Why don't strong engines enter WCCC?

Discussion of computer chess matches and engine tournaments.

Moderators: Harvey Williamson, bob, hgm

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Post Reply
User avatar
hgm
Posts: 25060
Joined: Fri Mar 10, 2006 9:06 am
Location: Amsterdam
Full name: H G Muller
Contact:

Re: Why don't strong engines enter WCCC?

Post by hgm » Mon Aug 04, 2014 3:57 pm

vittyvirus wrote:Most ~2800 engines do calculate around 100-500 kNodes/sec on my system. Rybka, which merely calculates 100 kNodes/ sec with single core on my system, is rated 3161
So Rybka obfuscates its similarity to other programs by counting nodes in a different and unusual way. And many ignorant users were fooled by it. It is also known to lie about its depth, by subtracting 2 or 3 ply from what it really searches.

This just shows how dangerous it is to have uninformed opinions on complex matters. Letting the engine report false node count and depth is the simplest trick in the book of engine cloners. No conclusion can be drawn at all from what an engine prints.

Sedat Canbaz
Posts: 3018
Joined: Thu Mar 09, 2006 10:58 am
Location: Antalya/Turkey
Contact:

Re: Why don't strong engines enter WCCC?

Post by Sedat Canbaz » Mon Aug 04, 2014 4:17 pm

hgm wrote:
vittyvirus wrote:Most ~2800 engines do calculate around 100-500 kNodes/sec on my system. Rybka, which merely calculates 100 kNodes/ sec with single core on my system, is rated 3161
So Rybka obfuscates its similarity to other programs by counting nodes in a different and unusual way. And many ignorant users were fooled by it. It is also known to lie about its depth, by subtracting 2 or 3 ply from what it really searches.

This just shows how dangerous it is to have uninformed opinions on complex matters. Letting the engine report false node count and depth is the simplest trick in the book of engine cloners. No conclusion can be drawn at all from what an engine prints.
I agree ...this is not right measuring to determine about which chess engine is a clone or not, e.g if we are looking at the engine's kn/s output values!

But Syed stated also that Rybka is rated 3161 Elo (more than + 400 Elo over Fruit) and what do you think about this BIG improvement ???

Is that be possible: any clone or derivative engine to have such Elo difference (+ 400 Elo) more than based original engine ???

Sedat Canbaz
Posts: 3018
Joined: Thu Mar 09, 2006 10:58 am
Location: Antalya/Turkey
Contact:

Re: Why don't strong engines enter WCCC?

Post by Sedat Canbaz » Mon Aug 04, 2014 4:34 pm

One thing more,
How many years is past...and we still discuss Rybka...and this would not be exaggeration,if we say:
- Vasik Rajlich is a really great programer !!

User avatar
Dr.Wael Deeb
Posts: 9772
Joined: Wed Mar 08, 2006 7:44 pm
Location: Amman,Jordan

Re: Why don't strong engines enter WCCC?

Post by Dr.Wael Deeb » Mon Aug 04, 2014 5:03 pm

vittyvirus wrote:Here's Houdini 3 Pro CRUSHING Deep Junior in just 21(!) moves with lots of sacrifices
[pgn][Event "CCRL 40/40"]
[Site "CCRL"]
[Date "2012.10.29"]
[Round "343.3.338"]
[White "Houdini 3 64-bit 4CPU"]
[Black "Deep Junior 13 64-bit 4CPU"]
[Result "1-0"]
[BlackElo "3044"]
[ECO "C25"]
[Opening "Vienna"]
[Variation "Hamppe-Allgaier Gambit"]
[WhiteElo "3250"]
[TimeControl "40/1800:40/1800:1800"]
[Termination "normal"]
[PlyCount "41"]
[WhiteType "human"]
[BlackType "human"]

