Rybka engine is geared towards positional play when analyzing and suggesting candidate moves, that is why it takes it a longer time and put tactical calculation into a deeper threshhold. that is what i learned from someone who prefer rybka than fritz! while most chess programs are set to find and calculate tactically, rybka was designed to be the opposite.
Rybka didn't find this move till 16 DEPTH!!
Sort:
Oct 10, 2009
0
#23
zxzyz wrote:
All the engines i have running on scid get it:
Toga2 about 10s ..rybka 2.2 about the same , stockfish1.5 in about 5 s.
Ok since I am wasting time now heres more:
crafty 2.3 also gets it in a few s. it says nxf7!
The best response for black seems to be nc5 instead of taking.
I have a quad processor though..
Congrats anyway .. I like that combo ..
Whats your rating
I don't have an official rating in USCF. But in my chess club I'm the second best player with the best being a 2400 rated guy who I have yet to beat. I have drawn a game with him though the other guy in our club who is good is rated 1800's 1900's USCF and I've got a winning record against him in 40/5 timed games.
My best performance is beating a life master (2 years ago) at a chess party in Los Angeles. I know nerds right? He played Alekhines Defense, told me I had talent for the game and needed to study theory. My thought is he let me win to test me.
Oct 10, 2009
0
#24
yeah that is a famous game...Life Master AJ has it on is site and really good annotation.
hehe immortal you're being treated like a god ;)
I must sympathize with black though - the smith morra's pretty annoying. Even after gambitting the pawn white is still winning by a fraction of a pawn - so black can barely say that he has "material advantage". But here's the worst part: when white "equalizes" in material, 99% of the times it doesn't mean that black has gave back a pawn for a winning attack, but rather the opposite: White's attack has crushed through and the pressure has toppled the roof.
gg by the way. Sorry about my rant on the smith morra. I hope I didn't hate on it too much =)
Oct 10, 2009
0
#26
No worries Shuttle. BTW everyone shuttle is a guy who I believe will be a master one day. His opening knowledge is already quite large, and his instincts in openings he doesn't even know blow me away ...example: My game against him using the Latvian Gambit. He'd never seen it before and still played all the current book moves. Amazing!
agreed nezmendiov is so deep
PerfectGent wrote:
you only gave it time to reach 16 ply. This is not enough for any engine to reach a correct analysis. In game play within the advanced group (centaur chess, engines allowed) discussion has decided that a minimum depth to reach a good move is 21 ply. I believe that the chess.com analysis only goes to 15 ply which leads to some discussion about its results.
depending upon your computers strength (processor power) then a suitable depth of analysis can take from hours to days.
You use a lot of subjective terms here. What is meant by "correct" analysis. Unless you've solved the game for that position, any result has the potential for being incorrect.
What was the typical search depth for Fritz when it defeated Kramnik? How about Rybka when it's winning it's matches. These searches don't take hours to days, they take minutes. And they are suitable for winning very high level games.
When I analyze my own games, engines find mistakes at 10-12 ply all the time. And I rarely play a good move that the engine doesn't also evaluate as good within 10-12 ply or maybe 13-15 ply. And this I think is more Immortalgamer's point. As an amateur chess player, finding a strong move at non-correspondence time controls that the computer misses until depth 16 is somewhat rare.
Sure, if you play centaur chess at correspondence time controls a depth 16 engine gets blown off the board. I fail to see how that's relevant to Immortalgamer's post about a live chess game between humans.
Yes, I agree with Loomis there. Depth 15 is more than fine for the extremely large majority of positions.
I believe chess.com uses Time limiting rather than Depth limiting on its analysis, which is why in complicated situations it simply misses moves.
immortalgamer wrote:
No no...it didn't take 2 hours. that is just what I gave it to think. It didn't come up with Nxf7 till 16 depth
Nice move.
Actually it came up with N:f7 right away. If it wasn't showing up in the variations in the analysis pane prior to depth 16 it is just because the evaluation of the line wasn't high enough. It was thinking about it at every depth though.
