My 'strong club player, class A level' chess computers

Yeno 532 XL

Year: 1989
Programmer: Ulf Rathsman
CPU: 65C02 @4Mhz
ROM: 32Kb
Elo level: 1824
(1760 FIDE)
CMhz: 4
Rperf: 101%
KT: 1685

The second Yeno chess computer (1st one was the 301XL), and the strongest one from this brand. With assets such as its 32K, its 4Mhz 6502, and its program close to the Mephisto MMII one (based on Plymate), it is a strong player, uneasy to find on second hand market. I spent a long search time before finding this one, bought for 80€. The small diods are modest, one per square, level with the playing surface like the modular board ones. They are sometimes hidden by the pieces, which are a bit large compared to square size, and too light and not magnetic enough for a good holding on the board. Fortunately the moves are displayed in two small LCD windows, offering in sequence various information. Unlike the 320 XT, the control panel is pleasant to use, the keys are push sensitive with as little pressure as the squares require. The one I own failed after circa one hour to one hour and a half playing, some contiguous squares ceased to respond to push. An area on the left of the chessboard was overheated. After turning the computer off and waiting a while in order to let it cool, the square sensors were restored. I performed a careful dismantling/reassembling, with a meticulous positioning of ribbon cables connecting the mother board to input/output devices (push sensitive chessboard, LCD displays); I reinforced a heatproofing Bristol board with a second layer of thick paper, and drilled holes under the hot area, in the black plastic bottom cover. Not a quite purist way to do as a collector of chess computers, but invisible while playing, and my 532 XL never failed again. Maybe the careful reassembling would have done the trick, however the hot area completely disappeared. The most important thing is to play again, isn't it?




Overall, a usual profile shared with many other chess computers from the same or so period of time (the program can be considered dating from 1985, birth year of the Mephisto MMII); featuring a strong game in the beginning phase, lowering slightly in the middlegame, especially due to a lack of strategic relevance, then dropping significantly in the endgame. Its strength is mostly based on tactics, with counterattack in clear preference to attack. Finally, the calculations ability is disappointing considering the 4Mhz 6502; this reveals an essentially brute force approach, unsuitable to selectively deep dive into variations. The global score is nevertheless quite satisfactory, the Emerald Classic is the only one achieving a higher score in this class!

Tiger Grenadier

Year: 1998
Programmer: Chrilly Donninger
CPU: SH7034 @20Mhz
ROM: 32Kb
Elo Level: 1833
(1767 FIDE)
CMhz: 23
Rperf: 94%
KT: 1541

I bought this Grenadier (60€ on Le Bon Coin) from a collector, who warned me: the playing style is somewhat disconcerting, but interesting. Actually I already had read such comments, and it confirmed part of my motivation to add it to my collection. The rest of my motivation: the author, the brand, and the processor (a fast and powerful Super-H Risc) at that time not present in my collection. This device is the single one authored by C. Donninger (it also exists in  few numbers branded Millenium and named Genesis, and  the Sakkara Vega uses the same program in a different housing). I am not fond of Nimzo, to my opinion its playing style isn't up to much; but I knew I would find something different with the Grenadier. It has a talent for playing unusual openings, then carry on playing surprising moves and displaying optimistic scores (proved false by any chess engine), later suddenly change his tune, admitting a negative score, and from then on, resume a more serious play and slowly get back on its feet, to score a draw that came out of nowhere, sometimes even win... I would qualify its playing style as speculative.

Another peculiarity of the program powering the Grenadier is to enable some tuning regarding its playing style, thru changes applied to evaluation criterias. My approach in this area is definitely to keep its original playing style, so not throw into confusion any parameter; but tune using marginal updates. The trigger for such tuning is blunders the program may play. I then look for which parameter can significantly modify the chosen move and avoid the blunder, and what value as close a possible to the default value can achieve this result. Next step is validation... I do it using three methods:
  - play a few games "under survey" of a strong chess engine (most often HIARCS 14). If the engine reveals a quite bad move played by the Grenadier, I take back the move, and resume computing using Grenadier's default value for the considered parameter. Should the standard one play roughly as bad, the new setting would be declared not guilty.
  - Eight games gauntlet against the very same opponents the standard Grenadier faced once ago, scoring 4 points out of 8 (a balanced score). A better performance using the new setting is obviously welcome.
  - Playing skill evaluation based on five classical games, playing white and black moves, providing a score per each move. The overall rating is measured using "Elo points". The absolute value does not matter, the point is to check whether the new setting do provide any significant gain (if not, better jettison it, at the benefit of the playing style) ; and check as well whether this gain is balanced enough across the five games (to avoid introducing any weakness in some situations). This test designed by Spacious_Mind is available for download here.
Currently, my best setting alters both two parameters: "Pawn" (pawns structure) raised from 0% to 2%, and "Shield" (king safety) raised from 20% to 25%. Excerpt from my 'Spacious_Mind' test results (best scores per game are underlined using yellow color):