1. e4 {0s} e5 {0s} 2. Bc4 {0s} Nc6 {0s} 3. Nc3 Nf6 {0s} 4. f4 {0s} Nxe4
{0s} 5. Nf3 {0s} Nd6 {0s} 6. Bd5 {-0.01/22 71s} exf4 {-0.11/19 48s} 7. O-O
{(d2-d4) +0.17/23 69s} Be7 {(h7-h6) -0.11/19 47s} 8. d4 {+0.31/22 49s} g5
{-0.11/18 31s} 9. g3 {+0.11/21 56s} Bf6 {(h7-h5) +0.08/17 25s} 10. Ne5
{(g3xf4) +0.19/19 44s} h5 {+0.50/17 67s} 11. h4 {+0.67/19 30s} Rh7
{(Bf6xe5) +0.80/17 71s} 12. hxg5 {+1.99/20} Bxg5 {+1.24/19 57s} 13. Bxf4
{+2.00/19} f5 {(Bg5xf4) +1.12/19 48s} 14. Qd3 {+2.76/21 61s} Nb4 {(Qd8-f6)
+1.45/19 95s} 15. Qe2 {(Qd3-d2) +5.02/19 13s} Kf8 {(Qd8-e7) +1.29/19 120s}
16. Bxg5 {(Ra1-e1) +8.81/18 14s} Qxg5 {+5.74/21} 17. Ne4 {+8.83/17 1s} Nxe4
{+6.16/21} 18. Rxf5+ {+8.68/16 0s} Qxf5 {+7.15/21} 19. Bxe4 {+8.38/15 0s}
Qe6 {+9.60/21} 20. Rf1+ {+17.82/22 50s} Ke7 {+9.60/20} 21. Ng6+ {+17.82/21
0s} 1-0
[/pgn]
Its clear that Houdini would win after move 13, and Deep Junior is programmed to not resign before move 20, or it would've resigned on move 15. Deep Junior is Junior, and I don't know why I've to say this, that giving it WCC Champion title is injustice.
This is also the shortest Engine win in 3000+ level.
It's amazing how did Deep Junior destroyed his kingside defencing structure at the begining of the game....

It wasn't even a counter attack that it strated on it's kingside,it was a chess suicide,nothing more,nothing less regards,
Dr.D
_No one can hit as hard as life.But it ain’t about how hard you can hit.It’s about how hard you can get hit and keep moving forward.How much you can take and keep moving forward….

User avatar
hgm
Posts: 25060
Joined: Fri Mar 10, 2006 9:06 am
Location: Amsterdam
Full name: H G Muller
Contact:

Re: Why don't strong engines enter WCCC?

Post by hgm » Mon Aug 04, 2014 6:19 pm

Sedat Canbaz wrote:But Syed stated also that Rybka is rated 3161 Elo (more than + 400 Elo over Fruit) and what do you think about this BIG improvement ???

Is that be possible: any clone or derivative engine to have such Elo difference (+ 400 Elo) more than based original engine ???
Of course. There is no limit to how much you can improve a well-written engine like Fruit. Even the strongest engines we know now use a search, alpha-beta pruning, reductions and extensions, null move, Pawn-structure evaluation with recognition of passers, backward Pawns, doubled and isolate Pawns. And King safety, paying attention to Pawn shield, number of squares under attack in th eKing neighborhood. They use hash tables and Pawn hash, repetition detection, material hash.

Fruit does all that too. And the code that does it in Fruit is directly usable in a stronger engine. In my estimate it would only require 10% extra code to add 400 Elo to Fruit. The other 90% could be directly copied from it.

Sedat Canbaz
Posts: 3018
Joined: Thu Mar 09, 2006 10:58 am
Location: Antalya/Turkey
Contact:

Re: Why don't strong engines enter WCCC?

Post by Sedat Canbaz » Mon Aug 04, 2014 7:03 pm

hgm wrote:
Sedat Canbaz wrote:But Syed stated also that Rybka is rated 3161 Elo (more than + 400 Elo over Fruit) and what do you think about this BIG improvement ???

Is that be possible: any clone or derivative engine to have such Elo difference (+ 400 Elo) more than based original engine ???
Of course. There is no limit to how much you can improve a well-written engine like Fruit. Even the strongest engines we know now use a search, alpha-beta pruning, reductions and extensions, null move, Pawn-structure evaluation with recognition of passers, backward Pawns, doubled and isolate Pawns. And King safety, paying attention to Pawn shield, number of squares under attack in th eKing neighborhood. They use hash tables and Pawn hash, repetition detection, material hash.

