I started by choosing the Black Hole! I had decided all the best choices must be made early in the match in case another program discovered the same strong moves, or weakness in another opponent.

KingsRow took the 25-22 attack in game 1 but Nemesis defended cleanly to a quick draw. In game 2 Nemesis surprised with the 21-17 attack, planned because of the false pp in the KingsRow web page. However Kingsrow knew the right moves to defend and corrected some bad evaluations Nemesis had in its transposition table. At one point Nemesis had a +90 score but quickly returned to zero and the game was drawn.

In games 3 and 4, the 3-move ballot 10-14 22-18 6-10 was chosen by random draw. In game 3 KingsRow played an unusual defence that was solid and drew easily. Game 4 and KingsRow kept varying from pp but Nemesis kept transposing back, in an even game.

Martin chose 9-13 22-18 11-15 for the first 2 games. This surprised me greatly as Nemesis knew this ballot backwards and had almost beaten Al Lyman in the Banks Ladder with this. I thought Cake++ would struggle with the many difficult variations when defending the weak side but Cake++ handled things with ease, playing a new defence 6-9 on the 19th move that will cause me to rethink the best attack on this ballot. Neither program had any difficulty in the end.

In Games 3 and 4 I chose the Gemini I opening which I thought had a high chance of scoring against Cake++ if the false 8-11 defence was tried. In the event Cake played 16-20, a very tough defence. Nemesis knew that 16-20 was ok to draw and given a choice of 2 moves to attack, gained a strong position but could not force a win. Had the alternative attack been played, Nemesis probably would have won.

In game 4 Nemesis played the classically correct 7-10 defence and drew from book without effort, but Cake was showing an evaluation of +108 at one point and stayed very high for much of the game. Anyone observing the game could see that Cake++ did not understand the right attacks and defences in this ballot and needed to do some more work to play this correctly. This was of interest to Ed Gilbert of course who planned to take advantage of this observed weakness.

Martin chose the Tyne opening for games 1 and 2. Both programs played the Frazier (15-18) defence with ease, and I was so impressed by this that I immediately altered Nemesis to play the same defence (previously 11-16 would have been played). Little did I know that Cake++ and KingsRow had planned deep attacks against the 11-16 defence, which they hoped Nemesis would play.

In games 3 and 4 the Skullcracker was selected. This has some fine points and some good anti-computer lines of play. The downloadable versions of Cake++ and KingsRow do not play this opening very well, but here neither defence was troubled. It was apparent to me that the versions on Cake++ and KingsRow here were worlds apart from the downloadable versions.

The Garter Snake 9-13 23-19 10-15 was selected for games 1 and 2. Both programs sailed through their defences without problems. In games 3 and 4, 10-14 23-19 11-15 was selected, On move 11 Cake played an unusual defence that seemed perfectly sound. KingsRow played the more classical defence and was also untroubled. This theme continued throughout the match - Cake++ played the most unusual defences, Kingsrow played more published defences, and Nemesis always played published defences.

Excitement as KingsRow selected the Twilight Zone for games 1 and 2. KingsRow took the expected 7-16 defence that is its trademark and Nemesis saw some advantage but it gradually dissipated. KingsRow defended very well in one of its expert ballots. For game 2 Nemesis shocked everyone by playing the 7-16 defence back on KingsRow. Both Cake++ and KingsRow had planned deep attacks against the 8-15 defence but they were no use here. Nemesis played the weak side almost all from book, having largely used the same technique as Cake and Kingsrow to construct this book (dropout expansion).

In games 3 and 4 the Black Widow was played by random selection. Nemesis played the published 8-11 and 4-8 attack to pitch a piece and go man-down but KingsRow knew all the play. The 4th game was almost a repeat.

After studying the Black Widow games I decided to select the same opening against Cake++ with a real surprise in store. However after rushing the preparation somewhat, this caused Nemesis to jump the wrong way to my horror giving a completely equal game on move 11. The "surprise" was never to surface in this match. In game 2 Nemesis played the game from book but referee Mac Banks commented that this attack by Cake was more potent than the ones played in the previous 3 games.

