[ Sofia ] [ Business ] [ prices ]
[ copy protection ] [ caught! ] [ million-gate ] [Kasparov ]
About ignorance, naughtiness
and being right
Once upon a time there was no (or only linited) Internet (unimaginable nowadys) and so it was with the general knowledge of computer chess, lmited and only reasonable well understood by the chess programmers themselves. How different are things now anno 2015, many (non programmer) end users are well informed due to the explosion of knowledge the Internet caused. But this is a fun story when knowledge was limited.
There was this German computer chess magazine and it was famous (in my view) for its longwinded high qualified reviews (Peter Schreiner in particular) evaluating new released chess programs describing new features into detail. Then when it came to judging (predicting) the playing strength of a program things often got blurry up to a point they sometimes used laughable reasoning, not limited to REBEL only BTW, the story below is one example.
I won't blame them because that would be unfair, knowledge was truly limited in those days.
Anyway, without mentioning the reviewer (they had several rotating reviewers and it wasn't Schreiner) a new REBEL was put on the rack and the review was excellent. Then when reviewing the playing strength the reviewer to my amazement stated:
We don't consider REBEL much stronger because
there hardly is an increase of the NPS.
Imagine my horror.
And so I decided to teach them a lesson, I artificially increased the NPS by counting the QS nodes double and awaited what would happen. And indeed one year later with the next REBEL the same reviewer stated:
This new REBEL is obviously stronger, it has gained much in NPS.
What a nonsense.
And speaking about lessons, in retrospect, there was a lesson for me also, don't do such things even when you are right. Although I am not completely sure which were the 2 versions in question I suspect (it is a long time ago) it was about REBEL 7.0 and REBEL 8.0.
And then the fun part of the story is that the reviewer (although using faulty reasoning) was 2 times right in his judgement about the playing strength of the two REBEL versions. As when the SSDF rating lists appeared (back then the only true measurement of playing strength) REBEL 7.0 was not much stronger than REBEL 6.0 while REBEL 8.0 topped the list with a 65 ELO lead on number 2.
Irony at its best.
Sofia, me, stupidity and a smile
in that particular order
The year is 1997 and it's the AEGON time of the year again, my favorite tournament. 50 human chess players varying from 2100-2650 elo against 50 computers and it's all happening in my town of birth, The Hague.
What we did not know back then it would be the last edition as the human pool of 2100-2650 started to lose from the computers (not good for the sponsor AEGON) as the computer pool became stronger and stronger every year and AEGON in return had to raise its budget attracting stronger players, every year AEGON invited more GM's to save the face of the human race and of course of our their own advertisement goals. The tournament became too expensive. We (the programmers) with our neurotic passion for the game and the elo hunger killed a nice tournament ourselves.
On a different level (the commerce) the same thing happened, the commercials of that time and the rivalry between them made the chess programs so good (rich features) and so strong that at a certain point there really was no good reason for the average chess player to buy the latest Fritz, Rebel, Hiarcs, Shredder. The version they bought 2-3-4 years back already satisfied their needs and why would you pay for a new version anyway when it's only doing (somewhat) better at what it is already good? And so the commercials in 10 years time (the 90's) killed their own chicken with the golden eggs.
But I am drifting away, back then in 1997 facing a GM at long time control was more or less a receipt for defeat for the computer, occasional draws were celebrated as victories, let alone a win. In this atmosphere consider my game against the lovely Sofia Polgar, Rebel having the white pieces.
It's a won position for REBEL but after 5 hours of play, it was already after midnight and with the prospect of another 2 hour drive home I wasn't thinking clearly any longer.
After the exchange 121..Kxg6 122. h8=Q I expected another 30+ moves before REBEL would get the a-pawn to promote in exchange for the black bishop and then the remaining KQKR ending (which knowledge back then wasn't in the progam) would take endlessly, maybe even ending in a draw because of the 50-move rule. And so I offered Sofia a draw.
Sofia looked up surprised, smiled accepted the draw and added: "I was just planning to resign!"
Oh well.... I was just forgotten I wasn't playing a computer opponent which operator clearly would played all the way to the bitter end but I was playing a human, who was as tired as I was and also that players on that level just resign in a lost position, it's the etiquette among them.
