TCEC Season 10 announced

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

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Modern Times
Posts: 2554
Joined: Thu Jun 07, 2012 9:02 pm

Re: TCEC Season 10 announced

Post by Modern Times » Thu Aug 24, 2017 8:02 am

A game that takes an hour isn't 'rapid" and on the massive hardware they use the "quality" is still exceptionally high. That to me is the point of the big hardware, to shorten the time control, make the games watchable, and not sacrifice quality.
.

Opinions expressed here are my own, and not necessarily those of the CCRL Group.

User avatar
Thomas Lagershausen
Posts: 269
Joined: Mon Jun 11, 2007 4:59 pm

Re: TCEC Season 10 announced

Post by Thomas Lagershausen » Thu Aug 24, 2017 8:49 am

Modern Times wrote:A game that takes an hour isn't 'rapid" and on the massive hardware they use the "quality" is still exceptionally high. That to me is the point of the big hardware, to shorten the time control, make the games watchable, and not sacrifice quality.
You make the point. I am a serious tournamentplayer and like to play real tournamentgames, but watching games is a complete different story. It is boring to wait for obvious moves and such situations often repeat.

Please stop this waste of time. 40 moves in 40 minutes and 30 minutes for the rest of the game is a good choice for me to play such a big tournament in a useful time.

Time in life is limited.

Thx for your attention.
TL

gotogo
Posts: 91
Joined: Tue Jan 26, 2010 4:03 am

Re: TCEC Season 10 announced

Post by gotogo » Thu Aug 24, 2017 9:09 am

looks like the stockfish weakening process has started...

Looking forward to Houdini taking the crown!

Damir
Posts: 2358
Joined: Mon Feb 11, 2008 2:53 pm
Location: Denmark
Full name: Damir Desevac

Re: TCEC Season 10 announced

Post by Damir » Thu Aug 24, 2017 10:09 am

Yes, deleting parameter Slow Mover which is there to regulate Stockfish time, so it does not lose on time was a big drawback in my opinion. It is possibly the dumbest move Stockfish Team has ever made so far.
Both online and offline Stockfish loses on time in draw game, and sometimes even in won position...

Branko Radovanovic
Posts: 65
Joined: Sat Sep 13, 2014 2:12 pm

Re: TCEC Season 10 announced

Post by Branko Radovanovic » Thu Aug 24, 2017 1:31 pm

I hate to state the obvious, and all this is not new, but...

What is the benefit of having a 44-core machine rather than a 24-core machine? 15 Elo, perhaps? It's not only that engines do not play significantly better - I'm sure many play worse with 44 cores. (And the developers have little chance of addressing this, because they don't have access to machines like these.) In particular, what's the point of renting an expensive server, which provides marginally better play at best, and then complaining about the finances?

What's the benefit of ridiculous TCs, especially in the final? More tedious to watch, with even more draws? In the past couple of seasons, I watched the qualifying stages, but skipped the final altogether because it was unwatchable. What is the purpose of creating a tournament in which the final - normally the most exciting phase! - is in fact the most boring?

Apart from that, I wish that the rapid tournament also doubled as Stage 1, so that the winner becomes the rapid champion, and that one half of the engines progress to Stage 2. I've never liked splitting Stage 1 into 2 groups - double round robin's outcome is both more reliable (being based on more games) and more fair (no good or bad luck due to being placed in the weaker or the stronger group). Also results in a shorter tournament - last year's May to November was excessively long.

royb
Posts: 520
Joined: Wed Mar 08, 2006 11:53 pm

Re: TCEC Season 10 announced

Post by royb » Thu Aug 24, 2017 2:14 pm

Branko Radovanovic wrote:I hate to state the obvious, and all this is not new, but...

What is the benefit of having a 44-core machine rather than a 24-core machine? 15 Elo, perhaps? It's not only that engines do not play significantly better - I'm sure many play worse with 44 cores. (And the developers have little chance of addressing this, because they don't have access to machines like these.) In particular, what's the point of renting an expensive server, which provides marginally better play at best, and then complaining about the finances?

What's the benefit of ridiculous TCs, especially in the final? More tedious to watch, with even more draws? In the past couple of seasons, I watched the qualifying stages, but skipped the final altogether because it was unwatchable. What is the purpose of creating a tournament in which the final - normally the most exciting phase! - is in fact the most boring?

Apart from that, I wish that the rapid tournament also doubled as Stage 1, so that the winner becomes the rapid champion, and that one half of the engines progress to Stage 2. I've never liked splitting Stage 1 into 2 groups - double round robin's outcome is both more reliable (being based on more games) and more fair (no good or bad luck due to being placed in the weaker or the stronger group). Also results in a shorter tournament - last year's May to November was excessively long.
I agree about the relatively small benefit of using a server with 44 cores vs a more modest amount of cores, thereby saving some money in the process. The strength boost that engines get from super-high numbers of cores has to be pretty small.

And time controls in the past were too long for my tastes. I have to work and could very rarely (on a weekend) watch a whole game from start to finish. I usually downloaded the PGN of a game afterwards and went through it on my laptop. It would be much more fun to watch the games live with games finishing in an hour or two with a higher likelihood of non-drawn games.

