Pawel Kobylarz, 30, from Gdynia, Poland. Author of the EGM 0.1 program.

Greg Strong, from Washington, DC. Author of the Chess V program.

Jeff Mallett & Mark Lefler, co-authors of the Zillions-Of-Games engine.

Bill Angel, author of the CapaGNU Modified program.

Uwe Auerswald, author of the Max Gothic program.

While Gothic Chess is a relatively new game (invented in the year 2000), programs with the ability to play on 80-square boards have been around since the days of the 16-bit DOS operating system. The "old man" of the group is clearly Bill Angel's pioneering effort, CapaGNU, which was designed to play Capablanca's 80-square chess variant. Using the program's "setup" feature, we were able to change its starting position to play the game of Gothic Chess for this tourament. For that reason, it was renamed CapaGNU Modified to distinguish it from its original incarnation.

The web page describing CapaGNU (http://www.chessvariants.org/programs.dir/capaprogdesc.html) was written at a time when the 80386 was a top-of-the-line computer system. From what I know of the branching factor of Gothic Chess, and chess searching innovations in the software world, I think it is safe to say after reading its description that the CapaGNU program is closer in performace to DOS software programs of the late 1980's than any of the previously contemporary 1990's counterparts.

If we fast forward a little bit from here to the 1990's, we encounter the "Age of Zillions". The Zillions-of-Games software engine is really an incredible innovation. Zillions can play any game that requires a search, such as Chess, Checkers, Go, Tic-Tack-Toe, Connect Four, you name it. It does not need any specifics about the game, just a "script file" that identifies the description of how to perform every legal move. You don't need to be a programmer to understand this language, nor do you even need to assign values to the individual pieces in a game. The software interpreter automatically assigns these values after it parses over the "rules" of the game that you create in the little script file.

The chess variant community was very familiar with the Zillions engine, and all of them had high praise for its ease of use and play. Someone entered Zillions into the tournament, after having spent only 15 minutes developing its "rule file". But how strong was it? This tournament would help answer that question in an objective setting.

With tournament entrants from the 1980's and 1990's, we enter the first decade of the 21st century with Gothic Vortex, TSCP Gothic, Chess V, S.M.I.R.F, a program created just for this tournament EGM 0.1, and a peer-to-peer application, Max Gothic. These programs were all started well after the year 2000, and they represented the potential to be the strongest competitors.

So, here we have a unique tournament, with participants spanning a chasm of technological epochs. A late 1980's DOS program, a 1990's interpreter search engine, and dedicated applications compiled in the first decade of the 21st century. You would expect all of the 21st century entrants to stomp their antiquated counterparts, right?

The time controls for this event were Game in 90 minutes +5 seconds per move. This would insure rounds completing within about three hours. Since some of the programs did not have the capability to play with such time management, we had a decision to make. How long should we let these programs search on each move? If we set them at 2 minutes per move, they would only make 45 moves, and they would lose on account of a time forfeit on move 46 every game. We tested Zillions and CapaGNU Modified, and they stick with the move they make at 60 seconds even if they are left to search as long as three minutes in over 90% of the cases. So, it was decided to set them at 60 seconds per move, which would guarantee they would play further into the game.

Although this was a dual round robin, and every program would play every program twice, it was decided before play began to "seed" the entrants according to their observed performance at some benchmark that would not betray anything about their playing style or other technologies incorporated within the software. (The rankings that were produced mirrored the final standings perfectly at the end, so I thought I had a pretty good idea as what to expect.) This way, we can avoid having the #1 seed play the #2 seed early on and take all of the suspense out of the tournament.

The strongest program (seed #1) was paired with the weakest program (seed #8), seed #2 vs. seed #7, etc., and in each subsequent round the top seed would play progressively stronger programs. This way, the last two games of the match would be #1 vs. #2.

Given this approach, I did not anticipate any early round upsets. Gothic Vortex mated CapaGNU Modified in 35 moves, TSCP Gothic mated EGM 1.0 in 30, so I was walking around to watch the other remaining games.

Brian Colgan operates the S.M.I.R.F. program against Serena Parker, sitting at the console for Zillions.