In a way I can sympathize with Kasparov when he resigned in a drawn position against Deep Blue that same year
Nevertheless the (human!) blunder it was a good tournament for Rebel, with 4½ of 6 together with Kallisto ending as best computer, only one loss against the tournament winner GM Yona Kosashvili (6 or 6) and a win against GM Gennadi Timoshchenco.
Collected AEGON games from the 1991-1997 period for download here.
The REBEL way
What happens when there are 2 groups of people, the first well dressed with tie and 3 part suit and the second group of 2 country boys wearing jeans and T-shirts who want something from the other. Quite some funny things can happen when 2 different worlds meet and the well dressed group doesn't understand the needs of the T-shirt group and that money isn't always the dominant factor for the T-shirt group.
And one of those issues was the limited hardware I had to work with and that progress was no longer possible with only an 8 bit processor at 5Mhz, 32/48Kb ROM and only 8192 bytes of RAM and that I wanted to expand my horizon because I am what I am, a chess programmer, I wanted hash tables, had other ideas which I couldn't do because of the limited hardware and speed and that I was already behind in the competition because of the limitations and that the tie-group didn't understand that more money (that's probably what they thought the T-shirt group wanted) was not going to help even when I said, no, no, the money is good. And the well dressed tie group remained deaf to the wishes of the T-shirt group and underestimated that the leader of the T-shirt group (although normally a nice rebel) in essence remained what he was, a REBEL. And so he bought himself a nice 32 bit RISC machine and had the time of his life. So much new to explore.
Of course I am speaking of my relationship with Hegener & Glaser, the producer of the Mephisto brand during the 1984-1993/4 period. It was a fine company led by Manfred Hegener a true gentleman in businessland. He always treated me well and I have good memories working with him.
Such a pity his company went into bankruptcy because the PC was on the march and eventually the 80486 outperformed his special hardware which preluded the end of an era, the death of the table top computer replaced by software on floppy disk which was a lot cheaper also.
And of course the threatening bankruptcy gave the inevitable tensions. Sales went down and down, the contract had to be terminated, yet mister Hegener wanted new (kind of last resort) projects from me, the RISC program, our first PC program. Understandable from his point of view and I wasn't unwilling.
And so it was a turbulent time, lots of negotiations, lots of times driving to Munich or meet each other somewhere elsewhere. In this light read a couple of typical business tricks that work well on other well dressed with tie and 3 part suits groups but do not work well on a couple of country boys from Holland.
Case-1: We (Rob Kemper and I, Rob being my right hand and developer of the REBEL interface) drove to Munich, about 900 km which for a Dutch living in pinhead Holland is quite some drive, from east to west here is just 200 km and arrived late in Munich we picked an hotel. A meeting with Mr. Hegener and his staff was planned the next morning at 9:00 AM. We arrived in time and were greeted warmly by the receptioniste. It became 9:00 and no Mr. Hegener. 9:05, no Mr. Hegener, 9:10 no Mr. Hegener. I went to the receptioniste and asked if Mr. Hegener had not arrived yet or if he had some sort of delay. But she assured me he was in the building and soon would pick us up. 9:15 no Mr. Hegener. I had enough. Classic businessman trick of mental manipulation to show you who is boss and have an advantage in the negotiations right from the start.
And I said to Rob, let's go home, I have enough of this puppet show. But we are just arrived Rob smiled, in reality Rob is even worse than I on issues like this and we waved to the receptioniste and left the building on our way back to the hotel. A few blocks further we were halted by a few coworkers who offered apologies and begged us to return which we did. The meeting began and we surely had an psychological advantage during the negotiations. A hollywood ending after all.
Case-2: we were in Munich again, can't rememeber what it was about but the negotiations didn't go well (you can't always get what you want - Mick Jagger) and I was pretty much irritated. Manfred must have noticed my facial expressions and as the genleman he was he suggested a break and have luch in one of his favorite restaurants and talk further afterwards. And before I knew it I popped it out, "sorry, I am not in the mood" and kept my (undoubtly) stubborn face. Oops....what did I just say? They, on the other side of the table looked shocked, that was new to them, for me as well BTW, unintended but I surely meant it. After some "hin und her" we went for lunch anyway and funny enough after lunch the atmosphere had changed and we went home satisfied.
