What is wrong with stats

All questions and comments about player stats should be addressed here.
User avatar
Posts: 2029
Joined: Thu May 26, 2005 8:40 am

What is wrong with stats

Postby The Spanish Inquisition » Tue Jun 26, 2007 6:17 am

It's not a question, or is it? We all know separate stats sites feedback into the actual game play and people start whoring out known exploits in the stats ranking system. This will always be the case regardless of the stat software being used, mainly because you can't recreate actual player intentions with their actions, so an algorithm does it's best to fill in the gaps.

If you rank by kills then you rank by time played. Total online time is a huge factor. Some folks only play a few hours a month while others play days. To compare these two is nonsense. You can argue that if you only measure things based on time spent on the server (kills per minute) then you eliminate any bonus/debit that length of time playing gives you in the ranking. A noble idea but this skews in favor for people who only play a little but happen to do good (they choose their time wisely).

I think the answer is somewhere in between. Players should be grouped by time spent playing. I'm not sure what the divisions or durations would be, maybe there is a natural breaks in the data but this would give players a better sense of where they rank against others who spend the same amount of time on that particular server. (Each server has it's own x factor as to the average skill so this would not be a measure of total playing time; some players may be playing on other servers.)

A totally random example of groupings based on time spent playing:
5hrs or less
5-15 hrs
15-30hrs
30+hrs

Also, I have added a new award for "most waking hours spent playing dod." This is based on 495hours of wakefulness in 30 days(720hr total) which is 7.5 hours avg sleep a night. The award is only valid on the last day of the month as it is incapable of determining days of the month. So in effect, you will see the percentage rise for all players as the month wears on, even if you stop playing. So if you have a percentage of 10% then you are spending that much of your awake time playing dod. Pretty simple.
Image Nobody expects the Spanish Inquisition!

User avatar
Posts: 1482
Joined: Wed Aug 17, 2005 9:37 am

Postby Cpt Ukulele » Tue Jun 26, 2007 7:23 am

So what happened to the Test B ranking? Combining several factors seemed like a good idea.
Image

User avatar
Posts: 541
Joined: Mon Sep 26, 2005 3:12 pm
Location: Calgary

Postby Denzildon » Tue Jun 26, 2007 10:30 am

How about ranking for every 30 mins of play time? Most people play at lease one map cycle anyways.
Image

User avatar
Posts: 2029
Joined: Thu May 26, 2005 8:40 am

Postby The Spanish Inquisition » Tue Jun 26, 2007 6:09 pm

Cpt Ukulele wrote:So what happened to the Test B ranking? Combining several factors seemed like a good idea.

It ran once I think.

I got a friendly call from some admin telling me that SQL call was pegging the cpu. The stats processing runs under a 'nice' call but any remote db calls it makes are not niced because they are running in a different process. It's on a shared server so you gotta be nice to others.

So, I can maybe run that Test B again once a day at like 3am and they won't notice but that means all awards get processed once a day too. You think that would be sufficient?

Denzildon wrote:How about ranking for every 30 mins of play time? Most people play at lease one map cycle anyways.

Like detailed stats for each mapsession?
Image Nobody expects the Spanish Inquisition!

User avatar
Posts: 1482
Joined: Wed Aug 17, 2005 9:37 am

Postby Cpt Ukulele » Tue Jun 26, 2007 7:19 pm

The Spanish Inquisition wrote:So, I can maybe run that Test B again once a day at like 3am and they won't notice but that means all awards get processed once a day too. You think that would be sufficient?


If it's not totally uncool for the other people on the server, I think once a day would be fine. Those are weekly and monthly awards anyway; probably not something players expect to be updated as soon as they log off.
Image

User avatar
Posts: 541
Joined: Mon Sep 26, 2005 3:12 pm
Location: Calgary

Postby Denzildon » Tue Jun 26, 2007 10:30 pm

Not really detailed per map but what I mean is and correct me if I have the wrong impression on how the stats work.

Currently the stats are computed like this.

RANK = number of kills, deaths, flag caps, achievements, etc. etc. calculated in an equation.

How about doing the calculation like this:

RANK = number of kills, deaths, flag caps, achievements, etc. etc. calculated in an equation / 30 mins

In this case, even if you have played all day long or just one map it would show your true skill compared to other players.
Last edited by Denzildon on Tue Jun 26, 2007 10:56 pm, edited 2 times in total.
Image

User avatar
Posts: 2029
Joined: Thu May 26, 2005 8:40 am

Postby The Spanish Inquisition » Tue Jun 26, 2007 10:46 pm

Total skill is a representation of mostly kills but also flag caps/blocks. I can turn on skill/30min. That will be a rough representation of skill/map.
Image Nobody expects the Spanish Inquisition!

User avatar
Posts: 541
Joined: Mon Sep 26, 2005 3:12 pm
Location: Calgary

Postby Denzildon » Tue Jun 26, 2007 10:55 pm

The Spanish Inquisition wrote:Total skill is a representation of mostly kills but also flag caps/blocks. I can turn on skill/30min. That will be a rough representation of skill/map.


In other words yes. Now some people would take some maps to their advantage but it's pretty hard to get those maps all the time!
Image

User avatar
Posts: 2029
Joined: Thu May 26, 2005 8:40 am

Postby The Spanish Inquisition » Tue Jun 26, 2007 11:16 pm

Have you seen who is on that list? Just a bunch of guys with no playtime and that is after I limited it to minimum of 4hours playtime. I bumped it to 8hrs.
As I mentioned, you can see the problem with doing stats by time, you have to set a minimum number of minutes/hours otherwise comlete no ones will all be listed first with highest skill. You can't please everyone with an arbitrary minimum filter based on how much you play.

The difference in skill/30min between people who have a minimum of 4hrs and a minimum of 8hrs is about a 800 points difference. Those who play more will fluctuate less and also gain less. I'm not really sure this statistic is significant enough to keep. If it fluctuates wildly with an 8hour minimum then it is garbage. Also because of decay this stat favors good players that only play a short time, mainly with breaks of 48hours or more. With decay of stats over time not played, they will always have skill to earn.

high skill/min with min time played filter.
4hrs min. ~1600points/30min
8hrs min. ~800points/30min
Image Nobody expects the Spanish Inquisition!

Return to Server Stats

Who is online

Users browsing this forum: No registered users and 1 guest