Author: Mark Young
Date: 20:56:53 05/24/99
Go up one level in this thread
On May 24, 1999 at 23:19:04, Robert Hyatt wrote: >On May 24, 1999 at 20:46:40, Mark Young wrote: > >>On May 24, 1999 at 16:16:03, Robert Hyatt wrote: >> >>>On May 24, 1999 at 01:36:50, Mark Young wrote: >>> >>>>On May 24, 1999 at 01:13:09, Dave Gomboc wrote: >>>> >>>>>On May 23, 1999 at 23:45:29, Mark Young wrote: >>>>> >>>>>>On May 23, 1999 at 23:28:45, Dave Gomboc wrote: >>>>>> >>>>>>>On May 23, 1999 at 23:13:59, Mark Young wrote: >>>>>>> >>>>>>>>On May 23, 1999 at 22:44:04, Roger D Davis wrote: >>>>>>>> >>>>>>>>>It's only one game, certainly, but it doesn't really look like Rebel had much >>>>>>>>>understanding of the game from the time it left book (other's appraisals, not >>>>>>>>>mine), and Rebel's lose supports the argument that the micros have their own >>>>>>>>>weaknesses, and aren't yet a real match for GM players. >>>>>>>>> >>>>>>>>>So where does that leave previous opinions about a DB board for the desktop? >>>>>>>>>We'll have to see what the next months of Rebel-GM games hold, but the market >>>>>>>>>for a DB chip is starting to look better and better. >>>>>>>>> >>>>>>>>>If Rebel consistently loses to the GMs, doesn't this just set the market up for >>>>>>>>>the entry of Hsu? >>>>>>>>> >>>>>>>>>Roger >>>>>>>> >>>>>>>> >>>>>>>>This question must be answerd first: Does the new prototype Rebel 10.5 represent >>>>>>>>the current state of the art? Rebel 10.5 is untested, Rebel 10.5 has not shown >>>>>>>>itself to be in the very top class of programs as CM 6000 and Fritz 5.32 and >>>>>>>>Hiarcs 7. It is clear that Rebel 10.5 had problems (26. f4???)....but the other >>>>>>>>top programs have no problem understand the positions in (Ed's key moves shown >>>>>>>>below). >>>>>>>> >>>>>>>> >>>>>>>>17. e3? >>>>>>>>Blocking its pieces to go the king-side. After this move Ne3, Rg3 (h3) are no >>>>>>>>longer possible. 17.e3 was a last minute decision of Rebel10.5 it had 17.Qd3 >>>>>>>>all the time. Black answers with 17..f5 and now black is the boss on the king- >>>>>>>>side. >>>>>>>> >>>>>>>>*Hiarcs7 plays the correct 17. Qd3* >>>>>>> >>>>>>>We'll have to see what Ed can come up with regarding Rebel's last-minute switch >>>>>>>to e3. In general, I don't think that this game provides significant evidence >>>>>>>that Hiarcs is dramatically better than Rebel. In any case, it would be >>>>>>>foolhardy to make such a conclusion from one game. >>>>>> >>>>>>And no one did, other then to say Rebel 10.5 has not been shown to be the state >>>>>>of the art. Not like the top 3 I have named. And the current state of the art >>>>>>programs like Hiarcs7 plays the correct moves. >>>>> >>>>>The suggestion that Hiarcs 7 is state-of-the-art and Rebel 10.5 is not implies >>>>>that Hiarcs 7 is dramatically better than Rebel 10.5. Both or neither, please. >>>> >>>>What is hard to understand here, Hiarcs7 is a well know program, with many games >>>>and one of the top 3 programs on the ssdf list. Rebel 10.5 is a unknown with one >>>>game to its credit that I know about. How can we assume it is state of the art. >>>>Thus my question...does Rebel 10.5 Represent the Current State of the Art? I did >>>>not answer the question, only asked it. >>>> >>>>> >>>>>>>>21.Ra5 exd4 22.Rb5 >>>>>>>>Going for the pawn, nice manoeuvre but definitely wrong. Rebel's score dropped >>>>>>>>from 0.7 to 0.3 (or so) when it saw the danger of Qe2 but that was to few togive >>>>>>>>up the hunt for the pawn. >>>>>>>> >>>>>>>>*Hiarcs7 plays this, but so far has shown that after 22.. Qd8 23. Rxb4! wins. >>>>>>>>Not 23. exd4?! as played by Rebel* >>>>>>> >>>>>>>This is inaccurate. No win has been demonstrated whatsoever. Autoplaying from >>>>>>>a position doesn't prove squat. >>>>>> >>>>>>Until someone can bust the line it is proof, but not proven. >>>>> >>>>>It's not proof at all! You can't just take a position, play some reasonable >>>>>game from the position, and assess it based on the result of the game! >>>>> >>>>>I know chessplayers who do this in regular post-mortems too. It's pathetic. >>>> >>>>What is pathetic is a GM and other on CCC saying white is lost at move 20 >>>>without showing anything at all to back it up. I showed my results and lines and >>>>reasoning for other to look at and give me feed back on. That what I want! If it >>>>is wrong I want to know... I am not saying white has a