With the downfall of Hegener & Glaser juridical issues became a topic. Natuarally I didn't wanted to go down with them and I asked from a change in our contract as the earnings no longer justified his exclusive right on my work. And Manfred didn't want it. Understandable from his point of view, his 3 programmers (Lang, Morsch and me) were the base of his company. A conflict was unavoidable. And so we went to look for advice and picked the nearest lawyer just a couple of km after the German border. German law is differnt than Dutch law and my contract was under German law. And now it was our turn to be flabbergasted as at a certain moment this lawyer suggested the following:
It would be nice to remember them on a letter you have sent them in the past, create
such a letter, date it 2 years back and the problem is solved.
We were speechless, finished our talks with this corrupt lawyer, went home to never return. We then (sic) drove again to Munich, consulted a top lawyer office (600 DM per hour back then), they sent one letter and a couple of days later things were solved. My first experience with lawyers, there are good and bad ones.
Nevertheless I did 2 more (hardly profitbale) projects for them out of loyalty, the Milano and the Mephisto Risc which were sold in poor quantities. The landscape had changed in a rapid tempo, the golden 80's had gone. The PC became the future.
The payment in advance cheque
The early 90's were turbulent and chaotic, with the downfall of Hegener & Glaser I had a 2 year financial loss and I was seriously considering to give up commerce and get a regular job. Unfortunate for my 2 coworkers too and I hated that thought. But in the meantime I had ported REBEL to the PC, Rob made his first interface and suddenly we found ourselves in the middle of attention. Saitek wanted to distribute it, Hegener and Glaser too and there was a third candidate Ossie Weiner. It was crazy, they were outbidding each other, suddenly the PC was everything. And at a certain moment during again another negotiation round one of the parties did something remarkable, he laid down a signed cheque on the table of $40,000 which I could cash immediately, I only had to promise that I would negotiate with them first, no further conditions, carte blanche. No doubt I must have looked surprised, it's a powerful visual attempt and tempting approach. Nevertheless I was keen enough to decline the offer within seconds. I just did not feel they were the right partner for us for reasons I can't disclose because that's not chic.
In the end I decided it was best to start standing on my own feet and no longer depend on others. A new world opened, instead of just sending one or two eproms a year by regular mail I now had to do everything myself, advertizing, making manuals nobody reads (yikes), establishing a dealer network, the many phonecalls, dealing with customers -:) surely I had to learn a lot of new things.
The crazy price setting of the first PC programs
Roughly € 200 nowadays
It was in the beginning of the 90's the 4 then strongest (and commercial) chess programs M-chess | Hiarcs | Rebel and Chess Genius released their chess program on a floppy disk (huh? what's that?), it was the beginning of a new era and nobody at that time had a clue what price to ask.
Chess Genius distributed by Ossi Weiner (as first one) set the price to $ 149 | DM 199 | HFL 199 (no € back then) and the rest (M-Chess, Hiarcs, Rebel, Fritz) silently followed, apparently people were willing to pay that kind momey. Later Ossi revealed he had no idea himself and the DM 199 for Chess Genius 1.0 was a gamble and try-out. Updates in that time were invented too, loyal customers were rewarded with a 50% discount (stlll roughly € 100) and again the rest of the pack followed. Well done Ossi :-)
This paradise lasted about 4-5 years till ChessBase decided to slaughter the chicken with the golden eggs and suddenly out of the blue announced a drop in price with 50% with their new Fritz. The timing could not be more worse, I just had released a new REBEL and I feared a collapse in sales.
As for some background information, a (any) popular game sells for 3 months, thereafter sales are in dribs and drabs meaning that as a game developer you have to make your earnings for a whole year in those 3 months. So I had no choice and responded immediately with the same price drop. Of course I had to compensate customers who just bought the new REBEL against the old price.
I will never forgive ChessBase for that :-)
Kidding aside it made me wonder why on Earth ChessBase would shoot in their own feet as well. Certainly not to pester me or the rest of its competitors. I know that at that time they had connections with some distributors of the mass market game industry and they could sell Fritz in stores, shops and warehouses and of course ChessBase had to meet the standard pricing for a game in that market area, a € 200 price would not be acceptable. One day I will ask Frederic or Matthias if my wild guess is correct.