For games 3 and 4 Cake++ chose the Skunk opening. Neithed program had difficulty with the 8-12 and 6-10 defences.

Cake++ chose Ryan's Cross 11-15 23-18 12-16 for games 1 and 2. Of course I had published a loss that an earlier version of Nemesis had played against Cake++ on this opening, and this loss had been corrected, it being a correction of Marion Tinsley and Don Lafferty. To be on the safe side Nemesis played the 7-11 defence, as did Cake++. In game 2 Cake tried a new attack and Nemesis had its worst evaluation of the entire match, a -9 score! However the draw was found easily enough.

In games 3 and 4 Cake++ I chose the black hole planning the 25-22 attack. Nemesis played the game out of book to draw the weak side easily enough. To my horror in game 4 Nemesis played the 21-17 attack , the 25-22 attack was planned. I had added 25-22 to the user book but didn't realise that 21-17 was also in the user book. In that situation Nemesis had a 50-50 chance of playing either move and unfortunately played the one I didn't want. Cake drew the game easily enough but I was disappointed that the chance of playing the stronger 25-22 was missed.

Ed had been studying the Gemini I game between Cake and Nemesis from earlier and decided to choose this opening again, considering a good chance of winning. It was the right choice and the only surprise to everyone was that Ed did not choose this opening earlier, in round 3 or 4. Martin had done some work on Cake++ to try and improve its defence but Cake still played the 16-20 defence. Kingsrow had prepared a different attack and broke through with a fine win. However later analysis proved that Nemesis had not missed a win in round 2. KingsRow was not troubled with the 24-19 attack in game 2.

In games 3 and 4 Cake++ chose the 9-13 22-18 12-16 opening, which has plenty of scope, but the programs seemed to play out the games without difficulty. So we have a leader in the tournament - KingsRow.

12-16 23-18 16-19 was drawn from the deck for games 1 and 2. Game 1 was easily drawn but in game 2 KingsRow played a new defence on Nemesis which seems to be sound.

For games 3 and 4 the Gemini 2 opening was selected. Nemesis tried the 27-23 attack rather than going for 22-18 and the Gemini 1 landing. KingsRow dealt with this easily enough. In game 4 KingsRow tried the 22-18 attack but Nemesis knows this opening very well and was not troubled.

10-15 24-20 15-19 was played first up and 2 quick draws ensued. Game 2 was the quickest game of the tournament and lasted only 2 minutes.

In games 3 and 4 the Tyne was drawn by random selection. After round 3 I had switched Nemesis over to the 15-18 Frazier defence which was a disappointment to Ed and Martin who had planned lines against 11-16 Like round 3, both programs found the draw easily enough making Frazier's defence the #1 choice.

Some discussion was made during these games about the alternatives to 15-18, such as 11-16, 6-9 and 6-10. Nemesis thinks that 11-16 and 6-9 are 100% safe to draw, but does not like 6-10 at all.

Since KingsRow had managed to defeat Cake++ in round 8 I decided to try the Gemini 2 opening on Cake here (Gemini I had already been selected before). Cake++ had the strong side on game 1 and Nemesis drew without difficulty. Game 2 and Cake showed that the problems had all been corrected, and played flawlessly this time.

Games 3 and 4 were from the 10-15 22-17 6-10 opening, and with the weak side Cake played a very weak looking defence which allowed Nemesis to open up the Cake++ double corner.. However Nemesis considered the move in question to be sound and after some discomfort for Cake a draw was seen. Post-Mortem showed Nemesis didn't miss anything better.

In games 1 and 2 the Black Widow was chosen for the 3rd time, and Cake++ played the same attack as was played in rounds 5 and 6, KingsRow of course drawing without breaking a sweat. With the red pieces KingsRow tried something new with 9-14 @ 5th but this led to an equal position, at least it was original.

9-13 23-18 6-9 was next up and KingsRow surprised with a new early defence that seemed perfectly sound. Both games seemed to be drawn without too much difficulty.