clear win here, but said >>>>Hiarcs7 thinks and shows it has a clear win playing the other programs and >>>>against self play. I did not say Hiarcs7 is correct and this is proven. I want >>>>people to punch holes in the move if they can. If I did not, I would just say >>>>"White has a overwhelming position and black is lost at move 20". >>>> >>> >>> >>>I wouldn't call it 'pathetic'. IE why insult GM Rohde. He did win the game, >>>and all your analysis has shown is that Hiarcs can beat other programs with Bg2, >>>for example. Doesn't say a thing about Hiarcs beating him. Once classic >>>example was the final DB vs Kasparov game... Someone used a computer to play >>>the white side and it was able to draw or win against other computers, because >>>black _is_ a piece up. But an IM took black and no computer was able to win >>>a single game with white, losing most and drawing a few. IE perhaps the GM sees >>>a bit _more_ than the computers do there? His comment was "after the Ra5 >>>business it overlooked the fact that I got both files... and white was probably >>>lost at that point." >>> >>>That was all he said, in kibitzing analysis after the game ended with a bunch of >>>us looking on. >>> >>>Telling him "just because you say white was lost is pathetic, because when I >>>play Hiarcs (white) against other computers, it never loses..." I can only >>>imagine his "so what?, a computer wasn't playing black. I was..." >> >>If that what I did, I could understand that, but I gave moves, lines, reasoning >>why I think white may be winning, I have seen nothing from others who just took >>the word of I guess any GM that white is lost. No insult intended, but if you >>are going to make a claim like this or support that kind of claim and the >>position is far from clear you better give some reasoning to justify that claim >>other them a GM said its true. >> >>Hiarcs7 is my tool of choice at looking at the game, that what programs are used >>for, it keyed me on this move and I made the judgement it looks pretty good. >>This has been done many times by other amatures to find holes in GM analysis. If >>would be a mistake to just dismiss what programs help us fish to find out of >>hand. > >Hiarcs may be great. And if it finds a tactical win I would believe it. But >if it only sees non-tactical things, I would much prefer to trust a GM's >understanding of the position. > >But that is only an opinion. I would quote Kaidanov the other day, when we >were all observing some GM game in the Sarajavo tournament... Crafty was >giving some incredible tactical continuation, and someone said "GM Kaidanov, >would you give us some _real_ analysis to let us know what is going on, we can't >trust crafty..." He responded "If you don't trust crafty in this position you >are _crazy_. It is far stronger than I am when the position is open like this." > >With the implication that in more quiet/positional cases, it is not as reliable, >which I would certainly agree with... > >Maybe Rohde will give some detailed analysis at some point. But we aren't >goingto be able to force him to say anything... an old proverb about honey >and vinegar comes to mind... > > >> >>Now what has all this looking told me about the Rebel 10.5 - GM Rohde game? It >>has shown me that Rebel 10.5 played one mean game of postional chess. What was >>thought to be positional errors may instead be outstanding positional play that >>even a GM did not see or understand while playing the game. Lets look again at >>Ed's key moves and see what they might tell us. >> >>17. e3? >>Blocking its pieces to go the king-side. After this move Ne3, Rg3 (h3) are no >>longer possible. 17.e3 was a last minute decision of Rebel10.5 it had 17.Qd3 >>all the time. Black answers with 17..f5 and now black is the boss on the king- >>side.>>>> >> >>If this is a positional error, it is slight and should not cost the game. >> >>21.Ra5 exd4 22.Rb5 >>Going for the pawn, nice manoeuvre but definitely wrong. Rebel's score dropped >>from 0.7 to 0.3 (or so) when it saw the danger of Qe2 but that was to few to >>give>>up the hunt for the pawn.