The idiotic copy protection
Besides the price setting for PC programs there also was this fear of piracy people copying the floppy disk. Again it was Ossi Weiner who also here set the tone. Chess Genius 1.0 was released with a copy protection and the rest of the pack (Hiarcs | Mchess | Rebel) followed.
To protect diskette's from copying you had to buy expensive software first, by head I remember we paid approx. $1000-$1500 for it. Steps:
1. At release time (first with Rebel 6.0 in 1994) we produced a master diskette and sent it to a copy house to make the first 1000 reproductions.
2. When the 1000 diskettes came back we on each one had to put the copy protection manually which took about 3-5 minutes per diskette, quite a self chastisement, first self punishment. The copy protection software did all kind of horrible things to the diskette such as changing the logical sequence of tracks which made the diskette uncopyable. And herein lies the first problem, customers couldn't make a backup of their expensive software, first customer punishment. As a result we had to replace quite some diskettes for free because diskettes weren't exactly the safest media, second self punishment, and counting...
3. Next problem, the number of installations, another crazy idea we took from the trend-setter Ossi Weiner. Chess Genius, Mchess, Hiarcs and Rebel all allowed 3 installations.The result, massive complaints, people lost installations, I don't know how they did it but some were able to spoil the 3 installations in a couple of days. Customer is king and always right and so we had to replace them free of charge, second customer punishment, third self punishment. After a month we had enough and only provided diskettes with unlimited installations, that helped. At the time we were just a bunch of amateurs trying to find our way in a relative new market segment, the rising popularity of the PC.
What then? Is copy protecting your work a bad thing? From the then available statistics it was known that commercial software for the PC was copied 3-5 times more than the number of legal samples so from that point of view it was the right and logical thing to do, so it seems.... But we learned something. Due to popularity of our promotional free REBEL DECADE many people bought the real thing. We came to the following risky conclusion: it might be true that unprotected software is copied 3-5 times but on the other hand it has 3-5 times more attention and maybe, just maybe, there are enough honest people among those who would buy the real thing after all.
And so in 1997 with Rebel 9.0 we took the gamble, it was released on cd-rom with no copy protection. And much to our relief we noticed no drop in sales. No more self chastisement and pestering our loyal customers. Oh yes, many illegal Rebel's out there ever since but life is one big compromise anyway.
Copy protection is a bad thing, don't underestimate the promotional effect the pirates cause, they make your brand name known free of charge. But that's of course knowledge in retrospect, at the time we did not know any better.
Last and this is funny and shows how deeply we were into this copy-protection-thing, we even developed a DATA copy protection system ourselves to protect our extra opening books and later our game databases from copying. How it is done? Well, during installation the data was encrypted with an XOR from the volume serial number of the harddisk, usually the day and time a harddisk is formatted. Meaning if user X would make a copy on diskette and mailed it to user Y it wouldn't work because the volume serial number of user X is never equal to user Y. As by instinct we felt that this would probably a bridge too far we never practiced it, but it worked.
REBEL 6.0 the cheating program
A storm in a teacup
That's what the title stated in the former mentioned German computer chess magazine (see first anecdote) some months after the release of REBEL 6 in 1994. They claimed I had cheated their rating list. As already stated these were the dark days before the internet and knowledge was limited back then and in their enthusiasm they had created a testset of 30 positions (the BT-2630 test) and (in a nutshell) the program that solved the most topped their rating list.
Yes, you have read well, a rating list of only 30 positions to measure the strength of a chess program. Ridiculous of course, only 1 game already contains 30+ positions and every move has to played well but (I suppose) that kind of knowledge (or insight) was only available to the chess programmers of that time. And back then there was only one source that deserved the title of a rating list, the famous SSDF folks from Sweden playing thousands of 40/2h games manually, the forerunner of CCRL and CEGT.
This as an introduction to the accusation that was made, which was: we have found 3 positions of our BT-2630 rating list inside REBEL 6 and the program profited from that. And the accusation was true, inside REBEL 6 there was a small database of test positions that I had created after reading about a new feature of the Spracklens in their MACH IV (I believe) that it could learn from its mistakes. And that it was done via hashing. That back then was new and it fascinated me and so I had to try myself, hence the 3 positions were inside and REBEL 6 must have profited from that because I remember it worked, they missed the fact there were also 2-3 positions of their former BT-2450 rating list inside. After the experiment I forgot all about it and continued with my regular work.
