TCEC Season 6 - Superfinal now live

[S.Taylor]

Martin, is there any way i can use chat, without going through a whole proccess of commitments and agreements with twitter or facebook or similar things?

You can register an account at www.chatwing.com and login to the chat with that.

They only require 2 things from you:

1: Username
2: Password

Thanks Martin!
It worked quite easily! Long last!

I was just going to say my name there, but i'm now thinking if my same name might be used on other chat forums where i don't want to be known from here! I did identify myself there already anyway. But on chessbomb i am "chess614". I won't be using that username anywhere else anyway.

[Ajedrecista]

Hello:

The current result is 15-11 (+4 -0 = 22) in favour of SF after 26 games. I decided to run my own simulations inspired by this post.

I supposed an a priori draw ratio of 22/26 ~ 84.62% and an Elo difference of 9 Elo in favour of SF based in TCEC ratings (I think it is much more). I did not coded an advantage for white. After these simplifications and fixing an initial score of 15 - 11:

Code: Select all

1e+7 simulations:

Stockfish wins:              9702702
Komodo wins:                  108191
Drawn matches:                189107

Approximated elapsed time:   40.14 seconds.

-------------------------------------------

1e+8 simulations:

Stockfish wins:             97022124
Komodo wins:                 1085963
Drawn matches:               1891913

Approximated elapsed time:  406.50 seconds.

It is too soon for kill Komodo options yet but my numbers speak for themselves. I would not bet in favour of Komodo (and it is supposing only 9 Elo difference). Just trying with an Elo difference of 50 Elo and the rest of parameters unchanged:

Code: Select all

1e+7 simulations:

Stockfish wins:              9999998
Komodo wins:                       0
Drawn matches:                     2

Approximated elapsed time:   42.20 seconds.

-------------------------------------------

1e+8 simulations:

Stockfish wins:             99999967
Komodo wins:                       1
Drawn matches:                    32

Approximated elapsed time:  418.31 seconds.

My only condition to declare a winner is check if any engine has more than 32 points after each game so I do not care about finishing the whole 64-game match or collect statistics about if the match is finished with 32.5-31.5 or 40-24. It is a trick for speed-up my simulator. I wonder Kai's results, just in case I did something wrong.

------------------------

I see that Martin posted a wallpaper in this thread. Here are more.

Thank you very much for your effort, Martin!

Regards from Spain.

Ajedrecista.

[Martin Thoresen]

Hello:
I supposed an a priori draw ratio of 22/26 ~ 84.62% and an Elo difference of 9 Elo in favour of SF based in TCEC ratings (I think it is much more).

Well whether you think it is much more right now is another matter, but the ratings in TCEC are based on all games in TCEC since Season 4 and is updated after each Stage.

So if Stockfish will go through the Superfinal undefeated, and perhaps with an even higher + score, it will gain more rating points while Komodo might lose some.

You can look at the rating as a sort of "TCEC performance rating", since it's based only on those previous games played in TCEC.

[IanO]

At the midway point, Stockfish is ahead by four points.

Code: Select all

 N Engine           Rtng  Pts  Gm     SB St           Ko          

 1 Stockfish 170514 3164 18.0  32 252.00 ············ + 6 = 24 - 2
 2 Komodo 7x        3155 14.0  32 252.00 + 2 = 24 - 6 ············

It was looking to be a one-sided match, but Komodo achieved two victories as white in the last six games.

[Ajedrecista]

Hello:

The result is 23-17 in favour of SF (+9 -3 =28) after 40 games out of 64. This means that SF will draw the match if it wins 9 points out of 24; SF loses if it wins 8.5 points or less; otherwise, SF wins TCEC Season 6.

Using my own programme of probabilities in a trinomial distribution, I have assumed an a priori draw ratio of 28/40 = 70% and SF being 9 Elo stronger than Komodo 7x, just sticking to TCEC ratings. Again, I have not take into account an Elo advantage when playing with the white pieces. I got 0.3807% of probability of a Komodo win, 0.7031% of probability of a tied match and 98.9162% of probability of a SF win in this final.

It is clear that SF has a clear edge (+6 after 40 games) but I wanted to quantify the chances of each outcome. The probabilitiy of SF winning between 33.5-30.5 and 36.5-27.5 is moreless 79.9425% ~ 80%.

I stay tuned for the final result.

Regards from Spain.

Ajedrecista.

[Milos]

Hello:

The result is 23-17 in favour of SF (+9 -3 =28) after 40 games out of 64. This means that SF will draw the match if it wins 9 points out of 24; SF loses if it wins 8.5 points or less; otherwise, SF wins TCEC Season 6.

Using my own programme of probabilities in a trinomial distribution, I have assumed an a priori draw ratio of 28/40 = 70% and SF being 9 Elo stronger than Komodo 7x, just sticking to TCEC ratings. Again, I have not take into account an Elo advantage when playing with the white pieces. I got 0.3807% of probability of a Komodo win, 0.7031% of probability of a tied match and 98.9162% of probability of a SF win in this final.

It is clear that SF has a clear edge (+6 after 40 games) but I wanted to quantify the chances of each outcome. The probabilitiy of SF winning between 33.5-30.5 and 36.5-27.5 is moreless 79.9425% ~ 80%.

I stay tuned for the final result.

Regards from Spain.

Ajedrecista.

I appreciate the effort, but assuming only 9 Elo points advantage of SF is nothing but laughable. At this TC SF has at least 30Elo advantage. More probable value is around 50Elo.
And regarding Komodo chances, well you can run your sims forever, but you would still not get a single case where Komodo wins, i.e. realistic chance of Komodo winning the match is well below 1e-9.