Clan Rankings - RPI Version [Archived]

[jpcloet]

The following ranking systems uses a variation of the College RPI System.

This ladder system was created after test running an ELO ladder which had a hard time dealing with a lack of data, and mid-level clans winning significantly over newer clans which led to some skewed results.

How this Version Works:

• Once a clan challenge is completed, it is added to the database (Excel File)
• Using some complex calculations and look-ups, it calculates out Win Percentage (WP), Opponent's Win Percentage (OWP) and Opponent's Opponent's Win Percentage (OOWP), and the Reliability Factor (RF). (See below for Factor rates)
• An RPI is generated for each team, generally between 1.00 and 2.00 using RPI = [(WP + OWP + OOWP) * RF]
• Based on all the RPI's, they are all scaled up to a Grading out of 100 with the top Clan at a 100 Rating

Reliability Factors:

• 0.00 if your clan is extinct
• 0.80 for 1 completed challenge
• 0.85 for 2 challenges
• 0.90 for 3 completed challenges and less than 61 total games
• 1.00 for 3 completed challenges and more than 60 total games

[Scott-Land]

Would it be too difficult to post along with the RPI's win percentages and overall records-- JP?

[jpcloet]

Would it be too difficult to post along with the RPI's win percentages and overall records-- JP?

I can do the win % easilly, overall record will be a bit of work.

Had several suggestions on new colours via pm as well.

[Blitzaholic]

nice work jp

[jackin_u_up]

Just Curious and maybe i missed it.... but are the results of our divisional 20 game sets going to actually go down on Record as clan war results ?

I understand they are being factored into the clan ladder RPI ratings but i looked in a couple threads and couldn't find this being discussed.

IMO i think they should go down on record as a clan war. anyone else

[reptile]

Just Curious and maybe i missed it.... but are the results of our divisional 20 game sets going to actually go down on Record as clan war results ?

I understand they are being factored into the clan ladder RPI ratings but i looked in a couple threads and couldn't find this being discussed.

IMO i think they should go down on record as a clan war. anyone else

it was my understanding they were going to. but yes i agree with you.

[jpcloet]

Check the new ladder, all 20 game set will be added. They are actually all set up in the file and actually linked to the main league file, so there are some partial results in there right now.

[jackin_u_up]

Check the new ladder, all 20 game set will be added. They are actually all set up in the file and actually linked to the main league file, so there are some partial results in there right now.

What do u mean by partial results jp ?

[jpcloet]

That means if I've updated the league file and only 7 games of a second series is complete, then the results of say 9-8 (3 games not comlete) are in the database.

[Blitzaholic]

Just Curious and maybe i missed it.... but are the results of our divisional 20 game sets going to actually go down on Record as clan war results ?

I understand they are being factored into the clan ladder RPI ratings but i looked in a couple threads and couldn't find this being discussed.

IMO i think they should go down on record as a clan war. anyone else

jp said he was going to count it if they are 20 games completed vs the same clan


however, I disagree and do not think this should be counted as 20 games, as 20 games is too small of a sample and I suggest none should be counted unless it is at least 40 games or more against one clan, perhaps we should put this to a vote. With 40 games or more, more games the better tends to help measure skill a little more and evens out the luck drops and dice factors, after all, we all would like this clan ladder rating system to reflect skill more than luck, I personally would like it 60 to 100 games, but, to be fair, a minimum of 40 is a good measurement.



respectfully, blitz

[hwhrhett]

Just Curious and maybe i missed it.... but are the results of our divisional 20 game sets going to actually go down on Record as clan war results ?

I understand they are being factored into the clan ladder RPI ratings but i looked in a couple threads and couldn't find this being discussed.

IMO i think they should go down on record as a clan war. anyone else

jp said he was going to count it if they are 20 games completed vs the same clan


however, I disagree and do not think this should be counted as 20 games, as 20 games is too small of a sample and I suggest none should be counted unless it is at least 40 games or more against one clan, perhaps we should put this to a vote. With 40 games or more, more games the better tends to help measure skill a little more and evens out the luck drops and dice factors, after all, we all would like this clan ladder rating system to reflect skill more than luck, I personally would like it 60 to 100 games, but, to be fair, a minimum of 40 is a good measurement.



respectfully, blitz

you really should have been a part of the discussion about this 6 months ago....

but for the record, i still think it is good that they will count towards clan ladder.

[Blitzaholic]

hwhrhett, I did not have access to that hidden forum where you all discussed it, I believe only david wain and andrewb did from our clan.

[jpcloet]

20 used to be the most common size although challenges are growing, more often that not the 20 game sets are for new clans and clans attempting to enter the CLA. The average size is about 38 now. You will notice that the RPI factor is unfavorable for clans who don't play many challenges or play many games. A clan with 3 challenge results, but only 60 games (3 * 20) is still not given a factor of 1.00.