And then it became release time, REBEL 6 was ready for it and in the hectic of the production phase I had forgotten to turn off the learning data and code. Actually that's a half-truth because during the production phase I DID remember but was unwilling to correct it and to restart the production for reasons of money loss, time loss and frustration. And I remember I thought, who cares. And that was a mistake. As I years later found out, people DID care, WORSE some chess programmers (my colleagues) cared also.
At the time I didn't care much about all the hoopla of the magazine, so much nonsense about nothing. That changed when I discovered the internet in 1995 and in 1996 joined RGCC. And somewhere in 96/97 some of my colleagues started to attack me over the BT-2630 issue which came as a complete surprise and it wasn't funny at all, definitely I had underestimated the case.
The positive side of the story is that I never lost my fascination with DATA and the possible ELO in there, first with REBEL 10 and the EOC approach (only minor elo gain), later with ProDeo 1.84 (20-30 elo) and now finally a breakthrough with ProDeo YAT resulting in a 3 digit number ELO gain.
and the battle for freedom
The year is 1996 and after the release of Rebel 6 with the introduction of a database system we wanted to offer our customers something new, a giant collection of games (called MillionBase) so they could do something with the database functions.
In those days there were only 2 companies who sold games, ChessBase and New in Chess (NicBase) and because of their monopoly position they could ask considerable money for few games. But to speak with Bob Dylan - The Times They Are A Changin' in 1964 the same counted for 1996 and the almost exponential growth of the Internet each year. The days considerable profit could be made selling games were already in
decline due to the Internet enabling people to exchange their game collections by email or website. 2 other companies (Chess4Less I remember, the other I forgot) were already offering game databases at a much lower price. Simply put, the golden (ChessBase & NicBase) days were coming to an end due to competition.
Some Dutch guy contacted me saying he for a hobby collected chess games and that he had over 1 million. In that days this was an enormous amount and so I became interested but not after seeking juridical advice first and confronting him with the hard questions on copyright. Chess games are freeware, a careful edited collection of chess games is not, the games are still freeware, the editing work (player name consistency, adding elo, comments etc.) is protected by copyright law. On my question if he copied ChessBase and/or NicBase stuff he firmly said no and so I bought his collection for 6000 dutch guilders (around € 10,000 nowadays) and started to clean up the 1 million games (with DBUTIL), name consistency, header consistency, elo consistency, a massive job of a week of 5-6 because the database was one big mess considering the non-consistency of the above mentioned data fields. And that (the mess) was exactly what made me feel comfortable that the Dutch guy had spoken the truth, that it wasn't just a rip-off of the well organized commercial databases from ChessBase and NicBase.
Still we wanted to play it safe and so we sent both ChessBase and New in Chess a preview cdrom with the announcement we were going to release it in december 1996 together with the question if (after all) there were games carrying the sign of their work.
Their response was furious, suddenly the 2 competitors made an alliance and soon a letter from a german lawyer ended on my desk threatening to sue if we would release MillionBase. As evidence they provided a couple of game collections which according to them carried the sign of their editing work. I wasn't convinced but removed the games they referred to, told them so and released MillionBase anyway. Nothing happened, no sight-seeing trip to Hamburg.
Likely they saw my preview cdrom as a sign of weakness and with some lawyer violence were trying to protect their monopoly position. But MillionBase was a low quality but high quantity product at a reasonable price in a changing world, not really a threat for the true connoisseurs.
Some years later I bought 1.5 million high quality games from the Chess Assistance guys as support for new REBEL releases, when I retired in 2003/4 I made an update with 240,000 newer games (MillionBase 1.74) freely for download, that's how fast Internet has changed the database business.
Not so much later - and because of the trouble with MillionBase - we became befriended with New in Chess and we did some cooperation projects. Somewhere in 2000/1 we gave ChessBase a licence for producing Chess Tiger under their own brand. Bottom line, solving your problems is much better than giving someone you don't know the control over your life while paying high lawyer invoices to make that gamble even possible.
