Training with engines

[Ponti]

I`m preparing for a 21 min KO tournament and decided to play some games using Arena or FritzGUI with some engines. The difficulty is to choose 10 engines at about my strengh - 2029 FIDE. I`d like to find some nominees... any help? Otherwise, do you think that it is a good test? Are there any rating list that is gauged with a human rating (van Wely and many players have already played with Rebel, is it possible to "calibrate" a rating list based in these results of Rebel and/or other engines?).

[nczempin]

I`m preparing for a 21 min KO tournament and decided to play some games using Arena or FritzGUI with some engines. The difficulty is to choose 10 engines at about my strengh - 2029 FIDE. I`d like to find some nominees... any help? Otherwise, do you think that it is a good test? Are there any rating list that is gauged with a human rating (van Wely and many players have already played with Rebel, is it possible to "calibrate" a rating list based in these results of Rebel and/or other engines?).

I think your best options are to either use an engine that is genuinely "weak enough" (like Eden, although it is far too weak for your level), or to use one of the strongest engines and limit its available time or CPU.

[Aleks Peshkov]

Are there any rating list that is gauged with a human rating (van Wely and many players have already played with Rebel, is it possible to "calibrate" a rating list based in these results of Rebel and/or other engines?).

You can try to compare known online players ratings with their FIDE ratings and then compare your own online rating at same online server. IMHO any own home-made records should not be used for rating approximation.

[Thomas Mayer]

I`m preparing for a 21 min KO tournament and decided to play some games using Arena or FritzGUI with some engines. The difficulty is to choose 10 engines at about my strengh - 2029 FIDE. I`d like to find some nominees... any help? Otherwise, do you think that it is a good test? Are there any rating list that is gauged with a human rating (van Wely and many players have already played with Rebel, is it possible to "calibrate" a rating list based in these results of Rebel and/or other engines?).

Well, I think it's not so easy to train with engines. Whereas they might achieve similar results they play totally different. From a human point of view their positinal abilities look always way weaker then their tactical abilities. E.g. a 2000 engine might be 2500 tactically but only 1200 positionally. So you might see quite strange moves followed by a tactical shot even a GM would have problems to see. But of course any training should be better then none...

Greets, Thomas

[menniepals]

I`m preparing for a 21 min KO tournament and decided to play some games using Arena or FritzGUI with some engines. The difficulty is to choose 10 engines at about my strengh - 2029 FIDE. I`d like to find some nominees... any help? Otherwise, do you think that it is a good test? Are there any rating list that is gauged with a human rating (van Wely and many players have already played with Rebel, is it possible to "calibrate" a rating list based in these results of Rebel and/or other engines?).

I think your best options are to either use an engine that is genuinely "weak enough" (like Eden, although it is far too weak for your level), or to use one of the strongest engines and limit its available time or CPU.

The strongest engines would crush a 2029 elo player even with 1 second per move limit. Try analyzing games using Fruit/Toga, Rybka Demo, Glaurung or Spike under Scid with just 1 second per move limit and you will be amazed with the depth those programs can see even with such limitations. I think the best way would still be to play online ( much more convenient than going to chess clubs, at least you can play chess even with just your underwear on and curse your opponent as much as you want without getting punched on the nose) and then use engines to analyze games, prepare your openings and play against it using thematic positions you usually encounter with your opening repertoire than experimenting it against real humans who will ridicule your stupidity and crush your ego. Against the computer, you can always turn it off and restart a new position. Better than disconnecting when you get the losing positions. You don't even have to buy commercial engines if you can't afford one. Fruit engines are good alternatives especially for people stingy with their money and should easily overwhelm any master up to 2500 ratings. But do not just play against them to prepare for a tournament unless you like pain and is a masochist. Just my opinion.

[Ponti]

What I thought was to play against some weaker engines, as I don`t believe the strongest ones can be set to lower levels - I tryed that many times and got crushed; of course I`m playing real chess, not anti-computer strategy which is useless.

