Some stats / NS, FFA

we can consider switching to BY in the future.

fwiw i love what indi says, in 4pc we could do what regular chess should do but never will: make it 960 by default.

I would get on board with that. Playing a different set up every game would give us the variety we want, since we can never agree on just 1 setup. They all have pros and cons.

It could also rotate between the top 4 setups: OS, NS, BY, BYG.

From the marketing point of view, and clarity of explanation for newbies, it would be much more complicated to tell them "well, it's one of the following 4 set-ups" (why this one? have you tested all of them? who told these were the best ones?...) rather than saying simply "the position of your Q and K is random, and same for all other players".

In addition, if you only have 4 possibilities, it's clear that the best and/or most experienced players will learn by heart the best (?) or most common moves for the openings. It makes only 4 set-ups * 4 colours = 16 possibilities... They won't really think in the beginning. Whereas with 16 set-ups and 4 colours, it's rather unlikely they try to memorise the 64 openings.

I don't see FFA ever becoming 960 by default. It's good the default setup is one and players can analyze openings and try to improve their opening play. But would like to see 960 and capture the king be added as part of main FFA ratings. This means selecting these options will not stop the game counting towards FFA rating and leaderboard place.

Both of these options 960 and capture the king slightly slow the game down making teaming a little more difficult and help FFA be played as more of an individual game. Altering king value could also be made rated FFA again like it used to be.

Experimenting with different rules should be encouraged to help improve the game but unless games are rated no one will play.

As for the setups we currently have the most balanced setup for FFA if the best opening moves are played. If blue does not play correctly then green can have an unfair disadvantage. But at approx 2500+ you expect blue to open the diagonals properly.

With regards to the voting above on setups how can we be sure some people have not made multiple accounts to vote from. Not sure how BY has so many voters when barely any rated games of that setup have ever been played. I question the validity of that vote.

This vote was not representative anyway, because a) the sample is extremely reduced (about 120!), and b) most voters didn't really vote with due experience (for instance, I have no experience in BYG, so didn't vote for it, and this creates huge bias).

I don't think @Spacebar meant the full 960. I would simply prefer a random position for Q and K, hence 16 variants.

me

I wonder if we could see the RY win rate by average teams rating graph.

I am not 100% sure that's what you want, but if I sort all these games by the average of the R and Y ratings, and divide the sample into 10 deciles, the proportion of games where R or Y wins is the following:

Decile 1 average rating above 2826.5 65.4%

Decile 2 average rating between 2761 and 2826.5 60.8%

Decile 3 average rating between 2719 and 2761 50.6%

Decile 4 average rating between 2683.5 and 2719 46.4%

Decile 5 average rating between 2654 and 2683.5 50.6%

Decile 6 average rating between 2626.5 and 2654 49.4%

Decile 7 average rating between 2600 and 2626.5 41.8%

Decile 8 average rating between 2576.5 and 2600 48.5%

Decile 9 average rating between 2551.5 and 2576.5 46.8%

Decile 10 average rating between 2501 and 2551.5 42.5%

Rather logical I believe...

What seems to me more interesting, that's to see what is the relationship between the average rating of the board and the victory per colour.

Here you can see the results:

I found it surprising. (NB, as for the first 2 groups, the samples are statistically narrow, the results of course are much less relevant).
There were strangely much less games for 2500-2600 players than for 2600-2700 for instance. Rather strange. But it's facts. (More accurately, not for "2500-2600 players", but "for groups of players where the average rating was between 2500 and 2600").

Anyway as for the three biggest groups, we don't see any real advantage for RY. But it's clearly better to be red.

And this is the HIGHEST RATED GAME EVER PLAYED IN NS:

https://www.chess.com/variants/4-player-chess/game/33487476

(average rating 3054.25, mentioned above in the table).

José won against Radon, Caleb and Neo.

I was asked to examine what are the chances of the players per colour of the player who is 4th.

Here it is.

If B is 4th, R wins in 40.9% of the cases, Y in 32.3% and G in 26.8%.

If G is 4th, R wins in 40.7% of the cases, Y in 34.6% and B in 24.7%.

If R is 4th, B wins in 38.2% of the cases, G in 37.9% and Y in 23.9%.

If Y is 4th, B wins in 37.2% of the cases, G in 35.3% and R in 27.6%.

What does it mean?

Well, to begin with, it's not good for you if your opp is 4th (which is usually the case when he is eliminated first). It's a no-brainer and everybody knows that unless it's a noob under 2200 who doesn't know some basics (like https://www.chess.com/clubs/forum/view/basic-ffa-aspects).

But more interestingly, it means that for R in the 2nd stage the chances are the same whether he helps to eliminate his left or right neighbour.

For Y it's better to kill rather G, his left neighbour.

For B it's slightly better to kill rather R, his right neighbour (but the difference is very low).

For G it's better to kill rather R, his left neighbour.

So, in the 1st stage RY should rather focus on G; BG should rather focus on R. That maximises the teams' final result.

Intresting

This proves that 3v1 is the best strategy

No color is bad lol

BY sucks though

I couldn't think that the more elo they get, the more bet they do. It lets me think that the winning rates of 4 colors are equal.