The known lack of development suffered by this program translates here into a rather irregular profile. It is also atypic (as expected), showing attacking skills plainly superior to defense ones, unlike most programs. I pointed out its behavior trend to select rare openings; the graph displays an aggravating factor: it does not master the opening transition to early middle game, once out of book. So, it is no surprise it often faces tricky situations during this very phase. On another hand, the middle game is its favorite domain, leveraging strong support from tactics and, which is unusual for a chess computer, a consistent strategy knowledge. It is exactly from this starting point it does resume the serious play I did mention above... and, should the occasion arise, it launches a counterattack to reverse the previously threatened situation. More commonly, endgame is weak and it does not know much about standard engame positions. Lastly, the calculations ability is deceptive with regards to the powerful processor, thus explaining further its low Rperf, worthy of a full category under this one! 

Mephisto Rebell 5.0

Year: 1986
Programmer: Ed Schröder
CPU: 65C02 @4.9Mhz
ROM: 32Kb
Elo level: 1834
(1768 FIDE)
CMhz: 4.9
Rperf: 101%
KT: 1506

This module is the missing element between the Mephisto MMII and the MMIV. It should have been named MMIII, but the program performed so remarquably during the 1986 world computer chess championship (WCCC), the brand chose to leverage its fame. Rebel finished this WCCC at 
fifth rank (and 1st microcomputer) after having been in a position to win it! Facts are it won its games against a cluster of 20 parallel computing 32bits SUN, a Cray XMP, and an Amdahl 470 whose program (Bobby) did win its own game against the title owner (and future final winner) Cray Blitz. And this using an Apple IIe, however featuring a bitslice accelerator card (most probably a Schaetzle & Bsteh DC65, launched in 1984 for the Apple II). Such a card is able to accelerate the original 1Mhz 6502 up to 11 or even 12.5Mhz; but it has been set to half-speed in order to keep control on the clock, resulting in 5.5 to 7Mhz depending on the source (Jan Louwman or Eric Hallsworth). So the later sold module provides a performance level close to the one of the WCCC hardware. It is the outcome of the very first joint work between Ed Schröder and Hegener & Glaser, and many other products will follow, stronger and stronger (MMIV and MMV, Polgar, Milano, Risc...) until 1994; then the author will set up in business on his own with the PC-Rebel series, starting from Rebel 6. As I much appreciated the PC-Rebels, I was keen on inserting an Ed Schröder's program in my collection. And this Rebell 5.0 is a true milestone... I bought it 118€, with the Modular Board, the chessmen, the  mains adapter and the user's manual, on German eBay.



The skills profile of the Rebell 5 displays much contrast, revealing so a young program still lacking some development: very relevant after exiting the openings book, it definitely fails the endgame standard positions exam. We can see the programmer took particular care over injecting a fair amount of chess knowledge to get the program to develop its position and play good positional moves (strategy). So, the Rebell 5 did study the chess manual, but not to the last chapter... The middlegame playing strength is rather average, the defensive skills as well, while attacking and sacrificing abilities are weak (but this is not unusual with any program from the same era), and the program is more comfortable with recognizing threats, with leveraging tactical patterns, and with counterattacking - there again, rather usual properties, driving some passivity, awaiting opportunities to appear. The most surprising result, with regards to the rather fast processor back in time, is the poor calculations performance. Nevertheless, the Rebell 5 is a strong player, thus reinforcing the feeling the programmer did inject much chess knowledge (endgame kept apart), resulting in a rather sophisticated evaluation function, so a bit slow one.