Fruit does all that too. And the code that does it in Fruit is directly usable in a stronger engine. In my estimate it would only require 10% extra code to add 400 Elo to Fruit. The other 90% could be directly copied from it.

I fully disagree with you !!

Or maybe you are joking, right )) ?

I am not an engine developer, but I have a little bit experience in computer chess

And from my experience I can say,
To improve + 400 Elo...is not so easy as it looks !
If it was so easy...the opponents would not prefer to sign and collect signatures against Vas

Hopes this time helps....

Sedat Canbaz
Posts: 3018
Joined: Thu Mar 09, 2006 10:58 am
Location: Antalya/Turkey
Contact:

Re: Why don't strong engines enter WCCC?

Post by Sedat Canbaz » Mon Aug 04, 2014 7:25 pm

Remember that too,
-Nobody can be on top by the work of others !

That means,
To be on top requires a lot of efforts, unique experience, own original ideas...!!

In other words,
Better own ideas than all, only then you can be on top !!!

PK
Posts: 863
Joined: Mon Jan 15, 2007 10:23 am
Location: Warsza
Contact:

Re: Why don't strong engines enter WCCC?

Post by PK » Mon Aug 04, 2014 7:58 pm

As for improving a derivative over an original: on a CCRL scale Sungorus 1.4 has 2309 Elo, Rodent 1.4 - 2790 Elo. Of course starting point was much lower (Sungorus had beautifully written, concise code, but no eval to speak of), but then I am much worse programmer than Vasik Rajlich, so these two factors might cancel out.

Tord Romstad
Posts: 1808
Joined: Wed Mar 08, 2006 8:19 pm
Location: Oslo, Norway

Re: Why don't strong engines enter WCCC?

Post by Tord Romstad » Mon Aug 04, 2014 8:01 pm

hgm wrote:Well, this is only partly true. Stockfish, for instance, could easily participate: its originality is not in question, the source is open and verifiable, and there would be plenty of people willing to operate it at their own expense.

It is just that some of its main authors for unfathomable reasons want to wage a personal vendetta against ICGA, or other organizers of over-the-board tournaments.
I can only speak for myself, but I don't have any sort of personal vendetta against ICGA, and I have played in (and immensely enjoyed) over-the-board computer chess tournaments in the past. The reason I no longer want to do so is that a few years back I arrived at the regrettable conclusion that although over-the-board tournaments are great when viewed as social events (online tournaments are terrible in this regard), the competitive side of computer chess does far more harm than good. I think the community would be better off with more focus on cooperation and less on competition. Excessive competition inevitably leads to cloning, accusations of cloning, suspicion, obfuscated node counts (whether to hide the engine's origins or to make it harder to figure out how it works), and endless flamewars, as can already be observed in this very thread.

I therefore consider it a good thing that many strong programs don't participate in tournaments like the WCCC. By making the tournament less prestigious, we maintain the positive sides of over-the-board tournaments (a rare opportunity to interact socially with people who share this unusual hobby), while eliminating or at least reducing the toxic effects of competition.

The decreasing number of participants, on the other hand, is a problem. I'm toying with the idea of participating with some experimental toy engine running on a mobile phone or smartwatch some time in the future, and I hope others will consider something similar.

Sedat Canbaz
Posts: 3018
Joined: Thu Mar 09, 2006 10:58 am
Location: Antalya/Turkey
Contact:

Re: Why don't strong engines enter WCCC?

Post by Sedat Canbaz » Mon Aug 04, 2014 8:32 pm

PK wrote:As for improving a derivative over an original: on a CCRL scale Sungorus 1.4 has 2309 Elo, Rodent 1.4 - 2790 Elo. Of course starting point was much lower (Sungorus had beautifully written, concise code, but no eval to speak of), but then I am much worse programmer than Vasik Rajlich, so these two factors might cancel out.
Nice...

Nowadays is not so hard to improve e.g 2000 Elo to 2300 Elo or from 2500 Elo to 2900 Elo,
Because there are many strong open source engines, where the programers can benefit...

Actually here I am talking about the 'number one' engines of 3300+ Elo,
Rybka, Houdini, Stockfish
I mean, this is a extremely special situation...!
And there is no doubt that those top mentioned engines reached best performance, due to better own original ideas than all !!!

Post Reply