TCEC 10

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, Harvey Williamson, bob

Dann Corbit
Posts: 8589
Joined: Wed Mar 08, 2006 7:57 pm
Location: Redmond, WA USA
Contact:

Re: TCEC 10

Post by Dann Corbit » Wed Nov 15, 2017 7:53 pm

Jouni wrote:If Stockfish goes stage 2 lossless I consider it already as winner. No interest to continue with possible K vs H final :!: .
Funny that the whole thing pivots for you on a single loss (as the others have).

Surely you realize how random that is.
Taking ideas is not a vice, it is a virtue. We have another word for this. It is called learning.
But sharing ideas is an even greater virtue. We have another word for this. It is called teaching.

Leo
Posts: 447
Joined: Fri Sep 16, 2016 4:55 pm
Location: Saint Paul, Minnesota USA

Re: TCEC 10

Post by Leo » Wed Nov 15, 2017 7:59 pm

Jouni wrote:If Stockfish goes stage 2 lossless I consider it already as winner. No interest to continue with possible K vs H final :!: .
I have similar feelings but I will respect the final winner also.
Brainfish fan. AMD fan.

JJJ
Posts: 1186
Joined: Sat Apr 19, 2014 11:47 am

Re: TCEC 10

Post by JJJ » Wed Nov 15, 2017 8:27 pm

gotogo wrote:It's game selection and it is to favor komodo! absolutely the worst tcec yet!
only three stages get rid of the competition early! Tcec is nothing more than a scam!
For the jerks that would think I might be arb well what can I say other than you're a jerk!
Hello Arb !

Dirt
Posts: 2851
Joined: Wed Mar 08, 2006 9:01 pm
Location: Irvine, CA, USA

Re: TCEC 10

Post by Dirt » Thu Nov 16, 2017 12:07 am

Dann Corbit wrote:
Jouni wrote:If Stockfish goes stage 2 lossless I consider it already as winner. No interest to continue with possible K vs H final :!: .
Funny that the whole thing pivots for you on a single loss (as the others have).

Surely you realize how random that is.
No more random than basing the winner on how it did against the weakest engines. An argument could be made that fewer losses is more important than more wins.
Deasil is the right way to go.

Nay Lin Tun
Posts: 269
Joined: Mon Jan 16, 2012 5:34 am

Re: TCEC 10

Post by Nay Lin Tun » Thu Nov 16, 2017 12:45 am

Dirt wrote: An argument could be made that fewer losses is more important than more wins.
No need to make arguments, 1 win with 1 loss will get the same score as 2 draws. (You can see in the rules,) But in comparison of scores between engines A -10 draws vs Engine B 9 win 1 loss. Engine B will advance. and that should logically be!

Dann Corbit
Posts: 8589
Joined: Wed Mar 08, 2006 7:57 pm
Location: Redmond, WA USA
Contact:

Re: TCEC 10

Post by Dann Corbit » Thu Nov 16, 2017 12:53 am

Dirt wrote:
Dann Corbit wrote:
Jouni wrote:If Stockfish goes stage 2 lossless I consider it already as winner. No interest to continue with possible K vs H final :!: .
Funny that the whole thing pivots for you on a single loss (as the others have).

Surely you realize how random that is.
No more random than basing the winner on how it did against the weakest engines. An argument could be made that fewer losses is more important than more wins.
Give me daring chess any day, over a draw master. But that is just personal preference. The winner is determined by points and SB.
Taking ideas is not a vice, it is a virtue. We have another word for this. It is called learning.
But sharing ideas is an even greater virtue. We have another word for this. It is called teaching.