It is definitely a good feeling though when you come up with a move that the engine takes a while to decide is a nice one, especially when the evaluation makes a nice jump.
AnthonyCG wrote:
Your friend must have been very quiet after Nxf7.
Sicilian players are a funny bunch sometimes.
They'll call anti-sicilians weak, but really hate to play against them.
Wonder why.
Hehe, I may be an exception, because I'm a Sicilian player and I enjoy playing against pretty much all of the anti-Sicilians. I play the Morra myself, so I know how to combat it; the closed Sicilian systems with Nc3 and g3/f4 I find pleasant to play against, and I find it easy to get a free and easy position to play against the Alapin.
It's one of the reasons I love the Sicilian - the open is great fun, and I enjoy playing against all the rest of the systems.
PerfectGent > fair point - maybe i should have said 'best' analysis.
So your claim is that the best analysis is at depth 21? What's wrong with depth 23, it somehow gives worse analysis than depth 21?
PerfectGent > old m/c was a 1.3Ghz proc with 512M ram it took 24 hours to reach 21 ply
new m/c is quad 2.gGhz proc with 4G ram and takes 2 hours to reach the same 21 ply.
so an octal processor 64 bit m/c with 16G ram will reach the same ply in minutes.
Care to indicate how you come up with the final number? Looks like you just grab it out of thin air. I don't think it's worth quoting a time per depth stat without an actual measurement of it. (Improvements with increasing RAM and number of processors/cores aren't linear, the returns are diminishing.)
PerfectGent > If you find arrors at 10-12 ply then i am afraid this just shows the low level of your game. (nothing personal or derogatory in this observation)
I have no illusions about the level of my game. The federation is very good about assigning numbers that reflect this. :-). On the other hand, our e-chess ratings on chess.com are not too different. I doubt anyone at our level can consistently play OTB chess without making mistakes that a computer will find in less than 12 ply depth. Given this fact, an engine can easily make the correct analysis that my move was a mistake in under 16 ply. Contrary to what you claim in your first post (#28).
In fact, I'd be surprised to find that all OTB master games contain no errors that a computer can't find in 12 ply or less. So, don't worry, I take no personal offense in your observation. ;-)
PerfectGent > my response was nothing to do with the time limits of the game analysed.
Sorry for my misunderstanding, I didn't realize you were going off topic.
Hi Loomis,
I think you're being a bit harsh on PerfectGent there.
He chose depth 21 because that's his (and the impression he gets from others) estimate of when the computer can truly come up with the 'best' possible move. Obviously further depth can only be beneficial.
He is grabbing the figure for the oct-quad somewhat out of the air, but you get his point - he illustrated the massive difference between his old and new computer, so he's simply saying that on a computer that's much more powerful than his current one, it'll reach the same depth much quicker again.
Anyhow, I'm roughly 2100 ELO standard and I've played 3 games this season, in all 3 games I have made at least a couple '10-12 ply' mistakes. You have to be really very good indeed not to make these mistakes in your games.
He chose depth 21 because that's his (and the impression he gets from others) estimate of when the computer can truly come up with the 'best' possible move. Obviously further depth can only be beneficial.
I wasn't trying to be harsh, and I know this will come off as even harsher, but please check your dictionary for the meaning of the word "best". And then quit saying stuff like "this is the best, but other things are even better."
If you want to claim there is a time constraint (this is true even in analysis since life is time limited) and the additional time used allowing the engine to search further than 21 ply is not worth the possibility of finding better moves, then make that argument. Of course in that case you have take into account hardware.
Oct 23, 2009
0
#35
"So your claim is that the best analysis is at depth 21? What's wrong with depth 23, it somehow gives worse analysis than depth 21? "
Just my view, and I'm not trying to put words into Perfect Gent's mouth, but obviously depth 100 > depth 50 > depth 23 > depth 21. The farther it looks the better. I'm just assuming that a cut-off on depth 21 is based on time. 21 is probably like an hour or two, so he might have just been saying that he likes to let the engine consider it for at least an hour.