Jeroen
Posts: 501
Joined: Wed Mar 08, 2006 8:49 pm

Re: TCEC Season 10 announced

Post by Jeroen » Thu Aug 24, 2017 2:40 pm

That, and the >90% draw rate.
The draw rate in the super final of season 9 was only 75%.
This was the new Cato strategy in S9 to reduce the draw rate
It was not Cato's strategy. Besides, after the superfinal of season 8 many people were asking for more unbalanced lines.
Unless you think it's funny to play games with voluntarily bogus book lines that give white a winning advantage.
None of the lines gave white 'a winning advantage'. Furthermore, there were only two 1-1 results with white winning both games.
And S10 is getting even worse. Book lines are getting longer instead of shorter...
In tournaments where real books are used, you'll sometimes see many, many more moves of theory then you have seen in TCEC.

If you watch the games in the Mar del Plata match between K and SF on the Tournaments and Matches forum, you'll notice that you can get great games using more moves.

Short move book = more boring chess, more standard play, more draws. Especially with a LTC match on fast hardware.

Hai
Posts: 548
Joined: Sun Aug 04, 2013 11:19 am

Re: TCEC Season 10 announced

Post by Hai » Thu Aug 24, 2017 4:40 pm

Jeroen wrote:
That, and the >90% draw rate.
Short move book = more boring chess, more standard play, more draws. Especially with a LTC match on fast hardware.
And that's the reason why we should use shortest opening book
"the 20 possible opening moves" = much better chess, less standard play, draw rate around 50% - from more or less 0.00 opening positions.

leavenfish
Posts: 282
Joined: Mon Sep 02, 2013 6:23 am

Re: TCEC Season 10 announced

Post by leavenfish » Fri Aug 25, 2017 1:17 am

royb wrote:
Branko Radovanovic wrote:I hate to state the obvious, and all this is not new, but...

What is the benefit of having a 44-core machine rather than a 24-core machine? 15 Elo, perhaps? It's not only that engines do not play significantly better - I'm sure many play worse with 44 cores. (And the developers have little chance of addressing this, because they don't have access to machines like these.) In particular, what's the point of renting an expensive server, which provides marginally better play at best, and then complaining about the finances?

What's the benefit of ridiculous TCs, especially in the final? More tedious to watch, with even more draws? In the past couple of seasons, I watched the qualifying stages, but skipped the final altogether because it was unwatchable. What is the purpose of creating a tournament in which the final - normally the most exciting phase! - is in fact the most boring?

Apart from that, I wish that the rapid tournament also doubled as Stage 1, so that the winner becomes the rapid champion, and that one half of the engines progress to Stage 2. I've never liked splitting Stage 1 into 2 groups - double round robin's outcome is both more reliable (being based on more games) and more fair (no good or bad luck due to being placed in the weaker or the stronger group). Also results in a shorter tournament - last year's May to November was excessively long.
I agree about the relatively small benefit of using a server with 44 cores vs a more modest amount of cores, thereby saving some money in the process. The strength boost that engines get from super-high numbers of cores has to be pretty small.

And time controls in the past were too long for my tastes. I have to work and could very rarely (on a weekend) watch a whole game from start to finish. I usually downloaded the PGN of a game afterwards and went through it on my laptop. It would be much more fun to watch the games live with games finishing in an hour or two with a higher likelihood of non-drawn games.
+1
Ironic isn't it
In real, human chess, consensus is that quicker time controls makes for more eyeballs on it...in engine chess, some think you have to have longer time controls and bigger and faster hardware.

The second ironic point is that the quality of human chess improves considerably with more time...not so much with engine chess. There is clearly a sweet spot TCEC is not even interested in getting to. And all the talk about the $$$...less cores = less cost.


:roll: :roll: :roll:

User avatar
Nordlandia
Posts: 2625
Joined: Fri Sep 25, 2015 7:38 pm
Location: Sortland, Norway

Re: TCEC Season 10 announced

Post by Nordlandia » Fri Aug 25, 2017 4:17 am

Code: Select all

It was not Cato's strategy. Besides, after the superfinal of season 8 many people were asking for more unbalanced lines. 
I think Cato should allow more semi-dubious lines to reduce draw rate.

[pgn][Event "?"]
[Site "?"]
[Date "????.??.??"]
[Round "?"]
[White "New game"]
[Black "?"]
[Result "*"]
[PlyCount "18"]

1. e4 c6 2. d4 d5 3. e5 Bf5 4. Nc3 Qb6 5. Bd3 Qxd4 6. Nf3 Qg4 7. h3 Qh5 8. O-O

Bxd3 9. cxd3 e6 *[/pgn]

[pgn][Event "?"]
[Site "?"]
[Date "????.??.??"]
[Round "?"]
[White "New game"]
[Black "?"]
[Result "*"]
[PlyCount "10"]

1. g4 d5 2. e3 e5 3. d3 h5 4. gxh5 Nf6 5. Be2 Nc6 *[/pgn]

[d]rn1q1r2/p1p1ppkp/1p4p1/8/2BP4/2N2N2/PP3PPP/R1B1K2R b KQ - 0 11

+0.80 | Advantage of the bishop pair & initiative

[d]r2qkb1r/ppp1n1pp/4p3/6B1/3pP3/5Q2/PPP2PPP/R3KB1R b KQkq - 0 10

[pgn][Event "?"]
[Site "?"]
[Date "????.??.??"]
[Round "?"]
[White "New game"]
[Black "?"]
[Result "*"]
[PlyCount "11"]

1. e4 c6 2. d4 d5 3. exd5 cxd5 4. Bd3 Nc6 5. h3 Nxd4 6. Bxh7 *[/pgn]

Post Reply