D
 
The hourly activity graph was very informative! Seems like I'm always playing when player activity is ebbing :/ Thanks, Antar :D
 
Ok, I'm just going to put the obvious out there;

What the hell purpose is Clear Sky's serving? Obviously NO ONE is playing the damn thing is the average wait time is over two damn hours. Even Little cup beats that by about 4 times. So why do we still have it? I know there are still a few people that whine about weather, but clearly they aren't playing Clear Skies, so why do we have it? I was fully in favour of it when it was created (let them have their own tier if it bugs them so much) but it turns out it was either a very vocal minority or that they were all talk and no substance. It's just a waste of space now, so why should we even keep it?
 
Clear Weather was created on a whim by Aldaron. It isn't an "official" tier or anything. It's there to play if you want to play it.
 
Thanks Antar!
I found the hourly test to be the most interesting. Some big-time ladderers (is that the term? lol) may use this to their advantage (ex. when trying to get reqs or when testing a team). Pretty obvious that weekends would be the most popular. It seems Wednesdays are about the lowest in activity, which makes sense.
Anyways, really cool and very informative
 
This is actually neat. I do feel the need to point out that the scale difference in "Battle Frequency" grows on the way up, making it difficult to compare tiers with each other.

But anyways: thanks, Antar!
 
This is great thanks *looks at poster* Antar! I however think RU battlers for me been quite speedy. But this is a great thing.
 
Lesser Smaug said:
I do feel the need to point out that the scale difference in "Battle Frequency" grows on the way up,
Click to expand...

Yeah, it's a log scale. Otherwise, it would have been really hard to show everything on the same graph.


Limi said:
Maaaaybe you could post stuff like that on a regular matter (in the usage stats threads maybe?) if it isn't too much work?
Click to expand...

I'd like to, but I need to fine-tune the script. As it is, it takes nearly as long to do this analysis as it does to do my full usage stats analysis (showing me where the bottleneck is--reading the logs into memory). I'll keep everyone posted.
 
So that's why it always feels like forever when I press the find battle button for a LC battle... However I can't really tell what the best time to Get on the LC ladder is because there are so few of them! it just looks like a 180 degree line ;_;.
 
Vix said:
So that's why it always feels like forever when I press the find battle button for a LC battle... However I can't really tell what the best time to Get on the LC ladder is because there are so few of them! it just looks like a 180 degree line ;_;.
Click to expand...

If you look at the Battle "Frequency" plot, it appears that 3pm-4pm is your best bet for finding a Little Cup (and really any) battle. The average wait time is only about 500 seconds (8 minutes).

I will say that the server seems to have been a lot more active these past few days...
 
ReshiramBlackZekromWhite said:
Do you know what the DW Ubers frequency is? That's all I play...
Click to expand...

I didn't graph it, but take the number of DW Ubers battles in November (947 rated) and divide 1 month by that number.

Google Calculator is your friend: 46 minutes on average. If it's anything like DW OU, you can halve that during peak hours.
 
You must log in or register to reply here.
Back
Top