Algorithm of 400 Clan Ranking (Sep 30, 2011)

[FarangDemon]

Hello Sportsfans,

I have been working together with Jigger1986 and now Gunn217 to maintain a performance-based clan ranking. We plan to update the ranking with the results of new clan wars once a month. I will post the updates to this original post.

Thanks also to jpcloet for providing me the clan war results files he tabulated up to March, 2011.

To find out how the algorithm works, please check out this thread:

viewtopic.php?f=438&t=140929&start=15#p3088157

Data

I'm using data that includes all clan wars, including Clan League sets that consist of relatively fewer challenges.

You can check out the data I'm using with the link below. Please notify me if you find any inaccuracies.

https://docs.google.com/document/d/1Ph3 ... t?hl=en_US

Ranking

Data as of Sept 30, compiled Oct 19th. I'm only including clans with 150 weight points.

Best Showing in 1 Year provides info on the highest ranked clan that each clan has beaten or tied, only counting wars of 40 games or more.

Click image to enlarge.

Unofficial Ranking

This shows all clans that have at least 50 weight points. This is not my official ranking, but I provide it to give these clans an indication of how they're doing.

http://farm7.static.flickr.com/6110/6272905579_e6f0f2daa0_b.jpg

Chart

Here is a chart of performance over the past year:

Click image to enlarge.

Basis of each Clan's Rating

Here is a breakdown showing the ratings and weight from each clan war that were used to determine each clan's overall rating. If you are surprised about a clan's rating, this is where you should look. The rating a clan derives from a clan war is equal to their opponent's score +/- a number from 200-600, depending on the margin of victory/loss. But if one of the clans has completed less than 2 clan wars, I use +/- 100-300.

Click image to enlarge.

Click image to enlarge.

Click image to enlarge.

[Leehar]

The weightings still feel somehow shoddy...

[Army of GOD]

Is there a reason the Spelunkers of Hell aren't on here...?

[FarangDemon]

I just realized that the 3 most recent KoRT games were not counted correctly because of a naming inconsistency in the data file. I'm updating results as we speak.

The weightings still feel somehow shoddy...

Please describe the problem in greater detail if you think there is a way I can improve it.

Note, the weight applied to each challenge is computed as the product of how many games in the challenge TIMES decay factor. I am using 2 years as the data window, so that means that a war from 2 years ago has 0% decay factor (means does not affect score anymore), 1.5 years ago has 25% decay factor, 1 year ago has 50% as decay factor, 6 months ago would be 75% as decay factor, yesterday would be 100% decay factor.

Is there a reason the Spelunkers of Hell aren't on here...?

I had it set to a 150 weight point cutoff, and at just 3 challenges you didn't accumulate enough weight points.

[danryan]

Looks pretty good actually, once the results of the CC are added in a lot of this is going to be turned on its head.

PACK is going to be interesting to watch as they stick out like a sore thumb in that grouping.

[Sniper08]

Looks pretty good actually, once the results of the CC are added in a lot of this is going to be turned on its head.

PACK is going to be interesting to watch as they stick out like a sore thumb in that grouping.

yea im kinda suprised we are ranked so high when i first when through the list i though we had been left out until i looked at the top . No losses so far with big % wins ofc after the NC finals are over we will see were we stand.

[FarangDemon]

Looks pretty good actually, once the results of the CC are added in a lot of this is going to be turned on its head.

Yeah, everybody has been talking about the increasing competitiveness of the top 10 clans as of late, and I think these results support that notion. It is such a very tight race between 4th and 9th place that practically every challenge is going to result in some reshuffling there.

PACK is going to be interesting to watch as they stick out like a sore thumb in that grouping.

yea im kinda suprised we are ranked so high when i first when through the list i though we had been left out until i looked at the top . No losses so far with big % wins ofc after the NC finals are over we will see were we stand.

Yes I was surprised at first, too, but PACK does have a strong record, 5-0 with high win margins. Definitely keep in mind that the highest ranked team (with 150 weight points) that they have beaten so far is BSS at 20th (though SOH is provisionally ranked right behind ID which is 15th on this same list).

[FarangDemon]

Is there a reason the Spelunkers of Hell aren't on here...?

Spelunkers of Hell have 128 weight points at the moment, so you were just below the 150 weight point cutoff. The choice of 150 as the cutoff is somewhat arbitrary but seems to work.

List containing all clans with at least 50 weight points on 06-13-11

Click image to enlarge.

Spelunkers come 18th in this list, right behind ID.

Also note, TFFS at 9th and AFOS at 11th. These clans will soon have enough weight points to be included in my more official listing, and they will make the top 10 even more competitive.

[benga]

What is weight and what is rating?

Explain in math please, I don't need a dictionary.

[Great-Ollie]

This looks very good guys, well done. I think this ranking system may be a little off with some clans, but really overall a very good job. This is a step in the right direction for all clans to become more competitve. As the leader of the PACK i am surprised we have jolted up the standings so fast. I know we are very strong, and i know we could compete with the best clans anytime. It is nice to have a performance based ranking system. I am looking forward to seeing if we can remain in the top ten after a top level clan challenges us. After the NC is done i hope we have earned enough respect to be able to get a top level challenge, only time will tell! Thanks guys, and keep up the great work!

