Algorithm of 400 Clan Ranking (Sep 30, 2011)

[FarangDemon]

July Update

Jigger1986 has sent me an update for all wars concluding in June. I've ran my script and updated the OP with the new results. I didn't feel like updating the ratings basis or accuracy tables at this time - will do so next update, as I think there will be more upheaval then as well.

Btw, I am still using the same formula as last month. I am considering experimenting with weighting peer-to-peer challenges more than lopsided challenges, but this will wait until I have more free time.

Big Eight

I notice that, based on this data, there is a top 8 cluster of clans that have over 1200 points - in order, TOFU, KoRT, THOTA, EMP, AOC, IA, TSM, LOW. Their scores cluster together (dropoff after 8th place) as well as their best recent performance - these Big Eight are pretty much the only clans that have been able to beat another Big Eight clan within the past year.

The exact details of this top cluster will change after certain wars finish, but it will be interesting to see how the results cluster after the next update.

Data

Feel free to check out the data file I'm using and please point out any error if you find one.

https://docs.google.com/document/d/1luEvDTpI-Hz3dMWTBmLpchvolmzBu1UBxcXm6QdVkec/edit?hl=en_US

[Dako]

I am not sure that "beat xx in 2010" should count league results. Yes, 11 games confrontation is a very small war, but I think it should not be counted as best showing.

[Commander9]

I notice that, based on this data, there is a top 8 cluster of clans that have over 1200 points - in order, TOFU, KoRT, THOTA, EMP, AOC, IA, TSM, LOW. Their scores cluster together (dropoff after 8th place) as well as their best recent performance - these Big Eight are pretty much the only clans that have been able to beat another Big Eight clan within the past year.

Just out of curiosity - so FOED versus IA doesn't count as a win against Top 8? And it's not our best showing "ranking" wise?

[FarangDemon]

I am not sure that "beat xx in 2010" should count league results. Yes, 11 games confrontation is a very small war, but I think it should not be counted as best showing.

I definitely agree with you - I could set some kind of minimum games requirement for this. What do you think, 41 games?

Just out of curiosity - so FOED versus IA doesn't count as a win against Top 8? And it's not our best showing "ranking" wise?

Sorry Commander9, the data jigger1986 sent me did not include the FOED vs IA war. I read the thread now and see that, according to Master Fenrir it has been officially over for 1 week at 29-12, though the topic does not say concluded. Maybe it didn't make it to the forum that he checks for clan war results.

Thanks for pointing this out. Unfortunately we don't have a clan war results archive, so we are going to continue to make errors like this until one is provided.

CC could add the capability to tag Clan War games upon their creation as being "Clan war KoRT vs AOC CC2 Quarterfinals" so that CC could automatically tabulate the results and post the results of completed wars to a Clan War Result Database. That would save everybody involved tons of time and produce more accurate reports in a more timely fashion. Has anyone suggested something like this before? Would CC be interested in using its development capabilities for the Clan Community? Maybe CC is already doing the same thing for tournaments?? Same idea could be applied to both.

In the meantime, anybody, feel free to look at our data file and let me know if any other wars are missing or if we have incorrect results.

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

I want to wait several days for any other possible discrepancies to be pointed out before going through the trouble of redoing everything.

[danryan]

I wouldn't blame anyone for not wanting to include that one.

[Bruceswar]

TOFU vs BFM was also missed. Or maybe not depending on when you put together the info.

[FarangDemon]

TOFU vs BFM was also missed. Or maybe not depending on when you put together the info.

Was it over before July 1? If so please tell me score.

Ok I see over June 30 at 22-19 (TOFU). So I'll add it along with the FOED - IA war.


I will redo the rankings soon...
The squeaky wheels get the grease, people, so if we left out a war that ended in June, start squeaking or forever hold your peace for this update.

[Leehar]

A simple check would be to see if you inputted all the Newcomers and Conqueror Cup ties that have completed.

I'm curious however on how you came up with the Best score in the last year if that one isn't actually better than the rating you get after a match?

