D400 (Official Clan Ranking) - Oct 2021 ranking

[Donelladan]

What is the weight and how is it calculated?

It's the sum, for each war, of number of games multiplied by decay factor.

[JPlo64]

Is this the lowest ID has ever been?

Probably, previously their lowest was 922 rating tho the rankings vary slightly so without running the F400 impossible to tell.

Well, the D400 can be run historically. SO one could see if this is their lowest D400 rating ever.
One assumes "lowest rating ever" refers to "lowest monthly rating ever". You can also look at the rating daily. So you could find the specific day a clan had their highest and lowest rating too.

[Donelladan]

Is this the lowest ID has ever been?

Considering only their ranked score ( when they had a weigth higher than 150) then the answer is yes.
You also have to consider their current weight, they aren't officially ranked and shouldn't be.

[Keefie]

All the image links are broken.

[Donelladan]

All the image links are broken.

Looks fine for me.

[Keefie]

Weird, every link was broken earlier, now they're fine.

[rockfist]

Is there a December ranking coming soon?

[Donelladan]

Yes. Sorry for the delay. Expect it in the hour.

[Donelladan]

show

[IcePack]

So you are entering each CL set as its own of 8, and not a combined 16? No more minimum 12 game sets and balanced results of home & away games?
How will that impact RL, are you going to do 6 game sets?

[Donelladan]

So you are entering each CL set as its own of 8, and not a combined 16? No more minimum 12 game sets?
How will that impact RL, are you going to do 6 game sets?

I am not taking care of collecting data, so I actually didn't check that, as I don't have much time at the moment.
You're right, will see about that.

[JPlo64]

So you are entering each CL set as its own of 8, and not a combined 16? No more minimum 12 game sets and balanced results of home & away games?
How will that impact RL, are you going to do 6 game sets?

That's my fault, I forgot to combine them into 16 game sets...
Will have to amend that...

[rockfist]

Thanks Don!

OSA ahead of Fallen and FOED? Wow...not saying OSA is bad but Fallen has beaten us and FOED personally gives me fits.

[benga]

Thanks Don!

OSA ahead of Fallen and FOED? Wow...not saying OSA is bad but Fallen has beaten us and FOED personally gives me fits.

Since we have double the weight, our rank is more accurate than TOFUs.

[rockfist]

Thanks Don!

OSA ahead of Fallen and FOED? Wow...not saying OSA is bad but Fallen has beaten us and FOED personally gives me fits.

Since we have double the weight, our rank is more accurate than TOFUs.

You guys participate in more things than we do that is for sure. I couldn’t keep up with your schedule and I’m a high volume player for TOFU.

[emilywink]

congrats Tofu!

[Arama86n]

Well done to all the Odd Fellows who have worked hard this season. Damn fine to see that #1 once more.

[fishydance]

Well done, Tofu, well done. Big congrats from da fish troops

[mdhill]

I appreciate you guys bringing these rankings back and doing the work to get it done. I found the numbers for the folks without enough weight to be ranked yet that you did in October and November interesting. Hopefully, with all the CLX games, we'll have enough weight to qualify for the main ranking by the rankings that incorporate January, but I still think that's an interesting feature to see where the newer groups are if it's not a lot of additional work.

mdhill

[mdhill]

Early in this thread, there's a good bit of discussion about wanting the ranking algorithm to be kept on the site so it isn't lost. Have you guys considered uploading the calculation to a Google Sheet so everyone can see exactly how it works, replicate it, and figure out where they stand at any time without having to bother you guys about it? This would have the added benefit of ensuring the ranking data and algorithm are permanently available. Not that I'm worried about Jplo or Donelladan disappearing, but this seems to have been a concern of some based on the previous disappearance of the rankings.

mdhill

[Donelladan]

The calculation I can give it to you; but I don't think you can do it without a program, so unless you know how to program you can't do it.
I have been talking to bigWham to see if we could implement it directly to the website.
It is ongoing. We'll see.

Here is the way it works.


So every clan start with a value of 1000. This isn't their score, it's their worth as opponent.

A clan score is the sum for each war of its performance, multiplied by a weight factor, multiplied by a decay factor.
Then divided by the sum of the weight X decay factor
So it is
Clan score = Sum ( performance x weight x decay ) / Sum( weight x decay)
weight for a clan war = number of game for this war.
decay = linear value between 0 and 1.
1 for today and 0 for 2 years.