The Skunk (12-16 24-20 10-15) was chosen for games 1 and 2 and after a regulation draw in game 1, KingsRow played a most unusual 21-17 attack in game 2 but could not extract any advantage.

10-14 23-19 11-15 was next out of the deck and was not expected to cause a stir. In game 3 Cake played another non-standard defence that was a feature of this tournament, and once again it proved sound enough.

At this stage of the tournament KingsRow was the big favourite to win, as results were so hard to come by. However KingsRow found round 13 very unlucky, when in the 4th game an extraordinary event occurred that changed the course of the tournament. I and everybody else watched in disbelief as we approached the ending and Cake++ announced a win out of nowhere, with KingsRow announcing a draw. They played out to the 4x4 ending and it was a win for Cake++.

Somehow KingsRow had an error in its database, or database access code. For this tournament KingsRow used the Chinook 8 piece database that could be downloaded off the Internet. Cake++ and Nemesis had their own 8 piece databases that Martin and Murray had generated themselves independently (these two had been verified against each other.)

This result meant that the tournament was equal again. Interestingly, before the tournament it had been suggested that we played 3 points for a round win, 1 point for a round draw, and 0 for a loss. If we had agreed this, Nemesis would be in last position now, despite never having lost a game. It was Martin who argued that we should play 2 points for a win, 1 point for a draw and 0 for a loss, to eradicate any possibility of "Buddy-Buddy" systems whereby 2 systems could agree to win a round against each other to increase their overall scores.

In the event, we played 4 points for a win, 2 points for a draw and 0 for a win, so the tournament was now all square. We were heading for an 11-man ballot shoot-out.

Mac also informed me of another rule that was withdrawn. In the event of a tied match, the rules once stated that the program that lost the least games would win.

Now the only moment of controversy in the whole tournament. Ed wanted to use the Cake++ endgame database and access code for the rest of the tournament. His argument was that, having had a long-term programming relationship with Martin, he could have implemented the Cake++ database into KingsRow BEFORE the tournament, but chose the Chinook database instead, as the Cake++ database wasn't ready at that time. Martin was agreeable to give the database and access code to Ed. I wasn't happy that Cake++ could take advantage of an error in KingsRow's system which I couldn't take advantage of, that didn't seem fair to me. And then there was the issue of "Original Code". In the rules it said in section 5.4.1 that "Computer Systems are defined as a program of original code..." but obviously in this case Cake++ and KingsRow were sharing code and databases so this seemed to be in violation of the rules.

So it was up to the Referee Mac Banks to make a ruling, and he decided that in the spirit of the tournament, we wanted the best program to win, and so upheld the request for KingsRow to use Cake's endgame database and access code.

From my perspective, I didn't want a "cheap" win over KingsRow due to an endgame database error, but on the other hand I had invested heavily into the production of my own 8 piece database, both in terms of time, effort and hardware expense.

I had made 100% sure my database and access code was correct, by using verification techniques. If I had been able to use somebody else's database and access code, that would have given me 6 extra months to work on other aspects of my program. Therefore I felt that if any system were to use the Chinook database and access code, then they would also be liable to pay for any errors or problems using them, just like any other programming error that could cause a program to misplay a game and lose it.

It was a difficult situation but I believe Mac made the right one. Certainly both me and Martin had gained from Ed's database error, as we had both got back into the tournament by taking advantage of this deficiency in KingsRow. Ed had lost his lead, and was faced with a rapid programming challenge to patch in the Cake++ database. Although I felt somewhat uncomfortable with the idea of Ed using Martin's database. Nobody was really happy and it was an unfortunate event all round. However I respected the referee's decision.

The last round of the day did not give Ed enough time to patch in the Cake++ database, so he was playing KingsRow with potential problems in its handling of endgames. Naturally I was going to play out endings in case the same problem arose. This visibly agitated Ed, but I did not feel that it was the wrong thing to do. When Ed had spotted the weak defense of Cake++ in the Gemini (which Nemesis exposed) he attacked it and won a game. That's exactly what I would have done. I knew KingsRow had a weak point in its endgame, so I was going to play out games to 4x4. If we don't play to win, there is no point in taking part.