[Dako]

40 games or more for the best showing I think.

[FarangDemon]

A simple check would be to see if you inputted all the Newcomers and Conqueror Cup ties that have completed.

I prefer not to check threads myself, that's why I delayed putting out an update until June when I found a volunteer to help me - jigger1986 helps me by checking the threads and tabulating the results every month. I contribute in my way, he contributes in his. If there are wars we have missed, I ask the clan community to contribute by pointing them out to us. If you want to contribute you could check the threads and compare them to the data file, see if anything is missing - that would be awesome.

But perhaps the best way the clan community could contribute toward the goal of obtaining accurate clan war results would be to push for the creation of a Clan War Results Database. Then nobody has to waste the time checking all the forums and tabulating results, no mistakes or omissions would be made, and members of the clan community would not have to waste their time posting about the absence of certain wars.

I'm curious however on how you came up with the Best score in the last year if that one isn't actually better than the rating you get after a match?

The score displayed in the "Best Showing in 1 Year" column is not the score of the clan itself, but that of the highest ranked opponent that the clan has beaten or tied to within the last year. It is this opponent's score at the time of the match's conclusion (before points changing hands from the match).

[Lubawski]

All in all I think Farang has done some really good work here. I also applaud his attitude when people jump into the thread with pitchforks and torches because something was missed.

[FarangDemon]

All in all I think Farang has done some really good work here. I also applaud his attitude when people jump into the thread with pitchforks and torches because something was missed.

Thanks very much, Lubawski.

I've updated the rankings in the OP to include the two omitted wars.

[IcePack]

GR (21) vs. KOA (27) of 48 [Final]
by jigger1986 on Sun Jun 26, 2011 9:41 am

This war is now officially over. Congratulations again to KOA for the win.

Cheme moved it to completed area July 2nd, tho it ended before July 1st.

27 KOA 21 GR

[jigger1986]

Just to clear a couple things up, I noticed several posts regarding missing data.

Only wars/challenges that were finished, moved and notated by a CD on or before 06/30 have been counted.
IA vs FOED still has active games (or at the very least still needs a thread update) and TOFU vs Botm finished on 06/30, but was not notated and moved by the CDs.

These results will be given to FarangDemon for the July calculations.

By the way, I love the "greatest accomplishment" idea, but would be in favor of a game limit of 41 games.
(eventhough my low ranked clan tied a very high ranked clan in a CL3 matchup and looks nice on paper to see that)

Thanks

-jig

[FarangDemon]

I meant to make it be > 39 games limit but there was a mistake in my code:

Code: Select all

if (outcomeString == "TIE" || outcomeString == "WIN" && RatingRecords[m].Games > 39)

I will fix it to:

Code: Select all

if ((outcomeString == "TIE" || outcomeString == "WIN") && RatingRecords[m].Games > 39)

[Dako]

Instead of doing 2 check for TIE or WIN why don't you do one check for != LOSS? Saves some processing ;).

[Leehar]

Lol, you missed a bracket, thats one of the most irritating things ever.

And with regards to the win/tie sentiments, could there ever be a situation where a loss could give you a higher rating then your wins? If so, then I think that could also be worth reporting...
(I'm actually surprised our 'loss' to thota didn't net us more points, since I think that was one of our best performances compared to a win over DB...)

[Dako]

Can you post top-5 ratings achieved during any moment? So it looks like this:

Best score:
#1 THOTA - 1375 - 2010 Oct
#2 TOFU - 1372 - 2011 Jun
#3 KORT - 1339 - 2011 Jun
...

Would be nice to see clan records there and it is not easy to calculate them.

[Lubawski]

Hey Farang, I don't think you have our most recent challenge against Nemesis in there from the Cup. Final score 30-11 win for EMPIRE.

[jigger1986]

Lub,

I can assure you that the EMP vs NEM war of June 2011 is in there, however because those two clans also had a war back in December 2009, the games,win/loss and % are likely being lumped into one large calculation. I am not the expert, FD is, but I thought I could offer a possible explanation. I do know however that 100% that war IS being calculated.