At WBEC Ridderkerk`s site there is a rating list which shows:

172 Geisterauge 0.07 2099 33 33 355 46% 2120 19%
173 Bearded 06-29 2099 57 56 121 58% 2021 21%
174 Bodo 0.2e 2098 30 30 428 43% 2148 18%
175 BigLion 2.23w 2090 44 44 216 55% 2026 15%
176 Ghost 0.13 2086 50 50 180 54% 2044 12%
177 OliThink 4.1.1 2080 43 43 208 46% 2112 18%
178 Atak 4.99 Linka 2072 62 62 101 56% 2000 22%
179 Fortress 1.62 2069 30 30 473 37% 2178 19%
180 Prophet 2.0 Epsilon 2 2060 81 80 67 58% 1964 9%
181 Armageddon 2.251 2058 31 31 378 47% 2075 24%
182 GK 0.90 2052 52 52 132 48% 2073 21%
183 King's Out 0.2.42 2049 126 117 33 71% 1822 9%
184 JanWillem 1.13 2039 33 33 382 56% 1969 18%
184b HfC 2006.03.01b uci 2038 124 116 33 70% 1826 12%
185 LaDameBlanche 2.0c 2037 24 24 675 47% 2058 22%
186 Ax 0.8 2027 30 30 459 53% 1997 15%
187 Matheus 2.3 2026 27 27 528 53% 1993 21%
188 LordKingIII 2020 64 63 111 56% 1931 15%
189 RivalChess 1.17 2015 127 119 33 68% 1823 9%
190 GNU Chess 5.00+ 2012 76 77 60 48% 2025 20%
191 Asterisk 0.4b 2000 30 30 444 48% 2008 18%
192 Gerbil 0.2 1997 24 24 714 47% 2008 15%
193 Feuerstein 0.4.5-b2 1992 104 112 34 34% 2105 15%
194 EnginMax 5.11c 1983 44 44 204 55% 1951 21%

I`m used to win against LaDameBlanche and Atak... at 5 0 KO.
Is there a rating list with time=21 min KO ? How does time affect these engine ratings?

I don`t trust ICC's ratings, do you? "Internet chess is not chess!" (Nigel Short).

If you get a top engine it crushes 90% of players, but such an engine`s strengh is _way_far_ from reality when you play tournaments at =~ 2000 level.

In this way I think there`s no other choice than choose many engines to play, because to weak an engine`s play is not easy, and results IMHO are ambiguous (i.e.: Fritz, Chessmaster, Rebel). I did not try Delfi because it is comercial and I think I already have too many engines (and there are so many for free!). The last one I bought was Rybka 2.1 but that is only for analysis...

[Ponti]

The new version of ICC`s Dasher interface comes along with a new feature: it is possible to play against the built-in engine (Crafty) and choose a level of play... (I thought someone did already that using Winboard). Other engines can be chosen too (UCI or Winboard, not Chessbase's) but it is not possible to play a tournament like in Arena or FritzGUI.

The question is : how does this "weakening" affects Crafty`s play? Won`t it make some stupid blunders? I mean, there are blunders that humans do not make at a slower pace of game very easily, as i.e.: not deberately moving a peace when attacked by a pawn if there`s no obvious compensation. I saw that kind of blunder many times made by Fritz when it is playing in "sparring" mode.

[AdminX]

I like the idea of playing online for training. There are only two problems you really have to overcome. One is the longer the Time Control the harder it is to find games, which means you have to wait a little longer for someone to pick up your game. Second it will take a while to build your friends list with players that give you a really good game.

Once those issues are overcome, training and preparing for events no matter what Time Control should be no problem. With a 2000 plus rating I don't think you will have any problem finding a game, however finding the good players that bring out the best in you ... Well that will take some time. As for the chess engine part, I think they are best used for help analyzing your games.

PS: Have you checked out Delfi, it sounds like what you need if looking to a good Chess Engine for Training.

[Aleks Peshkov]

At WBEC Ridderkerk`s site there is a rating list which shows:
/.../
I`m used to win against LaDameBlanche and Atak... at 5 0 KO.
Is there a rating list with time=21 min KO ? How does time affect these engine ratings?

I don`t trust ICC's ratings, do you? "Internet chess is not chess!" (Nigel Short).

FIDE and WBEC ratings are totally incomparable (2xxx number looks similar because they are artificially made looking so). Random "strong" computer may be (or may be not, WBEC simply does not tell) very weak against human master, and a strong FIDE player (like Adams) can be totally destroyed by a modest program. Online rating based on computer-computer, human-human and at least some computer-human games so it is at least a reasonable computer-human game results estimation.

[gerold]

I`m preparing for a 21 min KO tournament and decided to play some games using Arena or FritzGUI with some engines. The difficulty is to choose 10 engines at about my strengh - 2029 FIDE. I`d like to find some nominees... any help? Otherwise, do you think that it is a good test? Are there any rating list that is gauged with a human rating (van Wely and many players have already played with Rebel, is it possible to "calibrate" a rating list based in these results of Rebel and/or other engines?).

I never weaken an engine. It does change their playing style.
Spike,SOS,Ruffian,and my Favorite Crafty. These are very
good mix to improve playing chess.

Best to you,
Gerold.

P.S. Crafty has many diff. versions to choose from