When I stopped at the Zillions vs. S.M.I.R.F. board, I was surprised that only 11 moves had been played! What was happening? According to Brian Colgan, who was operating the S.M.I.R.F. program, S.M.I.R.F. was taking almost 8 minutes or longer to make a move early on in the game, then 7 minutes, then 6 minutes, etc. S.M.I.R.F. had used up over an hour of time to get to move 10, while Zillions used only 10 minutes! S.M.I.R.F. showed less than half an hour left for it and 79 minutes left for Zillions, a huge disparity in time. The position I saw is shown below on the far left after 1. Nc3 Nh6 2. e4 g6 3. g4 Nc6 4. g5 Ni4 5. d3 Be5 6. f4 Bd4 7. Af3 d6 8. j3 Nj6 9. Nh3 f5 10. exf5 Bxf5 11. Ah4 Ae6

What happened next was amazing. After a few more moves, S.M.I.R.F. really started blitzing off moves, each one more quickly than the next. The middle position above was reached just a few minutes later after 12. Ni5 Qd7 13. Ri1 Ag7 14. Ce2 a5 15. Bf3 a4 16. Bxc6 Qxc6 17. Qe1 a3 18. Ag2 Qb6 19. bxa3 Bxc3 was played. Now you see two types of exchanges being threatened, the "standard chess" exchange of Bishop for Rook, and also the "Gothic Exchange" of Archbishop for Chancellor should white play Cxc3. A very interesting battle has taken shape, with trades flying off the board 20. Cxc3 h6 21. gxh6 ixh6 22. Nxg6+ Bxg6 23. Ah4 Kf7 24. Axj6 Axc3 25. Qxc3 and you need a scorecard to figure out who is ahead. S.M.I.R.F. eventually pulled out the win after playing another 43 moves in under 10 minutes!

25...Ri8 26. Af2 Qa5 27. Qb3+ e6 28. Ag4 Qj5+ 29. i4 Qh7 30. f5 Bxf5 31. Axh6+ Qxh6 32. Bxh6 Cf6 33. O-O-O Cxh6 34. Qxb7 Ric8 35. Rdh1 Ch3 36. Ri3 Cf2 37. Qc6 Rxa3 38. Kb2 Raa8 39. Qd7+ Kf6 40. Qc6 Rcb8+ 41. Ka1 e5 42. Rc1 Cf4 43. Qc4 Cxc4 44. dxc4 Rb4 45. Rf1 Rba4 46. Kb1 Rxa2 47. Rif3 Ra1+ 48. Kb2 R1a2+ 49. Kb1 Ra1+ 50. Kb2 R8a2+ 51. Kb3 Ra3+ 52. Kb4 Rxf1 53. Rxa3 Rb1+ 54. Rb3 Rh1 55. c5 d5 56. Ra3 Rxh2 57. Ra7 Rh7 58. c4 d4 59. Ra6+ Kg7 60. Rc6 Kg8 61. j4 d3 62. Ra6 e4 63. Kc3 Rh2 64. Ra8+ Kh7 65. Rd8 Re2 66. c6 Ra2 67. Kb3 Rg2 68. Rd5 d2 0-1

I must admit, after seeing the way S.M.I.R.F. handled most of that game in severe time pressure, I had to wonder how it would play if it adjusted its settings to play a more balanced game. This game made me feel a little uneasy.

I thought that was going to be all of the excitement for the first round. What I did not realize was that the Chess V vs. Max Gothic game had long since drawn to a close, as... a draw! "What ?!?" I remember saying out loud as I ran over to their board. Draws are as rare in Gothic Chess as Solar Eclipses are to the outside world, so where there is one to behold, it is usually something of a curiosity. You see, draws in Gothic Chess are not long, weary battles. They came about more often as a result of one side winning some material, then furthering the acquisition, then a crafty player pulls a forced repetition of the position using one or both of the new piece types.

Here is Chess V vs. Max Gothic from round 1.

1. g3 d5 2. Nh3 Nh6 3. e4 d4 4. Bf3 c5 5. Ae2? Bxh3+ 6. ixh3 Let's just say for now this is not a recommended way to open your game in Gothic Chess. Already we see that Chess V has doubled h-pawns without an i-pawn, the equivalent of having doubled f-pawns with no g-pawn in regular chess. Kingside castling is definitely ruled out at this early stage.

6...e5 7. Ad3 Ae7 8. Axe5 Ag5 9. Af4 Axf4 10. gxf4 And now white has doubled f-pawns as well. Black's position looks like tranquility and peace compared to the "war zone" look of the tattered White camp. As a footnote, we mention that Black had a tactical shot with 8...Axj2 9. Rxj2 Cxe5.