[Leehar]

Lub,

I can assure you that the EMP vs NEM war of June 2011 is in there, however because those two clans also had a war back in December 2009, the games,win/loss and % are likely being lumped into one large calculation. I am not the expert, FD is, but I thought I could offer a possible explanation. I do know however that 100% that war IS being calculated.

It is there, just the basis points for each clans rating haven't been updated so you can't see how it counts.

It's the reason why they jumped from 1255 in 6th to 1279 in 4th and why we went from 1124 to 1100...

[FarangDemon]

Can you post top-5 ratings achieved during any moment? So it looks like this:

Best score:
#1 THOTA - 1375 - 2010 Oct
#2 TOFU - 1372 - 2011 Jun
#3 KORT - 1339 - 2011 Jun
...

Would be nice to see clan records there and it is not easy to calculate them.

I can do that but I'm not exactly sure what it would represent. What do you see it as representing?

I think number of clans at time X vs time Y could be a factor that might distort whatever this ranking is supposed to represent. It definitely could not be used to say Clan A can beat Clan B now - and I'm not sure it would be a reliable indicator that Clan A in their heyday could beat Clan B in their heyday ("heyday" meaning the time that the clan performed its best) due to this possible distortion caused by total number of clans at any given time.

How does total number of clans affect the highest rating achievable at that time? If there are only two clans, and the top clan beats the other clan at 75% win margin, the two clans would come to an equilibrium of 1200 and 800. This would happen no matter how good the two clans are, as long as the better clan always wins at 75% win rate. Who knows, maybe once there are 5 clans, this is not so much of an issue, but someone would need to examine this before drawing any conclusions.

And there is another factor - how good were the clans at time X compared to time Y? If the clans were weaker at time Y than they were at time X, then a higher rating over them might not be as meaningful as a slightly lower rating at time X.

Anyway, I will consider providing it just for fun, at least. But if it is proven to be meaningless due to the number of clans at time X factor, then I probably don't want to provide it.

Lub,

I can assure you that the EMP vs NEM war of June 2011 is in there, however because those two clans also had a war back in December 2009, the games,win/loss and % are likely being lumped into one large calculation. I am not the expert, FD is, but I thought I could offer a possible explanation. I do know however that 100% that war IS being calculated.

It is there, just the basis points for each clans rating haven't been updated so you can't see how it counts.

It's the reason why they jumped from 1255 in 6th to 1279 in 4th and why we went from 1124 to 1100...

If a clan plays another clan several times, each war is considered a separate event.
I will update the ratings basis shortly so you will see the impact of each war on your ratings.

And with regards to the win/tie sentiments, could there ever be a situation where a loss could give you a higher rating then your wins? If so, then I think that could also be worth reporting...
(I'm actually surprised our 'loss' to thota didn't net us more points, since I think that was one of our best performances compared to a win over DB...)

It is possible for a loss to give a clan a higher rating than a win.

The settings of Min delta per win/loss and Max delta per win/loss affect this. I have chosen 200 to be the minimum points awarded on top of an opponent's rating to a clan that beats their opponent. I chose 600 for maximum. So the actual value added/subtracted is between 200 and 600 and is linearly correlated to margin of victory.

So even if that NEM barely lost to THOTA, NEM's basis point for the match is THOTA's rating at that time minus 200.

If I used a min value of 0, then NEM would get a basis point near equal to THOTA's rating as THOTA's margin of victory was small.

But I found that using the value of 0 resulted in lower accuracy than using the values I am using.

[Dako]

It will not provide anything - just the numbers for the hall of fame, aka another blitz top list.

[FarangDemon]

It will not provide anything - just the numbers for the hall of fame, aka another blitz top list.

Ok I'll try to work it into the next update.
Btw, thanks for your advice. I implemented the != "LOSS" in my code.

[JoshyBoy]

As much as I hate to admit it, these rankings are more representative of where Nemesis is right now. We need to get our act together and be better than we have been.