> The consensus a few years ago was 150 Elo. This was expected to decline over time. I suspect that its still 100 Elo or so, which would qualify as significant in my book!
I agree that 100 Elo is significant. It's just that I am not a part of that consensus.
I was talking about cases where the hardware used is equivalent, so to a certain extent we may be talking past each other. I assume the case you have in mind is one where a freestyle player (or team) has more than one computer?
It would be interesting to see how well one could make use of this as a centaur. But 1) I would be surprised if one could turn this advantage into 100 Elo points and 2) given some practice, I think a very strong player will never be worse at this than a weak player.
I'm curious to hear your view on the following: Do you think that the decision process of centaur teams could in principle be automated?
I would guess that the advantage of the centaur under the conditions above would still be 100 Elo or so. I believe the strong player has an advantage all else being equal, but all else may not be equal. There are team and database management and bookkeeping issues that come into play, and the strong player may not have these skills.
In theory, the process could be automated (except for the part where two or more moves give similar engine eval, and the tie is broken by someone with a great degree of chess knowledge). This has not been done to date because it would be a lot of effort for little return.
> In theory, the process could be automated (except for the part where two or more moves give similar engine eval, and the tie is broken by someone with a great degree
> of chess knowledge).
I am puzzled by the part in brackets. What about good old-fashioned interactive analysis? Is there no place for it in a freestyle team?
> A second would be pressing interactive analysis on lines that the engine doesn't like. This second rational is becoming less frequent, but it's comforting to think it's still of some importance...
But I don't think very high OTB strength is needed for that, with my 1400 elo I've managed to force my ideas into the engines to show that they're better, or at least equal, to their ideas, specially on positions they have trouble with (when there's nothing to do and they just shuffle pieces around), positions in where they don't find the winning moves (e.g. if they trade down the Bishop pair for no reason or go into an opposite Bishop ending), or positions where there's a thematic key move that they showed me before in a different similar variation, but that they don't find in this one, but also works.
Discussion here makes it sound like GM strength is needed for great influence, but I think years of experience with chess engines can replace that even if understanding of what is going on at OTB level is not the same.
(this is about corr chess but I think the general ideas apply)
The question for you would be:
Would it be more difficult for the GM to learn your techniques, or for you to add 1200 Elo?
This is a much different question than: 'Do top GMs know better methods of analyzing with engines?', which Harvey alluded to above (which I am very skeptical about).
And i can give you another example about the current issue
I think the bellow single-processor engine is around on the same level as Top GM of 2700-2800 Elo human points,right ?:
Chess Tiger 2007
So...in other words,if we will start to test Chess Tiger 2007 engine against Rybka,Houdini,Critter
(i mean if we run the Top MP engines on latest i7 6 core or 12 core machines via Auto232 mode or via online)
And after a such test at 40/120,can we expect the Elo difference to be less than 500 Elo ?
For example,Auto232 engine match between Houdini 2.0c x64 6 core against Chess Tiger 2007 w32 1c
Lets take as example SSDF rating,which is very useful for a such comparison:
Deep Rybka 4 x64 2GB Q6600 2,4 GHz 3216 32 -29 642 78% 3001
Chess Tiger 2007 256MB Athlon 1200 MHz 2786 22 -23 966 39% 2862
Note:the current SSDF Elo difference is 430 Elo,what will be the Rybka's MP Elo difference on 3 times faster machine ??
For example, if Rybka,Houdini Critter will be played on decent fast hardwares,e.g on i7 990X @ 4.60GHz
Note that in case of running MP engines, i7 990X @4.60 GHz is approx.3 times faster than Q6600 2.40GHz
That means you will get extra approx.130 Elo points
In other words,i expect the SSDF Rating to be:
1.Houdini 2.0c Pro x64 6 core i7 990X @4.60 GHz 3400 Elo
2.Deep Rybka 4.1 x64 6 core i7 990X @4.60 GHz 3350 Elo
3.Critter 1.4 x64 6 core i7 990X @4.60 GHz 3350 Elo
4.Chess Tiger 2007 256MB Athlon 1200 MHz 2786 Elo
Btw,SCCT Auto232's rating has almost same Elo points :)
I really wonder too,in case of participating the Top Human GM in SSDF Rating (e.g Kasparov,Kramnik,Carlsen,Topalov,Anand... )
Probably then, i expect the Top Human Players would be rated around 2700-2850 SSDF Elo points !
