Awesome, thanks for analyzing the whole dataset!

And interesting, p is higher for mod 256 yet 387 and 388 were so high up the tiebreaker rankings. very interesting. so best call it a lower mod 256 rule of thumb then since the correlation is higher.

Edited 1/26/2023 15:29:27

Umm sorry. I call the parameter for alternative distribution p. P-value is something very different.

ah lol, pardon me noobing up the terms

[0, 0.9938650306748467, 1.0, 1.0, 1.0, 1.0, 0.9957805907172996, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 0]
[0, 128, 256, 256, 256, 256, 256, 256, 256, 128, 128, 128, 256, 256, 128, 256, 256, 256, 128, 128, 128, 128, 128, 128, 128, 0]

The first array is an array of point estimators for the best value of p for i-th season. The second is an array of best modulators for i-th season. I index from 0. If they are no data (0th season for example), there are both 0s.
If multiple modulators satisfy having the best value of p, I choose the smallest one, unless there is one which is the power of two - then I choose the biggest power of two.

Interpretation: the tiebreaker is clan ID mod 128 OR mod 256, which changes haphazardly every season.

Edit: season 6 is slightly weird - mod 257 eliminates less clans which have the same ID mod 257, so the best modulator is 257. But both modulator values have 1 broken clan.

Edited 1/26/2023 16:23:12

Things which are broken:

(511, 41, [(1, True)])

(511, 376, [(1, True)])

(199, 565, [(6, True)])

ignore the season 1 error with 511. the data goes off of wins, but there was weird phenomenon of the clans with 13 wins having different numbers of territories

511 (clan named Twelve) was awarded a 3rd territory with 13 wins
while
French Community and Varangian Guard also had 13 wins but only ended up with 2 territories.

Not sure if free wins or 2v2s were counted differently then or if clan 511 benefitted from some weird glitch. I believe it is supposed to take 15 total wins to have 3 territories (3+5+7)



Edited 1/26/2023 16:09:52

Here is my python code. You must name the data "vstup.txt" and have it in the same folder as your source code.

import re
import os
import math



vstup = open("vstup.txt", 'r',encoding="utf8")
Lines = vstup.readlines()
bestp = []
bestmodulator = []
for i in range (0, 26):
    bestp.append(0)
    bestmodulator.append(0)
pos = []
all = []
for i in range (0, 26):
    pos.append(0)
    all.append(0)
for line in Lines:

    numbers = re.findall(r'\d+', line)
    if int(numbers[0]) %modulator == int(numbers[1]) % modulator:
        pepa = 1
    elif int(numbers[0]) %modulator < int(numbers[1]) % modulator:
        for i in range(2, len(numbers)):
                pos[int(numbers)] += line.split()[2*i-1].count("True")
                all[int(numbers)] += 1
    else:
        for i in range(2, len(numbers)):
                pos[int(numbers)] += line.split()[2*i-1].count("False")
                all[int(numbers)] += 1
for i in range (0, 26):
    if all != 0:
        p = pos/all
        if p > bestp:
            bestp = p
            bestmodulator = modulator
        if p == bestp and math.log(modulator, 2).is_integer():
            bestmodulator = modulator

print(bestp)
print(bestmodulator)
[/code]

Edited 1/26/2023 16:20:15

So does the latest research by καλλιστηι mean that we have 99.9% solved what (not "why") is happenning?

So does the latest research by καλλιστηι mean that we have 99.9% solved what (not "why") is happenning?

Not necessarily. I found a very strong pattern. However, that doesn't mean that there isn't a more simple and better pattern.

Alright, if we convert 128, 256, 256, 256, 256, 256, 256, 256, 128, 128, 128, 256, 256, 128, 256, 256, 256, 128, 128, 128, 128, 128, 128, 128 into ones and zeroes and then transcribe the number in decimal, we get 8447103.

Could somebody call https://www.whitepages.com/phone/1-404-844-7103, please?

Looks like CW 25 was "a mod 128" season. Here are the first 4 tie-breaks:

Clan ID	mod 128	mod 256		CW Rank
236	108	236		15
379	123	123		16
				
325	69	69		18
671	31	159		17
				
7	7	7		19
553	41	41		20
697	57	185		21
				
444	60	188		29
257	1	1		27
696	56	184		28

I am so happy, that we together managed to deduce how this stuff works.

Edited 1/27/2023 16:49:37

I just want to suggest explanation why one clan get less territory than other clan with same number of win. Probably the winnings are happen on the last time slots. Games are not ended at the same time. Every time one game end, the winning registered and if there is enough win the clan will get one more territory. But at that last time slot all territories have been allocated. No more 'free territory'. All subsequent wins after that will not resulting additional territory, even if the total win should merit it.

I believe that wins that occur after the last territory is claimed do not count. They are clan war games that end up in purgatory that count for no season and end up neither red nor green but perpetually grey.

Dont know, why we are discussing the single CW outlier from an experimental CW season from 2 years ago, but here the "wins vs territories" breakdown between two clans:

Twelve		Games <> Games	French C…
Qaasuitsup	2	3	Philippines
Baffin-Nunavut	3	6	Borneo
Qeqqata		3	-	…
…		-	-	…
Unassigned	5	4	Unassigned

Definitely something was bugged.

---

At this point in time I am more interested in:
- Might there be any other natural explanation to the "mod N" behaviour other than "#BlameDB"?
- Why are we only now publicly discovering this and other stuff?

Edited 1/31/2023 04:29:16

Nobody cared before.

The weirdest thing is that modulator switches seemingly at random.

Also, what is the 2nd tiebreaker?

-pops popcorn and spectates-

While it no longer a mystery "what" happens, just a check on the recently completed season, that it is MOD 128 again.



For "why" this happens, Fizzer invited to send him a mail during his last AMA, but I guess no one bothered so far.

I am the programming guy. Social interactions are the job of somebody else.