Ignoring the league, here are the stats (round to the nearest 10)

Size----Count
100---- 1
60---- 5
50---- 3
40---- 23
30---- 7
20---- 13

The handbook does talk about both sizes. However, maybe we should include 40 games as being "The recommended size for determining skill"

This is actually off-topic now, I will be moving this to the ladder discussion.

[jpcloet]

Moved to ladder discussion.

[Blitzaholic]

20 used to be the most common size although challenges are growing, more often that not the 20 game sets are for new clans and clans attempting to enter the CLA. The average size is about 38 now. You will notice that the RPI factor is unfavorable for clans who don't play many challenges or play many games. A clan with 3 challenge results, but only 60 games (3 * 20) is still not given a factor of 1.00.

Ignoring the league, here are the stats (round to the nearest 10)

Size----Count
100---- 1
60---- 5
50---- 3
40---- 23
30---- 7
20---- 13

The handbook does talk about both sizes. However, maybe we should include 40 games as being "The recommended size for determining skill"

This is actually off-topic now, I will be moving this to the ladder discussion.

yes, agreed JP, thank you considering this: maybe we should include 40 games as being "The recommended size for determining skill" because all must admit 40 games is much better than 20 and as you said JP "A clan with 3 challenge results, but only 60 games (3 * 20) is still not given a factor of 1.00." and the clan ladder of counting 20 games hinders clans who only played a few challenges, the clan league should be different and seperate than clan challenges unless all play against same clan at least 40 games. again, ty jp for be open about this.

[jpcloet]

As a Clan Director, I'm trying to use the public forums more for debated topics and ideas. The CLA has some key goals and voting rights so it will remain somewhat quiet and a key place for some leadership agreements.

[hwhrhett]

hwhrhett, I did not have access to that hidden forum where you all discussed it, I believe only david wain and andrewb did from our clan.

sorry, i thought it had been a public discussion..

for the record, i was for only counting challenges 40+ in the ladder, however, since we came to the conclusion that was too big for some of the smaller clans at the time. it seemed obvious to me that if we were going to be playing 20 games against a clan and a 20-game challenge counts towards the ladder, then there would be no reason not to count it towards the ladder.

p.s. u should have one of those guys give up his place in cla so you can come in, as your typically alot more vocal in the forums....

[Seulessliathan]

maybe that forum can be opened so everybody can read it, but only members can write? atm it feels like a secret agency, and my clan already mentioned that they are not sure if they want to join the new league. I will try to push them to join the cla group ... again^^

[jpcloet]

maybe that forum can be opened so everybody can read it, but only members can write? atm it feels like a secret agency, and my clan already mentioned that they are not sure if they want to join the new league. I will try to push them to join the cla group ... again^^

Please pm me Seul. You've had access and seen what's in the CLA and the content there. I'm not sure why your clan feels like it's a secret society because it isn't. Maybe you need to send someone to the CLA who will report back to the clan more clearly. The 3 entrance rules should be easy to meet so it's never been an exclusive group. I'd also like to know why your clan is uncertain about the next league, if you have issues with the setup etc. please pm me your concerns.

[Lubawski]

I like how this works, but I don't think the CLA results should be used, at least not this time around. It only consisted of doubles games for one, and because, as I understood it, it would not count toward the ladder, I sent out 18-20 different people every week and experimented with teams and maps. We took it as a practice battle ground, not the real thing. It's like deciding after the fact that you are going to count a pre-test as the actual test. Not everyone prepared for it adequately. I do like the RPI version though. I think it will yield better and more accurate results.

[hwhrhett]

I like how this works, but I don't think the CLA results should be used, at least not this time around. It only consisted of doubles games for one, and because, as I understood it, it would not count toward the ladder, I sent out 18-20 different people every week and experimented with teams and maps. We took it as a practice battle ground, not the real thing. It's like deciding after the fact that you are going to count a pre-test as the actual test. Not everyone prepared for it adequately. I do like the RPI version though. I think it will yield better and more accurate results.

if we had not decided 6 months ago that they would count, then i would have done the same thing....

[Seulessliathan]

i think most of my clan don´t like 1v1 in a league and we don´t know if we would have to play them or not. I think Scott-Land is ok to join cla group, maybe that makes it easier.

[jpcloet]

Well we are on the same page, so don't jump to any conclusions. I'm not a fan of 1v1 for the league, let's leave those to clan wars and tournaments. I think only one person has suggested it, but I'm not a fan.

[Seulessliathan]

looks like Scott-Land and maybe a second of my clanmates will join the group , will be the easiest way to see what´s going on.

[hwhrhett]

Well we are on the same page, so don't jump to any conclusions. I'm not a fan of 1v1 for the league, let's leave those to clan wars and tournaments. I think only one person has suggested it, but I'm not a fan.

yeah, 1v1's are just too easily suaded by poor luck or good luck... no workey.....