Even more than 10 years ago,we noticed that the top engines are started to play as GM levels,here is another proof:
-The above 'Red' table (based on around 2000 years) does not include Hydra chess engine
-Now we are in 2012 that means the current top engines play much stronger than 2000 years engines
-Many of the Top Engines have been improved to play stronger approx.200-300 Elo points (in that period of time)
-Plus 10 yeas ago,the processors were much slower... .e.g in those years,Fritz Benchmark kns values were around 500-1000 kns
-But nowadays, (on latest fast i7 machines) Fritz Benchmark tool generates around 20.000 kns values
Check again the results please
It does not matter,take any other engine (instead of Chess Tiger 2007) which is rated about 2700-2850 Elo points (based on SSDF,CCRL,SCCT... ratings)
The Elo performance is expected to be rated almost same level as Top GMs
Just i'd like to mention again that 10-12 years ago, the top engines elo performance were around 2650-2750 Elo points
Nowadays,exactly the same engine versions are improved to play stronger at least 200-300 Elo points
In 10 (ten) years period of time,the processors are become at least 15-20 times more faster,that means you will get extra at least 350-400 Elo points
Just imagine in 2012...what will be Elo difference between Top MP Engines vs GMs
Personally i expect to see approx.500-600 Elo difference
The best answer:its will be great,if there will be a serious mach (played with many games) -Man vs Machine 2012
And its will be a BIG surprise for me, if the Top Engines will be performed bellow than 500 Elo
I would think it would be pretty easy to gain some insight into Hydra's strength. Just run a simple test. Have two machines.. One with Shredder 8 running at an average of over 2000kNps (which is the equivalent speed used against Hydra in 2004), and the other machine with a modern engine (Rybka, Stockfish, Houdi, Komodo, Strelka, Fritz, Shredder 12 etc take your pick) running at a mere fraction of the power that fueled Hydra's algorithms. Make a pgn file with all eight of the Hydra-Shredder games stripped all moves after Shredder's last book move in each game. Then put whatever modern engine you want to compare with Hydra's performance, in Hydra's place, starting out as white after 17...Qa1+. Ponder ON for both engines. => Now you get to see how a modern engine performs in comparison to Hydra's performance against Shredder 8. I would think that if the modern engine has Shredder 8 showing higher negative scores earlier than it did against Hydra in games 1, 2 and 7, and if the modern engine manages to win some of the games which Shredder was able to draw against Hydra, then it's probably reasonably strong evidence that the modern engine is stronger than Hydra.
-The standing is based on Computer vs Human GM games, played during 2004-2005-2006 years
-For a better conclusion more games is needed,but however i am impressed by the performance of the Top chess programs
-Even 7-8 years ago,(when the processors were much slower and the engines were much weaker )the Top engines performed approx.200 Elo better than the Top GM
-The current Top MP Engines are expected to be performed (on latest fast hardwares) at least 500 Elo points stronger than the Top Human Players
For more details:
Argh, that nice game Rebel played against Anand and then ruining a won position by capturing too early on "b2" thinking it had a won ending.
All well today
00:00:29 15.00 2.33 1..Bxb2 2.Bxb2 Rxb2 3.Qxb2 Rxb2 4.Kxb2 Qxh5 5.Rc1 h6 6.d7 Qe2 7.Ka1 Qd3 8.Bh3 Qd4 9.Kxa2 Qd5 10.Kb2 Qb5 11.Kc2
00:00:32 15.01 2.33 1..Qe6
00:00:47 15.01 3.29 1..Qe6 2.Bc6 R3b4 3.Re1 Rc4 4.Qd2 Qf5 5.Be4 Rxc1 6.Rxc1 Qxe4 7.d7 Rd8 8.Qb4 Qd5 9.Qe7 Qxd7
>Argh, that nice game Rebel played against Anand and then ruining a won position by capturing too early on "b2" thinking it had a won ending.
Agreed...it seems GM Anand was too lucky (against 2000 years machines)
Note:i mean those calculations (where ChessTiger 2007.1 is rated at 2550 Elo) are wrong, which are based on Shredder 12 1c's 2800 Elo
But however, CEGT, SWCR, Clemens (which are based on Shredder 12 1c's 2800 Elo) are doing great job -BIG thanks for their works,efforts... !
