F400 Clan Ranking - Needs Repair

[chemefreak]

That shouldn't be right Ice. We had to win something like 8 wars straight to get in the top 10. And one of those where in the top 10.

I just got home (worked some massive OT today) so I'll start checking some stats.

IcePack

Ice, no hurry man. There is nothing starting soon that will rely on the F400. Plus, we don't want you to burn out too fast as the "Keeper of the Ranks"

[IcePack]

That shouldn't be right Ice. We had to win something like 8 wars straight to get in the top 10. And one of those where in the top 10.

I just got home (worked some massive OT today) so I'll start checking some stats.

IcePack

Ice, no hurry man. There is nothing starting soon that will rely on the F400. Plus, we don't want you to burn out too fast as the "Keeper of the Ranks"

I don't think i'll burn out, this kinda stuff really interests me I only wish i could do more!
IcePack

[MudPuppy]

I don't think i'll burn out, this kinda stuff really interests me I only wish i could do more!
IcePack

Awesome attitude, Ice. Thanks for taking on yet another project to improve the CC/clan world.

[IcePack]

Notes re: OTP

Started before 02/24/12 w/ 1000 pts. No decay until "earned" points thru wars.

OTP vs AKA 28-13 02/24/12 decayed approx 20%
Earned 1012 PointsTotal @ 41 "Weight" (AKA rated 850) OTP Tied 18th
OTP vs BOTFM 8-4 03/28/12 decayed approx 16%
Earned 1038 Points Total @ 51 "Weight" (BOTFM rated 1073) OTP Ranked 18th
PACK vs OTP 7-5 03/28/12 decayed approx 16%
Earned 1056 Points Total @ 63 "Weight" (PACK rated 1242) OTP Ranked 16th
DYN vs OTP 7-5 03/28/12 decayed approx 16%
Earned 1011 Points Total @ 75 "Weight" (DYN rated 1051) OTP Ranked 20th
OTP vs RA 8-4 03/28/12 decayed approx 16%
Earned 1035 Points Total @ 87 "Weight" (RA rated 847) OTP Ranked 19th
OTP vs TNC 39-22 04/08/12 decayed approx 12%
Earned 1138 Points Total @ 148 "Weight" (TNC rated 974) OTP Ranked 10th
OTP vs TFFS 24-17 06/01/12 decayed approx 9%
Earned 1157 Points Total @ 177 "Weight" (TFFS rated 946) OTP Ranked 10th

Please keep in mind I had to lower one factor so they would "show" on rankings so early as their first war, the way its built they need 150 weight to be factored in (because it can skew results a bit without the weight) and they dont reach 150 weight until TNC war.

Essentially, since they've had so many wars factored in, in a short time period they are able to gain some points faster than other clans. Also, they / TNC were fairly "equal" when they faced each other (ranking wise) and so when they scored such a high win rate vs a "equal" clan it boosts them. Just like PACK war, they should have been "severely" out ranked by PACK but kept it close, so it actually gave them a (small) bump in the ratings. (this is why CL wars are less weighted I believe) Tho i am a bit surprised at how large a jump based on a single war but I believe its based on the algo expectations for war score were greatly exceeded.

One thing cheme to keep in mind (please take this as constructive criticism, if i may) but adding large amounts of clan results on the same day can actually have an effect on clan scores as the date and order of challenges in database does have some effect within rankings. So if we could more accurately submit them on day of completion (and avoid doing them all at once on same day...or if you do that, perhaps moving to completed but date them when the challenges finish) so the challenges are entered as completed. Its a small difference, but it does make a difference.

But, I redid the database until reaching each one of OTP matches and scored based on the end of those dates each time. All appears to be working the way it should. I will keep looking into it a bit more...

IcePack

[benga]

Can you explain then how KORT got 10 pts w/o any new results coming in for them
and OSA got only 31 for winning 2 wars??

This was meant as a change from last month.

[MichelSableheart]

So basically it's big wins against middle clans, small losses against strong clans, all relatively recent to guarantee that the results haven't decayed yet which gives them a strong score?

@Benga: I would guess (without looking at the actual data) that KORT gaining points without new result comes from a severe loss decaying.

[Qwert]

nothing strange here, this list are update once per month, that why people have surprised how some clans jump in table.

[Leehar]

Can you explain then how KORT got 10 pts w/o any new results coming in for them
and OSA got only 31 for winning 2 wars??

This was meant as a change from last month.

clan league?

