While the CW map ranks clans by the amount of conquered territories, then by something - it is not correct and misleading.
The real ranking at the end of a CW season is done by the amount of wins and then some unknown factor.

The unknown factors are different, but that is not my question.

The question is: What is that factor that is used to rank clans at the end of CW seasons in case of wins-tie?

What has already been tested and proven wrong by the WGCDMC (Warzone Global Chat Data Mining Community)
(thanks @TR-8R):
- last win
- last game entered
- when last territory was completed
- alphabetical
- previous season rank

Although we have already seen examples when "random" stuff makes it way to production Warzone, personally I still refuse to believe, that the ranking is random...

How about clan ID?

nope, not clan ID, order on clan map (also order on clan map appears to be a different random but repeatable order, at least within the same season), clan rating, or any of the columns shown on https://wz-clanwars.netlify.app/

I am starting to believe it is truly random in cases where clans have same number of wins

you can review last season 24 to compare several ties at 15th place, 35th place (4 way tie), and 39th place (5 way tie) for no consistency in any hypotheses I have come up with outside of randomness

Edited 1/24/2023 18:40:01

To draw some additional attention, I would like to mention, that there has been multiple times already, when that unknown factor decided who would get the Top-5 spot among multiple clans.

Edited 1/24/2023 18:47:56

I probably won't add much more info, but feel free to check my data to be able to not redo work that has already been explored:
https://docs.google.com/spreadsheets/d/1P3jYQKoiagUIbhqzNU7q1Y8jNmHMVT89JC-Pr2Az3Wg/edit?usp=sharing

To save people the time from googling, the PostgreSQL documentation (https://www.postgresql.org/docs/current/queries-order.html) says:

After a query has produced an output table (after the select list has been processed) it can optionally be sorted. If sorting is not chosen, the rows will be returned in an unspecified order. The actual order in that case will depend on the scan and join plan types and the order on disk, but it must not be relied on. A particular output ordering can only be guaranteed if the sort step is explicitly chosen.

So based on that, I believe it primarily sorts by # of territories, secondly by # of total wins, and after that just uses the random order the list came out of that database.

Update: We haven't figured out what criteria is the tiebreaker for wins. But 5S and I have noticed many consistent preferences shown in tiebreakers between clans (roughly 90% of the time, the tiebreaker will have the same outcome between 2 given clans, sample size close to 1000). A limitation to this is that the vast majority of the ties happen among the lower performing clans with few wins.

After staring at the near 5000 rows of data for hours, my hypothesis is that there is a random set order that clans are ordered in, many of the exceptions to this rule appear to frequently happen to clans that start playing clan wars for the first time in the middle of the season.

Others are much more capable of doing more thorough analysis than myself. What I've found so far are the clans mentioned below appear to be at the top of the tiebreaker random order (Note: Clan ID is the # found in the url of each clan's page and happens to be in order of when the clan was founded) (Note 2: The sample sizes listed represent each tie between clans, therefore a 3 way tie within one season would be 2 ties between each of the 3 teams)

These 3 are undefeated:
Clan ID 257 (CORP, sample size 55)
Clan ID 1 ([REGL] The Red Eagles, sample size 31)
Clan ID 512 (American militia, sample size 21)

Others with some defeats but very high in the tiebreaker random order:
Clan ID 513: The Holy Brandi Empire
Clan ID 514: Vigor
Clan ID 2: (AHoL)
CLan ID 3: {rp} The Roleplayers Guild
Clan ID 259: Owl Dragoons
Clan ID 260: Outlaws
Clan ID 261: Nova Scotia
Clan ID 389: AMK
Clan ID 517: The Connoisseurs
Clan ID 520: Menace
Clan ID 269: Legio Invictus

What I found is that any clan not listed above would be later in the tiebreaker order and would lose to any of those clans were you to end up with the same number of wins in a clan war season.

On the flip side, these clans appeared to at the bottom of the tiebreaker random order, I was not as thorough in looking for all the clans at the bottom but these stood out to me:

At least this one clan without 1 tiebreaker going their way Clan ID: 255 ([ACE], Sample Size 83)
Clan ID 510: ズKING
Clan ID 511: Twelve
Clan ID 230: Cacaoçian Empire
Clan ID 232: The Mafia
Clan ID 236: SPARTA
Clan ID 464: Nestlings
Clan ID 489: Python
Clan ID 376: Varangian Guard
Clan ID 507: The Last Men Standing

Fortunately, the vast majority of these ties are in scenarios where it doesn't impact the rewards for a clan. But regardless it would be nice to improve the wins tiebreaker from random to something like winrate or clan rating.

Thanks, TR-8R.

One may also spot from your lists that:
- the biggest "winners" usually have ID slightly greater than a multiple of 256 (or 128)
- the sore "losers" usually have ID slightly less than a multiple of 256 (or 128)

So that reinforces the theory, that it is somehow due to the programming backend of Warzone.
The most obvious then is the DB internal sort, like JK_3 stated.