Excalibur Ivan the Terrible

Year: 1996
Programmer: Ron Nelson
CPU: H8 @12Mhz
ROM: 48Kb
Elo level: 1844
(1775 FIDE)
CMhz: 7.8
Rperf: 99%
KT: 1409

Ron Nelson has left his mark on the chess computing world with his skills as an electronician engineer, but he was as well a pioneer programmer, author of the famous Chess Challenger series before the Spracklen started working for Fidelity. I got lately interested with the Excalibur brand story and discovered recently Ron Nelson's production in this context. Not only he designed the devices, but also he authored the program, including the attack bitmap concept and using the map generation process explained by Ken Thomson, the famous author of Belle. So I started my quest for one of his program running on H8, preferable to later ones for which the H8 was no more available. Thus the author has translated his program for another microcontroller, 6502 like, but with less RAM, this led him to remove the attack bitmap tables; at the same time he disabled the permanent brain to mitigate any bug risk and enhance the batteries life
(Ivan II the Conqueror, Alexandra the Great, both from 2003). The devices powered by a H8 are the Mirage (featuring robotic move of  the pieces), Ivan presented here, Grandmaster (a de luxe one with a high quotation), and Igor. Ivan and Igor seem to be very close, and feature voice and entertaining sounds, not necessarily enjoyed features by serious chess players. Indeed it increases the toy aspect of this chess computer, despite its strength, yet these features can easily be disabled. Ivan voice messages are chosen according to the game status, with a sense of humour. As an example, using the 'hint' key can result in hearing: 'a hint? You do not need a hint!' if Ivan evaluates itself in a loosing situation; or 'you want a hint? Resign!' if it has a neat advantage. It nevertheless generously suggests a move assuming you hit the key a second time. Cherry on the cake, the voice compression used has been designed by Ron Nelson as well! A nice buy from UK, for 51€.




I was keen on measuring Ivan's behavior while performing the Khmelnitsky test: Excalibur devices were distributed late in the market life (starting from 1993), and the market had started to shrink, facing PC programs competition. They mostly aimed at general public market and could not benefit from the reputation built on comparative testing, that was very popular since the early chess computers appeared. In addition, let us consider an unusual combination including a programmer initially known for his primitive programs fitting occasional players' level, but now leveraging re-introduced attack bitmap technology (old, but having stood the test of time), and a fast processor; the result being a definitely strong chess computer. To be honnest, Ivan didn't shine in this test: it often succeeded in evaluating correctly the diagrams (identifying who is best), but it too often failed in spotting the best move (scoring five points) while choosing a decent but suboptimal one (scoring only one point). This lack of accuracy augmented by too many fails with endgame positions resulted in a loss of hundred points or so, with regards to other devices within the same category. The skills profile is typical for early chess computers: ability to spot threats (thanks to the attack bitmaps), tactical strength, strong defensive and counterattacking skills; with a spike of strength at middlegame, clearly lowering at endgame. The already mentioned lack of accuracy often prevents it from finding the most decisive attacking moves. As a conclusion, Ivan fits its marketing target: a strong player (you would better try to resist to, until endgame is reached), but no analysis tool... Nevertheless, Ivan does not deserve a too negative perception: I strongly advise you to use the comparison tool, and overlap its profile with the Saitek Turbo King II D+ one. This Kaplan program deserves an excellent fame, and you can check by yourself Ivan is definitely on par regarding most of the domains! To conclude with some fun, I discovered while performing the test some retorts from Ivan, I had never heard during standard games: should you handover your turn-to-move to him, with a loosing position: "no way, I don't want your mess" (you then need to presse "move" again); and if exiting setup with a diagram without queens, he sometimes would moan: "what is this? Where is my queen?"

Saitek Turbo King II

Year: 1990
Programmer: Julio Kaplan
CPU: 65C02 @5Mhz
ROM: 64Kb
Elo level: 1875
(1799 FIDE)
CMhz: 5
Rperf: 103%
KT: 1506

To get a strong Kaplan program, my choice balanced between the Tandy Chess Champion 2150 and the Turbo King, and I finally, reasonably, chose the Tandy: better adapted to my budget and to my playing level. But the seductive power of the Turbo King did not die down completely... So, when I happened to spot this nice one offered on French Le Bon Coin for 30€, I went wild over it... Usually, one would have to offer slightly more than 100€ to buy this device! The EGR (End Game ROM) was not present, but tests I performed have revealed the software is the 'D+' version; quite different and performing better than the 'B' one used into the Tandy. In addition, Julio Kaplan's program being chess knowledge oriented, needs computing time to provide full extend of its strength; it improves a lot running at 5Mhz instead of 3 as it does on the Tandy. Showing 150+ 'Elo' points as an order of magnitude, the Turbo King is far from being a duplicate in my collection!