Also, in my experience with engines, once you get up to a certain point, getting that extra ply is extremely tedious. Could take hours.
bondiggity, that's essentially what I was suggesting that he might want to argue in the second paragraph of my post right before yours.
Oct 23, 2009
0
#37
immortalgamer wrote:
WOW!!!! You make it sound like BIGGG!!!! deal. Rybka 3 takes less then 1 second to see Nxf7.
PerfectGent, I think anyone can see that I responded to what you wrote and you have resorted to attacking me personally (insinuating I can't understand words of more than one syllable).
You're back to using subjective terms. What do you mean that below 21 ply engine results are not reliable? A 19 ply depth search will reliably beat a 12 ply search. A 21 ply search will be unreliable (lose to) a 23 ply search.
Amateur chess players can rely on top engines to find improvements in their OTB/live games way below 21 ply. Any amateur player will find this out in a few minutes, if you're not making any mistakes at 17 ply, you're playing at a very strong master level.
If your point is that finding moves that an engine misses at 15 ply is not interesting because it's well known that engines are only "reliable" at 21 ply or greater, I think you're missing the fact that this was a live human game and not a centaur correspondence game.
If your point is that Immortalgamer is not properly using the engine to analyze his game because he didn't let it go to at least 21 ply, then I think you're missing the point Immortalgamer is making. I think his original point is that the move Nxf7 requires a minimum search depth of 16 ply to evaluate it correctly and that given typical human errors in live games finding a tactically rich combination that requires this depth is interesting.
In your original post (#28) you claim that depth 16 is not enough for an engine to make a correct analysis. But cases where a player makes a mistake that an engine sees in 12 ply is a simple counter-example to that. And no matter how much you'd like to disparage me as a player, these mistakes are common at my level (as well as above my level to some degree). I hope that helps you follow my reference to your initial post.
Of course you're not in a restricted zone. If you want to go off topic and discuss engine seach depth and confidence of results, go ahead. It just confused me because I thought you were responding directly to the original post.
Oct 23, 2009
0
#39
philidor_position wrote:
immortalgamer wrote:
WOW!!!! You make it sound like BIGGG!!!! deal. Rybka 3 takes less then 1 second to see Nxf7.
Ummm.. Yeah Rybka 3 is rated around 3100. So yes I think it is a big deal. I think I just have an old slow computer and this is why it took till 16 depth. I was excited about it and if you don't mind, since you have Rybka 3. Could you let it run on the position for 10 mins. and see if there is a deeper "better" move that I missed?
Why the negativity toward my post? Don't understand people like you?
Oct 24, 2009
0
#40
immortalgamer wrote:
philidor_position wrote:
immortalgamer wrote:
WOW!!!! You make it sound like BIGGG!!!! deal. Rybka 3 takes less then 1 second to see Nxf7.
Ummm.. Yeah Rybka 3 is rated around 3100. So yes I think it is a big deal. I think I just have an old slow computer and this is why it took till 16 depth. I was excited about it and if you don't mind, since you have Rybka 3. Could you let it run on the position for 10 mins. and see if there is a deeper "better" move that I missed?
Why the negativity toward my post? Don't understand people like you?
Sorry, I just don't appreciate use of multiple exclamation points and capital letters on forums. I find it the same with just raising your voice in the middle of a discussion with no good reason.
By the way, depth and slow computers are usually not related to each other. On a slow computer, it just takes more time to reach depth 16, but the variation should be same with a faster computer that reaches it. (unless it's a multiprocessor version of the engine where the evaluation is not determinate, but still should be similar). So on a fast computer, the same version of rybka also would need to reach depth 16 to see the move.