In fact I never played out any of these ballots to a 4x4 ending, they all completed with 10 pieces on the board. The ballots were weak 3-movers. Now Ed had the overnight break and completed his integration of the Cake++ database and so never had this problem again in the tournament. Thank goodness this day was over.

With the problems of yesterday behind us, Cake++ and Nemesis had their last two rounds together. The Wilderness I opening came out of the deck as a random draw. Mac had been pleading with us all week to select this opening as he thought it could well produce a win, but Cake++ was avoiding it (with good reason!) and KingsRow must have been saving it for the last round in case it was needed then.

In Game 1 Cake++ played a published attack and Nemesis drew without trouble. Game 2 and Nemesis played the 24-19 attack @ 4th for the surprise factor. It got a strong position and Cake was out of book and in difficulties, but pulled through for the draw.

In game 3 we played 11-16 24-20 7-11 and Cake++ surprised me with yet another original but sound defence, and a game that the observers greatly enjoyed watching. Cake++ also played an unusual attack in game 4 but it didn't give Nemesis any problems.

Between the morning and afternoon rounds I had to decide which ballot to select, I had one last choice remaining from my 6 choices, and I could use it against Cake++ in the next 2 games, or against KingsRow in the final round (18). Although Martin wanted to play the next round straight away I said I needed to do this preparation and so we would start at the regulation time (2.30pm). I had 2 hours and studied the previous Wilderness game intensely (as I'm sure Ed was doing at the same time).

In the end I concluded that I could improve the attack, but it wasn't enough to force a win. Cake++ had played very well and would probably see the draw. I decided to choose the Wilderness I again (remember the previous 2 games were from random card draw) but this time go with the standard 18-14 attack @ 4th. I noticed that 18-14 was marked as a "User book move" and so I must have been doing some preparation on this that I had forgotten about.

In games 1 and 2 I didn't get the choice of ballot so we played 10-15 23-18 6-10 and Cake shocked everyone with a very weak 11-16 defence, termed "ugly" by onlookers. Many observers thought Cake++ must lose the game from there but Cake++ found a neat draw. Nemesis saw a strong position for several moves before conceding the draw was forced and nothing was missed. Ed Gilbert added that KingsRow would also have played 11-16 in that position! It certainly seemed sound. There was no such excitement in game 2 as Nemesis played it by the book with 11-15.

So for games 3 and 4 I chose the Wilderness I and Cake had the strong side on game 3, the attack mainly published and not causing Nemesis any harm.

Game 4 was where the Computer Checkers Champion of the world was decided, and with a beautiful win too, I termed it a "Computer Buster".

Round 16, Game 4, Cake++ - Nemesis, August 8th 2002

9-13 22-18 11-16 18-14 (Not 24-19 this time) 10-17 21-14 6-9 23-18 7-11 (1-6 is published to draw, but leads to extremely long and complex lines of play. A computer with an auto-generated book would prefer 7-11) 24-20 (This is the computer-buster cook that wins the world title. Before now only 26-23 had ever been played in important play) 16-19 27-23 11-15 (Probably the game was lost here. 3-7* will draw. Nemesis and Cake++ both had a strong positive evaluation. Martin remarked "What's going on?" and at that point I didn't really know!) 23-16 12-19 18-11 8-15 14-10 4-8 20-16 2-7 25-22 7-14 16-11 8-12 28-24 19-28 22-18 14-23 26-10 9-14 11-7 5-9 7-2 12-16 29-25 14-18 2-6 16-19 6-2 (All prepared play by Nemesis to here, and in a won position) 9-14 2-7 3-8 7-3 8-12 3-7 19-24 7-11 18-23 10-7 23-27 32-23 28-32 7-3 32-27 23-19 27-32 19-16 12-19 11-16 19-23 16-19 32-28 19-26 14-18 3-7 1-6 7-2 (Nemesis wins)

With this win Nemesis was leading the tournament and if the remaining games were all drawn, Nemesis would be the world champion.