Men fight and make up, women never forget ;-)
Playing against Kaspy
After the disappointing news the AEGON tournament of 1996 was their last one it was time to do something new. REBEL ending 3 times as best computer during AEGON tournaments plus the games it had played against GM's made me realize the gap became closer and closer each year. And so the idea was born to challenge GM's to play matches against them and the first one was played against Arthur Yusupov (at the time listed as no.10 on the FIDE rating list) in 1997, the same year of the immortal Kasparov - Deep Blue match. REBEL was doing approx.100,000 NPS on the hardware of that time. One year later we played Vishy Anand (already second back then). REBEL played on special hardware doing 200,000 NPS. Both matches (with mixed time control) were won by REBEL, this was something hard to imagine and tasted for more.
And in 1999 that dream almost came true. A company with a considerable budget contacted me with the idea for a sort of super AEGON tournament.
Looking at their budget I suggested Gary Kasparov and they were excited about the idea of a Kasparov - REBEL match.
I made contact with the manager of Kasparov (Owen Williams) and he agreed to a meeting. And so it happened we met Gary in Amsterdam. It was an enjoyable afternoon, first the social part and Gary was very charming. That later changed discussing the match when it came to the heart the meeting, the playing strength, his passion awoke. As only true champions can do he asked me the rhetorical question if I thought REBEL could beat him. I smiled and said something like miracles do happen (showing him my respect, understanding his emotion while at the same time thinking, yes I can) and then continued in the same tone that the miracle wouldn't happen and that he would crush REBEL. I kept smiling, that was the Gary he was also famous for and now I had experienced it myself. Excellent show.
The good news was that we came to an agreement, also the money part looked good. All signs on green and it looked like a formality that Mr. Williams and the company in question also came to an agreement. Unfortunately that never happened and I don't know why and I don't need to know as such things are none of my business, but frustrating it was,
One more remarkable thing, Gary (like Anand) was surrounded by various people that travelled with him. I haven't seen that with the other 3 matches I played against GM's, there were all alone. I guess that only top-players like Kasparov and Anand (and maybe a few others) can afford that luxury. I can imagine the life of a professional chess player is the life on a lonely one, always travelling, sleeping in hotels, probably cheap ones.
REBEL vs DEEP BLUE junior
The year was 1999 and because Christophe Theron author of CHESS TIGER had joined the REBEL team we were able to participate with 2 programs (REBEL and CHESS TIGER) in the World Championship, location the University of Paderborn, Germany. Jeroen Noomen was also present to prepare the openings for each round. Despite the good preparation the end results were rather disappointing, both REBEL and CHESS TIGER ended with a 4½ out of 7 score and only became shared fifth. But there was a surprise waiting for us which became a big consolation, and fun as well.
Most days 2 rounds were played, one in the morning, the second midday. On one day both Christophe and me had quickly finished our games and we had to wait for 2-3 hours till the next game would start. So after taking some fresh air first we came back in the tournament hall and the waiting became already a bit boring.
Then my eye fell on a big screen in the tournament hall where for the occasion Deep Blue Junior [DBJ] was installed, the portable online little brother of the famous Deep Blue that beat Gary Kasparov 2 years before. And the screen was inviting you to play against it. And I said to Christophe, let's play against the beast. And so I went first with my poor laptop, a Pentium 333Mhz doing only 30,000-40,000 NPS against DBJ advertised as doing 20-30 million NPS or it must have been the demo version of IBM, still doing 2 million NPS but it was not advertised as such. It was there to glorify and promote the brand name of IBM, text on the big screen such as: Deep Blue der Sieger of Kasparov made that clear.
I played 3 games and surprisingly got 2½ of out 3. Next was Christophe, he was even more handicapped with a Pentium laptop of only 133Mhz and still managed to get 1½ (one win. one draw, one lost). Many people witnessed the 6 games and were astonished by the weak moves DBJ sometimes played.
And then when I published the 6 games on the REBEL site all hell broke lose. IBM by the mouth of its programmer became furious, started a public campaign to discredit what happened also contacted me pouring his anger on me with all kind of excuses why DBJ had performed so badly without even considering the option his brainchild maybe isn't so strong as he thought it was and that there are a bunch of other people who also can write a chess program and maybe better than him. Typical.
The discussion went to nowhere and he made it worse by demanding me to remove the games from my server. And I did, because I was tired of the emails and was beginning to lose my respect for the person responsible for the great achievement of 2 years ago and I did not want that.
Anyway, 18 years after, I see no reason any longer to hold back those games. History is history.
Download the 6 Paderborn games analysed by Jeroen Noomen