PS. We always joked that the raffles were rigged, but it turns out that we should have looked closely at the other source of free coins in Warzone...

Did we do a check on winrate? So wins vs games played?

Did we do a check on winrate? So wins vs games played?

This has been tested and proven false.

Do you have individual sorted clan cohorts for each tiebreaker event? We could use that to test whether this sort is a result of how Postgres stores data under the hood. Interesting observation around 2^8, which makes me think it's some sort of implementation detail, because 8 bits make a byte.

We could even test this out by spinning up a Postgres database (I believe the version used by Warlight was announced at some point when Warlight had to do a db upgrade) and seeing how it sorts integer (probably w/ an index or a foreign key). Could also check postgres source: https://github.com/postgres/postgres

Try to sort clans by winning in tiebreaks. For example, if CORP beats Python in tiebreak, then CORP > Python.

Does it break somewhere?
If yes, then the result is random/ criteria change every season.
If not, then results depend on a constant (for example ID, name etc.)

Edited 1/25/2023 18:06:04

What about head-to-head outcomes as a tie-breaker?

Hey l4v.r0v

Yes, if you go to the other tabs of my google sheets linked earlier in this thread, you can see that. Shout out to 5S for retrieving the raw data.

Sheet 2 has all ties between clans shown, sorted by length.
You'll see cells filled with info like: (41, 197, [(16, True), (17, True), (19, True), (20, True), (22, True)])
where the first two numbers are the 2 clan ID, then whether the first clan ID won the tiebreaker in given seasons.

Copy of Sheet 2, I sorted all the cases that were not consistent to the bottom (ie the tiebreaker didn't go the same way between the same 2 clans every time) (X means inconsistent vs. Y means consistent)

Some of the exceptions were already discussed outside of this forum, such as season 1 having a weird case of clans 511 and 41 having the same number of wins (13) but a different amount of territories. Not sure if 2v2s or free wins were handled differently back then or was a glitch that wasn't ironed out prior to season 1. WZ obviously is going to rank the clan with more territories ahead, so was odd glitch where thereafter the 4 times that these clans were tied again, the same clan won the tiebreak.

Many of the other exceptions looked to be clans that played clan wars for the first time that season (for instance Clan 389 was put lower in the tiebreaker pecking order in its first cw season 7 compared to season 14).

A curious case is between clans 306 and 411, where they tied 6 times, but the results alternated. I will take this as the exception that proves the rule lol. The around 90% prevalence of consistent tiebreaker order does suggest this being a somewhat predictable trend compared to a completely different random ranking every season.

Edited 1/25/2023 18:18:32

Why not just ask in the next AMA?

I recall others asking him before. I don't think he even knows.

LMAO

Well if he doesn't know . . . good luck figuring it out

Of course, what he's really saying is "I don't remember and am not committing to spending cycles to look at the code to figure it out"

As others have theorized, he likely didn't intentionally put a 2nd tiebreaker (1st tiebreaker between same number of territories being wins), so it is in a random order (aka the order that the code will spit out per the programmers here)

I can't imagine that he doesn't have secondary criteria, b/c that would mean he didn't foresee the possibility of 2 clans competing for 1st place to have same # of wins ... he's smarter than that

For examples, all it takes is 1 territory left in CW map, 2 clans tied for 1st place with both at exact same # of wins, same other clan (3rd place or lower) finishes a game that takes the last territory ending the season . . .

I can't imagine that he wouldn't have foreseen that

But maybe people are asking the wrong question; perhaps it's not "how do you determine the winning order?" (to which the answer is just refer to the order in the ongoing list) but rather how do you sort the clans into the ongoing list. The one we see on the CW map. It's likely just populating that view, and taking the order of clans from there to decide placement. They are obviously ordered primarily by # of wins, so what are the secondary, tertiary sort keys? And if there are none, then it is like getting a CW match before the randomness update was implemented, how they were ordered by WZ join date (actually account ID which correlates to WZ join date). But since CW winners are at clan level and not individual player level, it could literally be any aggregate field such as most # of points won at a clan level, or just the primary key from some other table involved in the join/query ... so maybe just the order clans appear in this list https://www.warzone.com/Clans/List, which I think actually is sum of all points earned in the last ?month? (some period of time)

But as he doesn't know, it's likely NOT going to be some complicated active net new tiebreaker criteria like some "difficulty" factor to assess who got the harder matches that season such as sum of difference of CW ratings for each match, and the lowest (closest to 0 or lowest negative #) would be given the edge for higher placement (so for example, Cats beating ONE would be worth more than ONE beating Cats -- but only as pertains to breaking ties). He'd probably remember doing something like that.

I also confirmed that the order that they are listed on the map is different from the tiebreaker order. The order they are listed on the clans page is also different from the tiebreaker order. I believe the clans list is the same order as the previous clan war season list, but this list doesn't even list the clans in order of wins, but merely by territories and some other order different from the tiebreaker order.

Are you even trying? ;-) These were ruled out in minutes.

Edited 1/25/2023 18:54:57