Since we have actually a veritable GM Larry, whatever that means in human terms because he declares even the ipon list reliable without any reported game scores, we can further find arguments in Sedat et al who help to keep some mummies alive, e.g. Virtual Chess and CSTal, who are basically on 2900 level too if we recognize it in human terms and a free definition of Shredder on a fixed Elo in human terms of 2800 Larry terms wise. To my surprise I discovered that Junior as ex-WCh could well be lifted on Elo 4500 human terms Kasparov wise.
Overall it was an ethically brilliant move to outlaw Rybka because how silly all these attempts would look if Rybka had continued to win all the tournaments what would result in a top score of (Larry human terms wise) of 7120! That would be absolutely unacceptable for Fabien, SMK or Christophe Th., also considering the depressing sales quotes right now.
P.S. It's a must to mention the forgotten SSDF side by side with Ipon at least 5 times a week. Next task: Who is stronger Hydra, Fritz or Rebel or Komodo MP? Is BELLE still competing?
>Since we have actually a veritable GM Larry, whatever that means in human terms because he declares even the ipon list reliable without any reported game scores, we can >further find arguments in Sedat et al who help to keep some mummies alive, e.g.
the IPON list is probably the most reliable rating.
The IPON list has the best conditions regarding uniformity of opposition, books, hardware etc.
No,no,no... i have the best rating list-40/40:
Can anybody prove about my rating list is best (true or not) ???
Rank Name Elo + - games score oppo. draws
1 Houdini 2.0c Pro x64 6c 3443 21 21 10000 61% 3380 53%
2 Critter 1.4 x64 6c 3380 22 22 10000 49% 3384 59%
3 Deep Rybka 4.1 x64 6c 3366 16 16 10000 60% 3309 55%
4 Stockfish 2.2.2 JA x64 6c 3364 23 23 10000 52% 3353 58%
5 Ivanhoe 46hm x64 6c 3356 22 22 10000 50% 3358 66%
6 Robopolito 0.10 x64 6c 3355 23 23 10000 58% 3304 62%
7 Fire 2.2b xTreme GH x64 6c 3344 23 23 10000 53% 3327 65%
8 Vitruvius 1.0C HEM x64 6c 3334 24 23 10000 56% 3297 54%
9 Naum 4.2 x64 6c 3296 29 29 10000 54% 3267 46%
10 Strelka 5.1 x64 1c 3248 15 15 10000 56% 3212 52%
11 Chiron 1.1a x64 6c 3235 27 27 10000 48% 3247 50%
12 Deep Shredder 12 x64 6c 3233 26 26 10000 50% 3236 44%
13 Deep Fritz 12 w32 6c 3226 25 25 10000 45% 3254 49%
14 Hiarcs 13.2 w32 6c 3223 28 28 10000 50% 3222 44%
15 Deep Junior 13 x64 6c 3222 28 28 10000 50% 3225 48%
16 Protector 1.4.0 x64 JA 6c 3212 31 31 10000 52% 3202 41%
17 Spike 1.4 Leiden w32 6c 3200 25 25 10000 40% 3263 44%
18 Spark 1.0 x64 6c 3196 25 25 10000 44% 3234 50%
19 Komodo 3.0 x64 1c 3194 12 12 10000 50% 3169 39%
20 Zappa Mexico II x64 6c 3171 26 26 10000 43% 3219 44%
> Can anybody prove about my rating list is best (true or not) ???
It can't be proven because you don't provide any games, just like the IPON!
SCCT Hardware Tournament (same conditions,exception hardware speeds,see the Elo differences):
SCCT - Ponder OFF/ON (same conditions,exception Ponder OFF/ON,see the rank differences):
Hope this helps...
It's great to hear what you and Don are doing and the confidence you have in komodo. Are you expecting to release now in March?
> Probably we will release next version about a week after I get a debugged MP version.
Ah ok, fair enough. I was under the impression from some comments at Talkchess that you were going to wait until after the Peter Skinner Tournament though.
By the way, when you did your recent opening book, how many cores did you have for the IDeA stuff? I've got a poxy 8 and it takes ages and I'm not letting it think for as long as you did.
> CEGT's credibility has gone south of cheese!
I love CEGT and I understand their decision but it needs a corrector of +100 elo I think.
One thing I hate about the engine is that it's oblivious to kingside attacks and blockades as you can see here where it allows f5 opposite an amatuer... when it must exchange on f4 as rybka knows...
There was another win opposite H2 which is not there anymore were the same trick was used.
Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill