Author: Uri Blass
Date: 15:29:11 06/12/05
Go up one level in this thread
On June 12, 2005 at 17:23:49, Christopher Conkie wrote: >It is the engine that must play chess not the GUI. > >There are for example positions that are legal as well where engines have no >concept of what is going on. > >Uri posted one below. I know of many such secret positions both legal and >illegal. I have collected them for years. I cannot say them to you or they will >be countered and against. > >We have a problem with clones as you know and these positions can tell us how an >engine behaves with both via willingness to play AND output. > >They are truly fingerprints of what the engines are and are not capable of. > >For example Uri's position is...... > >[D]K7/8/8/8/Q7/1k6/B7/8 b - - 0 1 > >In this position there is a simple capture to make. Movei does this...... > >FEN: K7/8/8/8/Q7/1k6/B7/8 b - - 0 1 > >Movei 0.08.295: > 1 00:00 62 62 +12.89 Kc3 > 2 00:00 163 163 +13.19 Kc3 > 2 00:00 263 263 +13.36 Kc3 Kb7 > 2 00:00 265 265 +13.36 Kc3 Kb7 > 3 00:00 917 917 +13.39 Kc3 Qd1 Kb4 Qd4+ Ka3 Qc3+ Ka4 > 3 00:00 1.394 139.400 +13.39 Kc3 Qd1 Kb4 Qd4+ Ka3 Qc3+ Ka4 > 4 00:00 2.062 206.200 +13.69 Kc3 > 4 00:00 3.592 359.200 +13.77 Kc3 Kb7 Kd2 Kc6 > 4 00:00 3.600 360.000 +13.77 Kc3 Kb7 Kd2 Kc6 > 5 00:00 16.091 402.275 +13.77 Kc3 Kb7 Kd3 Kc6 Kd2 > 5 00:00 16.105 402.625 +13.77 Kc3 Kb7 Kd3 Kc6 Kd2 > 6 00:00 16.179 404.475 +14.07 Kc3 > 6 00:00 48.256 438.690 +14.10 Kc3 Kb7 Kd3 Kc6 Kd2 Kd5 > 6 00:00 48.310 439.181 +14.10 Kc3 Kb7 Kd3 Kc6 Kd2 Kd5 > 7 00:00 151.552 473.600 +14.10 Kc3 Kb7 Kd3 Kc6 Ke3 Kd5 Kd2 > 7 00:00 151.674 473.981 +14.10 Kc3 Kb7 Kd3 Kc6 Ke3 Kd5 Kd2 > 8 00:01 513.758 471.337 +14.26 Kc3 Kb7 Kd3 Kc6 Kd2 Kd5 Ke1 Bc4 > 8 00:01 514.627 472.134 +14.26 Kc3 Kb7 Kd3 Kc6 Kd2 Kd5 Ke1 Bc4 > >And so on forever...... > >The point is that these kinds of positions tell us something about the makeup of >engines. How it works, in other words. > >They tell us the minutest details. There is no escape. I in recent months have >become aware of more areas of chess programming that I can bear to relate here. > >I have looked at so much output you would not believe. I know what is inside >them. > >If you want something done, sometimes you have to do it yourselves. That is why >we made the CDS Suite. We made it as a way to catalog all known engines in every >format. > >Every few months another new engine/version comes along and that is what we >have, and continue to, catalog. We have a vast internet and when you are >interested as much in AI and computer chess as I and the people I know are, you >have a huge pool of resources at your disposal. We are all collectors you see, >as well as being interested. I can go back in time to the very beginning of the >Winboard and UCI protocols that are the accepted ways today as well as forwards >to new protocols, like those being developed with regard to FRC and Chess960. > >The engines in each given environment behave certain ways. It's a fact. > >It is a very hard thing to make a good clone, so much so that one is better off >writing ones own code from scratch. > >If the commercial chess environment is every to recover then they must realise >that a strong amateur environment is required. > >Clones don't breed a strong amateur scene. New and original ideas are what lead >to a strong amateur environment. > >This is the remedy we have been waiting for and we like it. > >Sorry for banging on but you must understand that we are good people and that we >are sick of being f----d about. That is why we made this. It is also therapy for >the chess mind as well. I am truly interested in it. > >Look..... > >I let Movei running...... > >FEN: K7/8/8/8/Q7/1k6/B7/8 b - - 0 1 > >Movei 0.08.295: > 9 00:03 1.623.998 474.853 +14.25 Kc3 Kb7 Kd2 Kc6 Ke1 Bc4 Kf2 Kd5 Kg3 > 9 00:04 1.632.202 474.477 +14.25 Kc3 Kb7 Kd2 Kc6 Ke1 Bc4 Kf2 Kd5 Kg3 > 10 00:04 1.956.280 434.728 +14.55 Kc3 > 10 00:04 1.961.593 435.909 +15.25 Kc3 > 10 00:06 2.938.692 451.411 +49.91 Kc3 Qd1 Kb4 Kb7 Kb5 Bc4+ Ka5 Qc2 Kb4 Qb3+ >Kc5 Qb6+ Kxc4 > 10 00:07 3.397.930 452.454 +49.91 Kc3 Qd1 Kb4 Kb7 Kb5 Bc4+ Ka5 Qc2 Kb4 Qb3+ >Kc5 Qb6+ Kxc4 > 11 00:07 3.435.997 457.522 +50.21 Kc3 > 11 00:08 3.654.878 456.289 +50.91 Kc3 > 11 00:13 6.363.059 470.640 +M7 Kc3 Qd1 Kb4 Kb7 Kb5 Qb3+ Ka5 Qc3+ Ka4 Bb3+ >Kb5 Qc4+ Ka5 Qa4+ > 11 00:19 9.253.019 480.177 +M7 Kb2 > 11 00:23 11.479.198 487.853 +M22 Kb2 > 11 00:31 15.261.099 491.659 +50.69 Kb2 Qc6 Ka1 Qb5 Kxa2 Kb7 Ka3 Kb6 Ka2 Kc5 >Ka1 > 11 00:31 15.261.099 491.659 +50.69 Kb2 Qc6 Ka1 Qb5 Kxa2 Kb7 Ka3 Kb6 Ka2 Kc5 >Ka1 > 12 01:00 30.354.140 505.902 +50.87 Kb2 Qb4+ Kxa2 Kb7 Ka1 Kb6 Ka2 Kc5 Ka1 Kc4 >Ka2 Qc5 > 12 01:00 30.378.474 506.307 +50.87 Kb2 Qb4+ Kxa2 Kb7 Ka1 Kb6 Ka2 Kc5 Ka1 Kc4 >Ka2 Qc5 > 13 01:19 39.858.075 504.532 +51.17 Kb2 > 13 01:19 39.916.095 505.267 +51.87 Kb2 > 13 01:45 54.671.513 520.681 +M7 Kb2 Qa5 Kc1 Bc4 Kc2 Qe1 Kb2 Qd2+ Ka3 Qc3+ >Ka4 Qb3+ Ka5 Qb5+ > 13 01:45 54.671.946 520.685 +M7 Kb2 Qa5 Kc1 Bc4 Kc2 Qe1 Kb2 Qd2+ Ka3 Qc3+ >Ka4 Qb3+ Ka5 Qb5+ > >Kc3 and now Kb2. Both squares that are not check but no capture. > >Uri, if you are there, I have only seen this type of behavior in early versions >of EnginMax if I remember, but not in this exact position. I hope that helps. >Maybe it had the same problem. > >:-) > >Christopher > > > >Christopher I know what is the problem of Movei. It assumes that the piece that threats the king is the bishop but assumes also that the king is attacked from the direction of the queen. In that case moving in the direction of the attack is allowed only when you capture the attacker. The queen is not considered by movei as the attacker so it believe that it is not allowed to capture it. Here are more positions to confuse Movei. [D]8/8/8/8/1nkn4/8/Q3B3/2K5 b - - 0 1 [D]8/8/8/8/1nkn4/8/B3Q3/2K5 b - - 0 1 [D]8/8/8/1B1B4/1BkB4/1BBB4/8/2K5 b - - 0 1 [D]8/8/8/1B1B4/1BkB4/1BBB4/NP6/KRn5 b - - 0 1
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.