Dirt
Posts: 2851
Joined: Wed Mar 08, 2006 9:01 pm
Location: Irvine, CA, USA

Re: TCEC 10

Post by Dirt » Thu Nov 16, 2017 1:56 am

Nay Lin Tun wrote:
Dirt wrote: An argument could be made that fewer losses is more important than more wins.
No need to make arguments, 1 win with 1 loss will get the same score as 2 draws. (You can see in the rules,) But in comparison of scores between engines A -10 draws vs Engine B 9 win 1 loss. Engine B will advance. and that should logically be!
The problematic situation is Stockfish with no losses but three fewer wins versus one loss and three more wins for Houdini and Komodo. The rules do say that Stockfish failed, but I'm not at all sure that it would be the weaker program.
Deasil is the right way to go.

jhellis3
Posts: 340
Joined: Fri Aug 16, 2013 10:36 pm

Re: TCEC 10

Post by jhellis3 » Thu Nov 16, 2017 2:13 am

The problem is with the tournament format. It either needed another RR before the superfinal, or to be a knockout match style tournament. As it is now, there are 8 engines 4 of which have a 0% chance of advancing. So 25% of the games are effectively meaningless and can only serve to add noise do to the random nature of the openings selected.

A knockout style tournament would have none of these problems. The engines are seeded and play 4 matches 1&8, 2&7, 3&6, 4&5 say of 10 openings. That gives 20 games per match x 4 matches = 80 games. The 4 winners do so again in 2 matches of 20 openings for the same number of 80 games. And the final is 40 openings for the same 80 games. 240 games total for the entire event.

This completely avoids random chance benefiting one engine over another in the openings given to it. Of course, without any controversy, what would people complain about?

Dann Corbit
Posts: 8589
Joined: Wed Mar 08, 2006 7:57 pm
Location: Redmond, WA USA
Contact:

Re: TCEC 10

Post by Dann Corbit » Thu Nov 16, 2017 2:21 am

jhellis3 wrote:The problem is with the tournament format. It either needed another RR before the superfinal, or to be a knockout match style tournament. As it is now, there are 8 engines 4 of which have a 0% chance of advancing. So 25% of the games are effectively meaningless and can only serve to add noise do to the random nature of the openings selected.

A knockout style tournament would have none of these problems. The engines are seeded and play 4 matches 1&8, 2&7, 3&6, 4&5 say of 10 openings. That gives 20 games per match x 4 matches = 80 games. The 4 winners do so again in 2 matches of 20 openings for the same number of 80 games. And the final is 40 openings for the same 80 games. 240 games total for the entire event.

This completely avoids random chance benefiting one engine over another in the openings given to it. Of course, without any controversy, what would people complain about?
The other games except for the top 3 are still meaningful.
The weaker engines play the role of spoiler.
No matter what format you choose or how many games you play two of the top three engines among Houdini, Stockfish, Komodo are going to advance and the format chosen will not change that.

As for a knockout tournament, you won't be able to reverse the books. The books are deliberately slanted as far as having intentionally unbalanced openings so that we will not have 110 draws and two decisive games for the contest.

In the end, there are 3 real engines with a chance for the finals. One will be left in the cold, perhaps a little unfairly. But a contest like this does not tell us which engine is strongest. It only names a champion.

And as far as knockout format for a world chess championship...
Alexander Khalifman.
Need I say more?
Taking ideas is not a vice, it is a virtue. We have another word for this. It is called learning.
But sharing ideas is an even greater virtue. We have another word for this. It is called teaching.

jhellis3
Posts: 340
Joined: Fri Aug 16, 2013 10:36 pm

Re: TCEC 10

Post by jhellis3 » Thu Nov 16, 2017 2:28 am

No matter what format you choose or how many games you play two of the top three engines among Houdini, Stockfish, Komodo are going to advance and the format chosen will not change that.
Strawman.
As for a knockout tournament, you won't be able to reverse the books.
Um... yes you can. If you had actually bothered to read what I wrote before responding.... you might have noticed I said have each of the 2 engine pairs play a 10 opening match with 20 games total. Hmmm.... I wonder why that could possibly be. To repeat those 10 openings perhaps?
Need I say more?
Not if you are going to keep posting drivel like that.

Post Reply