Martin didn't have a great deal of time to try and correct Cake++ from making the same mistake that was made against Nemesis. Some pretty desparate measures were taken in approximately 2 hours that remained before the next round.

Games 1 and 2, Martin chose the Double Cross, 9-14 23-18 14-23. Nothing exciting happened here as both programs (and Nemesis) know this opening very well. The stage was set for games 3 and 4 where Ed could choose the Wilderness. Of course, he did!

Games 3 and Cake++ had the strong side, playing 25-21 against the 6-9 defence and KingsRow defended perfectly.

So finally in Game 4 we had a titanic struggle. Martin decided to force the 16-20 defence instead of 6-9 to avoid the loss of the previous round. With little book play on this opening Cake++ was soon out on its own playing crossboard. This led to Cake++ selecting a weak looking move (11-15 @ 11th) and KingsRow was soon out of book too. Both programs showed a big advantage for white.

The game progressed slowly and still white had a huge advantage and everyone was expecting KingsRow to score the win, except Ed who was more conservative and commented that the score wasn't increasing so a draw was more likely.

I thought Cake++ must surely have the draw sewn up but on move 29 jumped 6-13 instead of 5-14 and seemed to allow KingsRow another attacking opportunity. WIth both programs crossboard chances were perhaps missed for better play on both sides but it was very exciting. We ended in a 10 piece bridge ending that Cake++ had just enough defensive resources to draw, but only just. That was in reality KingsRow's last chance to get back into the tournament. Now there was just the final round between Nemesis and KingsRow. All Nemesis had to do was draw 4 games to be champion, but Ed had a ballot choice left, whereas I had used all my six choices. Actually Ed had 2 ballot choices left, but only 1 round remained. That could have been a costly error. Also after the game it turned out that if Ed had chosen the Wilderness 2 opening rather than the Wilderness 1, he could have forced the same line of play with which Nemesis won in round 16. Cake++ had prepared a variation but in my opinion it was too late to avoid defeat.

We had an overnight break before this final round and I was wondering which ballot KingsRow would select for games 1 and 2. I thought perhaps he would choose the Black Hole as there are so many attacks, despite the fact we already played this in round 1. I did some searches and found that some previously safe-looking endings in the Black Hole were critical, almost like the "Rattlesnake" endings. I found 2 attacks that were causing problems, enough to make me fear the opening. When we got to the hall I was relieved to hear Ed choose the Wilderness 1 opening. I Hadn't had time to add the correct play to the Black Hole (the lines of play do draw with care, I subsequently discovered).

We played out a quick draw with Nemesis on the strong side of game 1. Not expecting this opening I hadn't planned a custom-prepared attack for KingsRow. In game 2 KingsRow played the 18-14 and 22-18 attack, Nemesis replied with the 7-11 defence that was the downfall of Cake++ in round 16. However Nemesis played out this defence correctly. KingsRow then played a pitch to go man-down. It looked dangerous but Nemesis had seen it before and knew the next 3 moves, showing a small red advantage. It was a good prepared line by Ed but not enough to score. We were down to the last 2 games.

I had to choose a random selection from the deck and pulled out a really easy ballot, 11-15 24-20 12-16. ("Easy" compared with many of the others in this tough deck.) KingsRow played the strong side first and immediately varied from published play with a fine 23-19 move. Thinking crossboard on move 7 Nemesis saw a database draw and that's how the game finished but it was a good strategy from KingsRow. In game 4 KingsRow played a defence that was weak and was getting a negative evaluation for a few moves (Nemesis correspondingly saw a good position) but not enough to score a win.

When the advantage had disappeared I looked up to Ed and offered a draw. He graciously accepted and offered his congratulations. Referee Mac Banks immediately began preparation for the presentation. Nemesis became World Computer Checkers Champion!