Algorithm of 400 Clan Ranking Algorithm

[Leehar]

I didn't say you were wrong in your opinion on "each clan against score of the division" method. I haven't said your point didn't have sense ether. Please read my post more careful next time as it will save breath for both of us .

Then I'll say it.
I vehemently disagree Josko, specially with the Div A/B weaker than C/D. It's almost like saying in the Cricket World Cup Group A was overall better than Group B because they had 3 semifinalists. After all, over there the 2nd in group B made the Final, while the 1st lost in the quarters. It all just depends on who performed better on the day with the conditions, and I think that analogy holds some truth over here as well.

Again, I'm not sure why I still have an issue with the rankings, but something still feels a bit off

[josko.ri]

@ Leehar

do you think that all Divisions were equally stronger/weaker or some Divisions were stronger and other ones were weaker?

[danryan]

Division C was clearly the strongest because as awesome as FOED are we didn't even make the playoffs. Quod erat demonstratum.

[Dako]

@ Leehar

do you think that all Divisions were equally stronger/weaker or some Divisions were stronger and other ones were weaker?

What does equally stronger mean? I reread your sentence like 3 times and couldn't grasp what you were trying to say.

[josko.ri]

@ Leehar

do you think that all Divisions were equally stronger/weaker or some Divisions were stronger and other ones were weaker?

What does equally stronger mean? I reread your sentence like 3 times and couldn't grasp what you were trying to say.

equally strong (or) weak. means that noone Division were stronger than any onther Division. true or false?

[Rodion]

Imagine division's strenghts could be measured from 0 (weakest possible division) and 10.000 (strongest possible division).

Think of division A: what number would you give it?
Think of division B: what number would you give it?
Think of division C: what number would you give it?
Think of division D: what number would you give it?

If you didn't give everyone the same exact number, then divisions were not all equally strong in your opinion.
In my opinion, they were not all equal, but I'm not going to say which ones I considered stronger.

If you want to rank clans against divisions, you need to create several "Frankensteins" for it to work.

TSM vs. rest of Div D - rest of Div D is KORT + L4D + G1 + GR + TLW
G1 vs. rest of DiV D - rest of Div D is KORT + L4D + TSM + GR + TLW

As you can see, both of them faced "rest of div. D", but they were differents "rests", thus each should have its own division rating. In the end, we'd be comparing 24 clans vs. 24 different "rest of divisions" and giving it a weight factor of 90. I'm not sure if that would be different from comparing each clan vs. the other 5 clans with weight factors of 18 each, however.

[josko.ri]

I wanted to say, that no matter in which Division someone played, it is important which clans they played with. as this system of calculation validate all historical data of all clans, then CL3 results should be validated according to clans faced and their ranking in moment when match was done, and it would be very unfair to just give 55 (wins achieved) vs 45 (average wins in Division) because some Divisions are weaker and some are stronger. numbers and rankings(from previous wars) of clans from particular Division will show which clans were weaker/stronger and therefore validate performance of a clan according to strenght of opponent's clan, not according to average number of wins in a Division.

[jpcloet]

If you want to rank clans against divisions, you need to create several "Frankensteins" for it to work.

Nice term, but yes that is what I had to do before.

[Dako]

equally strong (or) weak. means that noone Division were stronger than any onther Division. true or false?

Equally strong == equal (to me at least). I was just confused of you adding "stronger" and "weaker" because those adjectives presume you are comparing them to something else. Just my misunderstanding.

[Leehar]

I wanted to say, that no matter in which Division someone played, it is important which clans they played with. as this system of calculation validate all historical data of all clans, then CL3 results should be validated according to clans faced and their ranking in moment when match was done, and it would be very unfair to just give 55 (wins achieved) vs 45 (average wins in Division) because some Divisions are weaker and some are stronger. numbers and rankings(from previous wars) of clans from particular Division will show which clans were weaker/stronger and therefore validate performance of a clan according to strenght of opponent's clan, not according to average number of wins in a Division.

But that was pretty much what was roughly done when Jpc drafted the divisions. I don't remember his process exactly but he did seed all the clans and divide them equally along the divisions. You're pretty much commenting on strengths after the fact, but that does wring somewhat hollow. Again there also all the other incremental factors like how much 18 game sets can truly count etc

[josko.ri]

I wanted to say, that no matter in which Division someone played, it is important which clans they played with. as this system of calculation validate all historical data of all clans, then CL3 results should be validated according to clans faced and their ranking in moment when match was done, and it would be very unfair to just give 55 (wins achieved) vs 45 (average wins in Division) because some Divisions are weaker and some are stronger. numbers and rankings(from previous wars) of clans from particular Division will show which clans were weaker/stronger and therefore validate performance of a clan according to strenght of opponent's clan, not according to average number of wins in a Division.

But that was pretty much what was roughly done when Jpc drafted the divisions. I don't remember his process exactly but he did seed all the clans and divide them equally along the divisions. You're pretty much commenting on strengths after the fact, but that does wring somewhat hollow. Again there also all the other incremental factors like how much 18 game sets can truly count etc

it has decay factor, what is (if I realized good) parameter of2 variables (time passed and strength of opponent), and I totally support that system.

for small number of games, it has Weight factor, so # of games played, be it 18 or 60, is inluded in calculation like it should be.

[FarangDemon]

I'd prefer to discuss the algorithm itself in this thread and have a second thread devoted to maintaining an up-to-date ranking using this system.

The algorithm itself

I'm happy to discuss the method, and maintain and improve my script but I don't want to maintain the Clan War data that it runs on. And I don't want people complaining to me that I did or did not use such and such data. CC isn't comping me in any way for this, so I will only do what I think is fun. I've changed the name of this thread to "Algorithm" to reflect this.

An official ranking based on the algorithm

This would be the thread that discusses data and contains updated ranking lists / graphs.

You'll need someone to maintain the clan war data file and I can link to it from my script page. Then whoever wants to generate an up-to-date ranking can just click on the link, copy and paste the data into my script and churn it out whenever. Or if I improve my script, I can eliminate the need to open the file and manually copy and paste.

This is the format for each line (Clan1, Clan2, Clan1 Wins, Clan2 Wins, Date) separated by tabs.

http://www.killersapp.com/AllData.txt

The file below has the same war results, but each war has a description of whether it was Clan League 1,2,3 or a regular clan war. This should help to refine dates. If you want accurate results, somebody is going to have to look up when each clan war from Season 1 and 2 actually ended (I'd choose 3 weeks after the 2nd and final Home/Away set between the two clans). Dates from Season 3 are pretty good already.

http://www.killersapp.com/AllDataMerged.xls

[FarangDemon]

Ok looks like there is no interest at the moment to use the algorithm to create an official ranking.

[qwert]

lol, so all these job,its that someone else work ranking instead you.
Look on mine elo ranking, i doing all job, and these is how you also need to do.
Jpcloet all ready have official ranking.

[FarangDemon]

lol, so all these job,its that someone else work ranking instead you.
Look on mine elo ranking, i doing all job, and these is how you also need to do.
Jpcloet all ready have official ranking.

All I want is somebody else to maintain a data set at a fixed URL that we ALL can use (in the format I can directly use, or in a different format that I could write a script to parse). If there are 10 people with different ranking systems, it's dumb for each person to have to maintain the data from the exact same games separately. Very inefficient.

I'm happy to use the algorithm to post updated rankings now and then, though I've published the website, making it possible for anybody to do this whenever they want. All you need is your own data to paste in. All calculations are automated using user-defined parameters.