Here's the Rybka 3 Human analysis around 10:30.
immortalgamer - NN, Online Chess 2009
r1b2rk1/1p1nbppp/pn2p3/4N3/8/1BN5/PP1B1PPP/3RR1K1 w - - 0 1
Analysis by Rybka 3 Human:
15.Nd3 h6 16.Bf4 a5 17.h3 a4 18.Bc2
= (-0.11) Depth: 8 00:00:00 0kN
15.Nd3 h6 16.Bf4 a5 17.h3 a4 18.Bc2
= (-0.11) Depth: 9 00:00:00 0kN
15.Nxd7 Nxd7[] 16.Bf4
= (-0.13) Depth: 10 00:00:00 61kN
15.Nxf7
= (0.09 !) Depth: 10 00:00:00 81kN
15.Nxf7
+/= (0.29 !) Depth: 10 00:00:00 94kN
15.Nxf7 Nc5 16.Ng5 Nxb3
+/= (0.41) Depth: 10 00:00:01 152kN
15.Nxf7 Kxf7 16.Rxe6 Nc5
+/= (0.49) Depth: 11 00:00:01 226kN
15.Nxf7 Kxf7 16.Rxe6 Nc5 17.Rxb6+ Nxb3[] 18.Rxb3 Rd8 19.Ne4 Bf5 20.Rf3 Kg6 21.Rg3+ Kf7 22.Re3 Rac8
+/= (0.39) Depth: 12 00:00:02 437kN
15.Nxf7 Nc5 16.Ng5 Nxb3 17.axb3 Rf6
+/= (0.51) Depth: 13 00:00:08 1362kN
15.Nxf7 Nc5 16.Ng5 Nxb3 17.axb3 Rf6
+/= (0.51) Depth: 14 00:00:09 1551kN
15.Nxf7 Nc5 16.Ng5 Nxb3 17.axb3 Rf6 18.Re4 Rg6 19.h4 h6 20.Nf3 Nd7 21.Be3 b5 22.Red4 Nf8
+/= (0.63) Depth: 15 00:00:15 2374kN
15.Nxf7 Nc5 16.Ng5 Nxb3 17.axb3 Rf6 18.Re4 Rg6 19.h4 h6 20.Nf3 Bf8 21.Ne5 Rf6 22.Rg4 Nd5 23.Ng6 b6 24.Nxf8
+/= (0.52) Depth: 16 00:00:19 3073kN
15.Nxf7 Nc5 16.Ng5 Nxb3 17.axb3 Rf6 18.Re4 Rg6 19.Nf3 Nd7 20.Ree1 b5 21.Bf4 Bb4 22.h3 Ba5 23.g3 Rf6 24.Ng5
+/= (0.53) Depth: 17 00:00:36 5715kN
15.Nxf7 Nc5 16.Ng5 Nxb3 17.axb3 Re8 18.Nxe6 Bd6 19.Ng5 Rxe1+ 20.Rxe1 Bf5 21.Nge4 Bb4 22.f3 Rd8 23.Be3 Nc8 24.Nc5 Nd6 25.Bf2 Rc8
+/= (0.65) Depth: 18 00:01:46 17553kN
15.Nxf7 Nc5 16.Ng5 Nxb3 17.axb3 Re8 18.Nxe6 Bd6 19.Nd4 Bd7 20.f3 h6 21.Ne4 Bf8 22.Bf4 Bb4 23.Re2 Nd5 24.Bd6
+/- (0.73) Depth: 19 00:06:39 88205kN
15.Nxf7 Nc5 16.Ng5 Nxb3 17.axb3 Re8 18.Nxe6 Bd6 19.Nd4 Bd7 20.f3 Bf8 21.Be3 Rad8 22.g4 Bc8 23.Bf2 Rxe1+ 24.Rxe1 Nd5 25.Nxd5 Rxd5 26.Re8
+/- (0.72) Depth: 20 00:08:29 108mN
Nifty game! You are above 2100. You don't think you could reach master?
Let me clarify. I think I could reach master if I had the time to study everyday. Since my chess study mainly consists of 10 1 min games and maybe 1 standard game per day: I don't think I will.