For a war, performance = score of opponent + alpha*((w-l)/abs(w-l)) + beta*(w-l)/(w+l)
w = games won l = game lost
alpha = 100 if less than 2 wars and 200 otherwise
beta = 200 if less than 2 wars and 400 otherwise

So everytime u calculate the performance of a clan, you need to know the score of the opponent clan at this time.
Which mean you need to recalculate it everytime because of the decay factor that change.


So for example let's assume clan A face clan B.
Clan A has a score of 1400
Clan B has a score of 1600
Clan A win 33 to 27
Clan A performance will be = 1600 + 200 + 400 * (33-27)/60 = 1840
Clan B performance will be = 1400 - 200 + 400*(27-33)/60 = 1160

[mdhill]

Very helpful. Thanks, Donelladan.

[Doc_Brown]

The calculation I can give it to you; but I don't think you can do it without a program, so unless you know how to program you can't do it.
I have been talking to bigWham to see if we could implement it directly to the website.
It is ongoing. We'll see.

Here is the way it works.


So every clan start with a value of 1000. This isn't their score, it's their worth as opponent.

A clan score is the sum for each war of its performance, multiplied by a weight factor, multiplied by a decay factor.
Then divided by the sum of the weight X decay factor
So it is
Clan score = Sum ( performance x weight x decay ) / Sum( weight x decay)
weight for a clan war = number of game for this war.
decay = linear value between 0 and 1.
1 for today and 0 for 2 years.

For a war, performance = score of opponent + alpha*((w-l)/abs(w-l)) + beta*(w-l)/(w+l)
w = games won l = game lost
alpha = 100 if less than 2 wars and 200 otherwise
beta = 200 if less than 2 wars and 400 otherwise

So everytime u calculate the performance of a clan, you need to know the score of the opponent clan at this time.
Which mean you need to recalculate it everytime because of the decay factor that change.


So for example let's assume clan A face clan B.
Clan A has a score of 1400
Clan B has a score of 1600
Clan A win 33 to 27
Clan A performance will be = 1600 + 200 + 400 * (33-27)/60 = 1840
Clan B performance will be = 1400 - 200 + 400*(27-33)/60 = 1160

Don,
That's very helpful. Do you set this up as a linear algebra problem and solve it via matrix inversion?

[Donelladan]

Huh not at all.
I simply make the sum.
I calculate the performance for each war ( from the 1st clan war recorded on CC), and store it in a table. I also store in the same line of the table the date of the war and the weight.
Then everytime I need to calculate the score of a clan at a given time, I calculate the sum for all war of weight X performance X decay factor ( that I divide by the sum of decay X weight).
The "difficult" part is that to calculate a performance for one war, I need to calculate the score of the opponent at the time of the war, since score of a clan is different every day because of the decay.
Score of a clan isn't an information that I can store.

[Fewnix]

Apost I made in our clan forum.

Appreciate all that you have done, are doing and will do.

Thank you

So now you know how one ranking for clans, D400, works, simple huh?



I am not a tech type my understanding is that, as for individual cc members who start at 1000 points and a ? New Recuit ranking unti 5 games are finished, clans start at a ? unknown ranking until some clan wars hae been completed.

Therafter you go up in points and ranking as you win wars, beating a higher ranked clan giving you more points than beating a lower ranked clan, losing to a higher ranked clan cost you less points than losing to a lower ranked clan, e.g. might only win 2 out of 5 clan wars, losing 3, but your points go up

Greater weight is given to reeently completed wars than older wars- so a clan might have had its glory days when it beat top clans regularly and was high ranked but those days are over now. Few wars in the past year, few victories, the points from old wars decay and don;t cost as much as they used to, more weight given to the recent was, the wars lost to lower ranked clans.

OR vice versa

A new reconsituted RGV will not carry the burder of early unsuccessful gamed for long, and as we beat higher ranked clans in recent CL10, the CUP and wars to come, we go up in rank.

something like that

Re: D400 - November ranking
Unread postby Donelladan on Fri Feb 08, 2019 2:38 pm
A post I made in my clan forum
Apost I made in our clan forum

Re: D400 - November ranking
Unread postby Donelladan on Fri Feb 08, 2019 2:38 pm

Huh not at all.
I simply make the sum.
I calculate the performance for each war ( from the 1st clan war recorded on CC), and store it in a table. I also store in the same line of the table the date of the war and the weight.
Then everytime I need to calculate the score of a clan at a given time, I calculate the sum for all war of weight X performance X decay factor ( that I divide by the sum of decay X weight).
The "difficult" part is that to calculate a performance for one war, I need to calculate the score of the opponent at the time of the war, since score of a clan is different every day because of the decay.
Score of a clan isn't an information that I can store.