[qwert]

its will be interesting if you can present ranking only to result from 2011,to see if we have some similarity in ranks

[Chuuuuck]

Let me first say I think you have done a great job and I am always happy to see people putting time in to try and make things better instead of just bitching about why other people's ideas are stupid.

In saying that, I will now put my time into bitch about how stupid this is.... just kidding!

I do have a small worry though and am curious for a talking point if there is anyway it can be improved on. Someone like The Pack being up there scares me because if the clan world ever decides to put a lot of weight on your rankings and everyone identifies it as the official pole, then it will change the clan world essentially for the worst (IMO). Instead of there being a TOFU/BotFM match, or a KoRT/THOTA match, then you would just have all of the clans at the top racing to sign up big wars against the 10 worst clans just so they can get staggering lop sided numbers which appears it would improve their rankings here.

No offense to The Pack, I don't really know much about them, but at a quick glance it seems that they have done little more than beaten up new clans that would probably get a brutally lop sided loss to anyone in the top 10-15.

I am also sorry if I just didn't take the time to understand this well enough and my concern has already been taken into consideration and I just didn't see it.

Anyway, like I said at the beginning. Great work and glad to see someone actually trying to do something about it as well.

[FarangDemon]

Let me first say I think you have done a great job and I am always happy to see people putting time in to try and make things better instead of just bitching about why other people's ideas are stupid.

In saying that, I will now put my time into bitch about how stupid this is.... just kidding!

I do have a small worry though and am curious for a talking point if there is anyway it can be improved on. Someone like The Pack being up there scares me because if the clan world ever decides to put a lot of weight on your rankings and everyone identifies it as the official pole, then it will change the clan world essentially for the worst (IMO). Instead of there being a TOFU/BotFM match, or a KoRT/THOTA match, then you would just have all of the clans at the top racing to sign up big wars against the 10 worst clans just so they can get staggering lop sided numbers which appears it would improve their rankings here.

No offense to The Pack, I don't really know much about them, but at a quick glance it seems that they have done little more than beaten up new clans that would probably get a brutally lop sided loss to anyone in the top 10-15.

I am also sorry if I just didn't take the time to understand this well enough and my concern has already been taken into consideration and I just didn't see it.

Anyway, like I said at the beginning. Great work and glad to see someone actually trying to do something about it as well.

Thanks for your kind words and support. You raise a very valid concern.

I think the best way to solve the problem would be to leave the ratings formula unchanged, yet extend the definition of 'provisional' to include those clans that have not beaten their peers. I prefer to try this way rather than to alter the ratings formula because as a side effect from altering the ratings formula to disincentivize the higher-ranked clans from playing the lower, the higher-ranked could get penalized too much for playing the lower-ranked clans that they might be forced to play in various leagues.

So we need a rule to determine whether or not a clan has "beaten its peers". I think this is much easier to define with a rule than solving the problem I stated in the above paragraph. A simple rule could be:

A clan's rating is provisional if they have not beaten any clan 3 places below their current standing or better within the past 6 months.

So if The PACK is now ranked #7, this would be considered provisional if, within the past 6 months, they had not beaten any clan ranked #10 or better (rank of enemy clan at time of clan war).

I'd be interested to hear feedback on this proposed method of determining provisional/official status. How recent should the war be, how many peers should they have played, how close do peers need to be?

For example, an alternative formula could be of the form:

A clan's rating is provisional if, within the past 6 months, they have neither beaten a clan the same rank as themselves NOR beaten two clans ranked within 3 ranks of themselves

This rule would mean that The PACK's score would be official if they either beat one team ranked #7 or above OR two teams ranked #10 or above.

[Frederik Hendrik]

Beautiful!

I have tried to create a ranking system in somewhat the same way last year, but wrestling with formula's, a failing database and RL business caused me to quit. I recognise the problems in the system, but also the great oportunities of it to act as a well working ranking.

One point of advice you might want to try. I found out that intergrating the clan league results helped the ranking greatly. The matches theirselves have very small weight, but the total games played in the league are huge. And the beauty is that so many clans play eachother, that it really smoothes out the ranking perfectly.

Clans (often new ones) that have gathered many points beating low-ranking clans remain the biggest problem in this system. A logic solution would be to never let a clan get to a higher position than the highest-ranked clan they have ever beaten (within the decay-time), regardless of their points. To add "provisional" to the ranking of a clan makes its "real" position unclear and to refuse a clan a higher position than just above the best clan they have ever beaten is both logical and makes them want to play the higher ranking clans.

I want to repeat that your work on this is great!

FH

[FarangDemon]

One point of advice you might want to try. I found out that intergrating the clan league results helped the ranking greatly. The matches theirselves have very small weight, but the total games played in the league are huge. And the beauty is that so many clans play eachother, that it really smoothes out the ranking perfectly.

These results do include all three seasons of Clan League. I never did state what results I was using, but I intend to post the data file to a URL so everybody can see for themselves, and use as a basis for their own ranking systems, to save them time tabulating results.

Your other advice regarding the issue Chuuuuck brought up, I will think about more and comment on later.

Thanks for your support and constructive feedback as well.