The Turbo King II transitions perfectly from the end of the openings book to a solid middle game, leveraging tactical skills (recognize threats, calculate, defend and counterattack): chess computers often provide evidences for such properties. Considering the program is well known for being chess-knowledge oriented, and so being rather slow (310 nodes per second as an order of magnitude), its lack of knowledge for standard endgame positions and its relative weakness in positional skills (strategy) are disappointing. Compared to other versions of Kaplan's chess computers, while neutralizing their various processors power impact, it runs roughly half-way speed between the "B" version of the Tandy 2150, and the much lightened version of the Saitek Blitz. So, this D+ version has probably been lightened as far as chess knowledge is concerned, in order to speed up its evaluation calculation, and so be able to analyze deeper. The result is far from being uninteresting; its profile is very similar (of course weaker) to the one of the Mephisto Berlin, with which it shares a certain amount of passivity (attack). Considering the domination of Lang's programs, leading towards a similar pattern makes sense. In a nutshell, a mini-Berlin/Vancouver, except their endgame strength!

Novag Emerald Classic

Year: 1996
Programmer: Dave Kittinger
CPU: H8 @26.6/2Mhz
ROM: 32Kb
Elo level: 1975
(1877 FIDE)
CMhz: 8.65
Rperf: 105%

KT: 1755


This computer is on the verge of expert level, thanks to the 32K Kittinger program powered by a slightly overclocked H8 CPU running at 13.3 Mhz. While showing close statistical performance to the hereafter GK 2000, it outdoes it as far as chess skills are concerned, being thoroughly strong at any stage of the game. Its playing style is somewhat passive, rather positional, often leading to one of those two endings: if its opponent keeps the game under control, the Emerald Classic tends to slowly get extinguished, and ends up losing or drawing without having fought much. On another hand, if it takes an advantage, it starts attacking quite convincingly, and wins. It is a nice item as well, with a pleasant finish, yet a disadvantage is that the LCD display is pushed away from one's eyesight and, above all, lying flat thus uneasy to read, unless one uses the board on a coffee table. I bought it 85€, a fair price.




Before performing the
Khmelnitsky tests, I spotted more chess skills in this program, compared to its rival GK2000. Well, this is confirmed, with over 100 more Elo points achieved! On another hand, I thought it was evenly strong, while the test underlines some weaknesses in opening and endgame, and particularly a lack of theoretical endgame knowledge. The skills profile is relatively atypic as, while strong tactics and capacity for recognizing threats are usual for programs, the prominent weakness for defense is surprising (on par with the Excellence and the Systema Challenge, so one category underneath!). This is consistent with the feeling I shared of a toneless play if under domination. The weak attacking skills confirm the noticed passivity (but this trend is fairly universal) whereas the capacity for counterattacking is outstandingly pronounced. Its expert skills domain is the middlegame...

Saitek GK 2000

Year: 1992
Programmer: Frans Morsch
CPU: H8 @20/2Mhz
ROM: 16Kb
Elo level: 1998
(1900 FIDE)
CMhz: 6.5
Rperf: 108%

KT : 1648


A 16K amongst 32K and larger! I bought this chess computer bare (without pieces, mains adapter, manual nor original packaging) for price no doubt a bit high on German eBay (70€). Fortunately it runs perfectly, and it finally inherited the Saitek/Mephisto 
Chess Challenger 10Mhz set of chessmen. I was looking forward to add to my collection a H8 microcontroller; I had none at that date. Here the H8 is clocked by a 20Mhz quartz, reprocessed by a divider to provide 10Mhz at CPU level; the computing power exceeds the one of a 6Mhz 6502. Frans Morsch's program is impressive with its tactical strength, it is based on Fritz 1 and fully benefits from the H8 speed. Its playing style shows the characteristic features of computer chess, including some weakness in strategy. It is the Mephisto Europa big brother, and the strongest 16K program!



 
Being a strong player once leaving the openings book enables engaging games on a solid ground, but its playing level gets back to rather mediocre during middlegame and endgame, with a worthy of note very weak knowledge of standard endgame positions. It is neither a great strategist nor a strong calculator; and even less a great assailant. On another hand, it does
efficiently spot threats and tactical patterns, to mightily counterattack. Avoid mistakes in order to reach the endgame, and you will beat it!


next category: 'expert player, candidate master and higher level' chess computers
previous category: 'average club player, class B level' chess computers
back to 'my chess computers'