[Chariot of Fire]

Can you explain then how KORT got 10 pts w/o any new results coming in for them
and OSA got only 31 for winning 2 wars??

This was meant as a change from last month.

clan league?

That's what I'd have put it down to, but there are no June results for KORT in the list under the OP.

Brownie points?

[IcePack]

@ benga...leaving for work, but happy to look into this as well. Although the previous explanation seems to fit my initial thoughts

[chemefreak]

One thing cheme to keep in mind (please take this as constructive criticism, if i may) but adding large amounts of clan results on the same day can actually have an effect on clan scores as the date and order of challenges in database does have some effect within rankings. So if we could more accurately submit them on day of completion (and avoid doing them all at once on same day...or if you do that, perhaps moving to completed but date them when the challenges finish) so the challenges are entered as completed. Its a small difference, but it does make a difference.

It doesn't matter when I add them. The date we provide has the end date of the war. So if I do 5 wars today, I list the completion date separately for numbering purposes. Please feel free to update your data accordingly. Here is the link:

https://docs.google.com/spreadsheet/ccc?key=0Atustsi2x0-2dDR6TVo4YnhQcWZzeUs0YXNlMHZrd3c&hl=en_US#gid=0

The end date is the date the last game of that war was completed.

[IcePack]

I don't see any CL results there? The reason I mentioned that is there were quite a glut of same day entries in the info I was provided which effects the rankings.

Examples: 3/28/12 as 4/1/12

I'll take my info and compare to that link.

IcePack

[chemefreak]

I don't see any CL results there? The reason I mentioned that is there were quite a glut of same day entries in the info I was provided which effects the rankings.

I'll take my info and compare to that link.

IcePack

We don't keep track of CL results. Only wars. Our file is the official record. If there is a glut of entries on certain dates in our data that is because those wars ended on that date. Nothing we can do about that! Also, if wars end on the same date, we try to find the last game played of each war and determine which one ended first. This was extremely important for the ladder and even though that is not active anymore we still pay attention to that variable.

[IcePack]

I don't see any CL results there? The reason I mentioned that is there were quite a glut of same day entries in the info I was provided which effects the rankings.

I'll take my info and compare to that link.

IcePack

We don't keep track of CL results. Only wars. Our file is the official record. If there is a glut of entries on certain dates in our data that is because those wars ended on that date. Nothing we can do about that! Also, if wars end on the same date, we try to find the last game played of each war and determine which one ended first. This was extremely important for the ladder and even though that is not active anymore we still pay attention to that variable.

Understood. However for CL results that you don't track, do you do the same date entry check when you move to completed? IE: if you move all CL first rounds together do they all bear the same date or do you check the final date on those as well?

IcePack

[chemefreak]

I don't see any CL results there? The reason I mentioned that is there were quite a glut of same day entries in the info I was provided which effects the rankings.

I'll take my info and compare to that link.

IcePack

We don't keep track of CL results. Only wars. Our file is the official record. If there is a glut of entries on certain dates in our data that is because those wars ended on that date. Nothing we can do about that! Also, if wars end on the same date, we try to find the last game played of each war and determine which one ended first. This was extremely important for the ladder and even though that is not active anymore we still pay attention to that variable.

Understood. However for CL results that you don't track, do you do the same date entry check when you move to completed? IE: if you move all CL first rounds together do they all bear the same date or do you check the final date on those as well?

IcePack

They are not moved all at the same time, however, for instance, Phase I - Division X - CL4 was not moved until the final game ended AND the TOs had completely updated all the charts, games, etc. then asked me to move them. (Which could have been days, if not weeks later) So let's say Otpisani had 5 series in the their division. The 1st series may have been over 7-10 weeks prior to it being moved to closed. You will have to go through the games and determine when the last one ended for each series...thus the reason we don't track those!

Also, let me point out one very important thing that goes all the way back to jpcloet...the clan leagues are TOURNAMENTS and not wars. Accordingly, they involve "forced" pairings. jpcloet never wanted to include these in official rankings since they may involve "unfair" challenges. Remember when we would not issue medals for the CCup? This was jpcloet's mandate because the tournament style events encourage unfair match-ups, especially at the beginning. Now, what changed with the CCup was the "go live" of the leap ladder. Basically, in the leap system, the rankings would not be effected by a high ranked clan demolishing a low ranked clan since there would be no movement on the ladder. This is what finally allowed the CCup match-ups to get medals and have their results included on the war tables. The CL match-ups are forced, way too small to yield accurate results, and a bitch to track the "mini-wars" contained therein. They also include map restrictions that apply to other challenges. So within each mini-challenge you can only use a map once. So by playing a clan earlier you have a wider range of maps to choose. Thus, the later games are not entirely reflective of the strength of your clan since you were map limited by the tournament structure.