> >> >>This manoeuvering might not only be positionally correct, but possablly winning. >>Was not this the monoeuvre the GM's laughed at. >> >> >>24.Rxe8?? >>The main positional mistake in my opinion. After this move Rebel is definitely >>lost in all variations. >> >>No mistake here...it seems white can still draw, and is as good as 24 Qd1. >> > > >It is definitely a mistake. White had a pretty good position... but this was >all pretty much forced after Ra5. And with Ra5 he ceded the E-file and the a- >file. So there is little room for changing mind after the rook leaves the a >file. And if the rook isn't leaving the a-file, Ra5 is then foolish. > >Whether this _move_ is a mistake or not is an issue. But the point is, if >it is forced, it is forced because white made a mistake earlier. That forced >it to have to play this move which gives up a lot... > >> >>26.f4? >>>>I don't now why Rebel10.5 played this move. It's just a matter of search depth >>>>to see this move is losing material on short term. Rebel10.5 needs 12 plies >>>>to see the loss and then changes to 26.Bg2 (the good move). I tested the >>>>current REBEL-CP on this position and nowhere in the log I see the ugly >>>>f4?, >> >>Positional play did not kill rebel 10.5 but a bug?? or something wrong in its >>program, why else play 26. f4 that loses by force and that other program >>including Rebel 9 see. This *bug* may have also cost rebel the win if this >>caused rebel to miss 24 Rxb4!, and for almost sure the draw if it had played 26 >>Bg2!. >> >>Tactics not positional play killed Rebel 10.5, and I see only the need to fix >>the tactical bug in Rebel's programming. This is why it is dangerous to take >>just the word of a GM without proof, thus thinking that your program played >>positionally stupid. >> > >Nothing says Rb4 wins yet. Nothing is a pretty strong word, I know some very good chess players along with some very strong chess programs that disagree with the word nothing. It is not importent to me if GM Rohde ever comments on this game, there are other very strong players that can look at the game, and I am sure they will. And nothing really says f4 was a bug, as other >programs like f4 for a while and then change their mind. This is only speculation on my part, I know other rebel's see that f4 is wrong, and play the correct move. Ed is the only one that can answer this for sure...does Ed's comments about why Rebel 10.5 played f4 hint that he thinks its a bug?...im not sure. Question is, would >they have had time to find that f4 is bad in the game... > > > >> >>> >>> >>> >>> >>>>> >>>>>>I played many program against this position, and myself have tried to bust the >>>>>>move, all attempts has lost to 24. Rxb4. >>>>> >>>>>Show the attempts and their refutations, then it will be more like a proof. >>>>>Concrete analysis for a variety of Black tries and responses by White is far >>>>>better than an autoplay result. >>>>> >>>>>>I said nothing inaccurate here, I did not say it proved anything...other then >>>>>>Hiarcs7 has shown this and thinks it to be a win. So this move must be shown >>>>>>wrong to prove the GM Rohde is right that the position is won by move 20. So far >>>>>>I don't see it, other then to see a win for white and for GM Rohde. >>>>> >>>>>The inaccuracy referred to is the claim that 23. Rxb4 is clearly winning. With >>>>>some backup variations, it might become clear, but it is far from it now. >>>>> >>>>>>>I currently think that White is not only alive and well after Rxb4, but probably >>>>>>>even better. However, "white is winning" is another matter entirely. >>>>>>> >>>>>>>>24.Rxe8?? >>>>>>>>The main positional mistake in my opinion. After this move Rebel is definitely >>>>>>>>lost in all variations. It gives away the crucial e-line. IMO Rebel should have >>>>>>>>played 24.Qd1 >>>>>>>> >>>>>>>>*Hiarcs7 plays the Ed's correct move 24. Qd1, but I have seen no line that shows >>>>>>>>24. Rxe8?? is losing.* >>>>>>> >>>>>>>Qd1 and Rxe8 lead to roughly (or exactly) the same position, don't they? >>>>>> >>>>>>I agree and said so. >>>>>>> >>>>>>>>26.f4? >>>>>>>>I don't now why Rebel10.5 played this move. It's just a matter of search depth >>>>>>>>to see this move is losing material on short term. Rebel10.5 needs 12 plies >>>>>>>>to see the loss and then changes to 26.Bg2 (the good move). I tested the >>>>>>>>current REBEL-CP on this position and nowhere in the log I see the uglyf4? >>>>>>>> >>>>>>>>*Hiarcs7 plays the correct 26. Bg2!!, and no line shown so far has been able to >>>>>> >>>>>>>>prove a win for Black after the correct 26. Bg2* >>>>>>> >>>>>>>Dave >>>>> >>>>>Dave
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.