Everyone likes the F400 because it pops out rankings that seem right. Which is fine with us. We are glad that it seems to work for everyone. However, if you really look into it, you will see real shortcomings of a formula like this when you start tracking different types of events using only one formula.

My advice would be to go through every CL event and determine when each "mini-war" ended to accurately get your results. However, since everyone's results would be skewed equally, perhaps pick the end date of the last one to end overall and use that result for EVERY "mini-war" for that certain event, at that level. Then your results may be skewed, but all done so equally, especially with the decay factor coming into play.

[IcePack]

Ok - I think I see where this is going let me use one quick RL example to clarify something.

EMP LHDD 8 4 04/01/12
TFFS MM 8 4 04/01/12
LHDD LEG 6 6 04/01/12
EMP MM 10 2 04/01/12

These CL matches/dates I have to assume come from "completed" clan area date. Was those dates (and many other CL result posted on 4/1/12) the actual date of completion or when it was moved to completed?

I understand the war / tournament bit, I just need to know how you handled things like this example so I can go back and accurately adjust the database to reflect what it should reflect.

IcePack

[IcePack]

MichelSableHeart...So basically.....

-----

Yes, basically.

[chemefreak]

Ok - I think I see where this is going let me use one quick RL example to clarify something.

EMP LHDD 8 4 04/01/12
TFFS MM 8 4 04/01/12
LHDD LEG 6 6 04/01/12
EMP MM 10 2 04/01/12

These CL matches/dates I have to assume come from "completed" clan area date. Was those dates (and many other CL result posted on 4/1/12) the actual date of completion or when it was moved to completed?

I understand the war / tournament bit, I just need to know how you handled things like this example so I can go back and accurately adjust the database to reflect what it should reflect.

IcePack

Those dates appear to be the date that the entire thread was moved to completed. It is not the date any of those ended.

[IcePack]

Ok. I will hve to go back and adjust some CL stuff then in database.
Please know that for August report this could cause some minor changes in scoring and decays so July - August wont necessarily "fit" together perfectly.

IcePack

[Gunn217]

The impact of those miniwars will be so miniscule that going back and keeping up with the date of each completed war will be more work than it's worth. At least when I was doing it.

[betiko]

so basically when we (the pack) faced OTP in CL and won 7-5 we lost points and they went up? not sure it's totally fair either.

[IcePack]

so basically when we (the pack) faced OTP in CL and won 7-5 we lost points and they went up? not sure it's totally fair either.

By definition you "should" have won by a larger margin. The point gain / loss is minimal.
The way the actual 400 works if the number 1 clan played the bottom clan they'd auto lose points, FD adjusted it so this wasn't the case and is minimized quite a bit.

If clan played Fallen for example, you would be "expected" to win by a good margin.
If the last place clan played first and the war ends 21-20 in favor of #1 - you don't think it's fair to say that last place performed well and achieved "above" their score and vice versa?

These are things it considers. But for cases like PACK/OTP the effect is minimal (it was CL result which is already toned down) and the score was close, so again minimized.

IcePack

[IcePack]

The impact of those miniwars will be so miniscule that going back and keeping up with the date of each completed war will be more work than it's worth. At least when I was doing it.

It is fairly minimal, but it does alter quite a few results minimally which ultimately could adjust a few positions. Since I got the time il just research and readjust finished dates to reflect the best / most accurate possible score the F400 can come up with.

IcePack

[Keefie]

By way of an example Ice could you show the impact of the recent CC3 war between HH and IA. We were ranked 20 something and IA 6th and the result was 23-18 to IA. So in my book HH performed better than the ranking would suggest.

[IcePack]

By way of an example Ice could you show the impact of the recent CC3 war between HH and IA. We were ranked 20 something and IA 6th and the result was 23-18 to IA. So in my book HH performed better than the ranking would suggest.

Sure I can do that. While I have plenty of time to help people analyze the system recreating each war / scenario is fun and all...but let's try not going crazy lol

To make it easier on me if you could provide me the date that specific war your requesting took place I'd appreciate it.

IcePack