10...Qh4 11. O-O? Black makes a lunge with the Queen, threatening h3 with check, and White goes from the frying pan into the fire by castling directly into an attack. 11. Bg2 was more patient in this damaged position.

11...Qxf4 12. d3 Qj4 13. Cg2 Ce5 14. Kj1 Cj5 How is White going to survive? There is only one move to postpone instant disaster. This move, Bg5, would have been so much better on move 14, as it would have disrupted the kingside slide of Black's Chancellor. But now, for one side to have two majors so aggressively poised while having two of their own minors undeveloped is a sure sign that somebody has made more than one inaccuracy. Max Gothic gets high marks for making the most of the opportunities.

15. Bg5 Ci5 16. Ch4 White could have mended some fences with 16. Bi3, turning the Bishop into a surrogate sheltering pawn. 16...Qxh4 17. Bxh4 Cxh3 18. Bg3 Ni4 19. Ki2? Cxj2+

The idea behind 18. Bi3? comes a little too late now, as Black would have had 18...Cxf2, winning the f-pawn and threatening the Queen. Instead of 19. Ki2? White had 19. Bd6+, not just a "spite check" since it was better than moving the King to no avail, or 19. Bg2 which would secure the weak h3 square. After 19...Cxj2+ Black can force a draw now at will by repeating the position, but there is a much better way to proceed. Perhaps time was a factor and Max Gothic elected to take the quick way out. After 20. Ki1 Ch6 is a logical destination for the Chancellor, with ...Nxg3+ as one path that leads to victory.

20. Ki1 Ch3+ 21. Ki2 Cj2+ 22. Ki1 Ch3+ 23. Ki2 Cj2+ 1/2-1/2

Gothic Vortex cruised past CapaGNU Modified in this round, turning in a 31 move checkmate, so I was relieved by this early result. TSCP Gothic mopped up EGM 0.1 quickly, and the operator resigned on move 20. TSCP Gothic was still winning in fewer moves than Gothic Vortex, and I was trying to convince myself that this did not matter. Chess V turned in a stellar result, sweeping aside Max Gothic in only 20 moves as well. As these games were all over in short order, all eyes were on the S.M.I.R.F vs. Zillions game, which had a tragic consequence as a finish:

1. Nc3 Nc6 2. d4 e6 3. Nh3 Nh6 4. j4 g6 5. e3 f5 6. i3 g5 7. Bj3 Cd6 8. Ae2 g4 9. Nb5 Cd5 10. c4 Cb4 11. Nf4 Ae7 12. d5 Ne5 13. a3 Ca6 14. Bd2 Ag5 15. Bb4+ Kg8 16. d6 cxd6 17. Nxd6 Nc6 18. Bc3 Bxc3 19. Axc3 e5 20. Nd3 e4 21. Nf4 Kf8 22. b4 Ni4 23. Kg1 Ne7 24. Qd4 Rg8 25. b5 Nxh2 26. bxa6 Axi3 27. Nxc8 Nf3+ 28. Kf1 Nxd4 29. Axd4 Qxc8 30. axb7 Qxc4+ 31. Ce2 Rb8 32. Ae5 Qxe2+ 33. Kxe2 Rxb7 34. Rjb1 Rxb1 35. Rxb1 Nc6 36. Axd7+ Kg7 37. Axc6 Kh8 38. Ad4+ Ki8 39. Ag7+ Kj8 40. Axf5 Af6 41. Nd5 Ag5 42. Axg4 j6 43. Nf6 Rf8 44. Bh5 Ki8 45. Rb5 Ah6 46. Axh6+ ixh6 47. Nxe4 Rc8 48. f4 Rc2+ 49. Kf3 Rc6 50. f5 a6 51. Rc5 Rb6 52. f6 Rb8 53. f7 Kh8 54. Nf6 Kg7 55. Nd7 Ri8 56. Rc4 Rd8 57. Rf4 Kh8

This leads to the diagram shown below:

Pawel Kobylarz, the EGM 0.1 programmer, sent an updated program prior to the start of the round. Tragically, this version played moves resembling a mini-max coding error (selecting the move leading to the worst material loss). On move two EGM 0.1 loses its Archbishop for a mere Bishop, and on move four, it gives up its Chancellor for a Bishop. In a strange twist of irony, once it was down a good bit of material, it started playing a somewhat respectable defense, in a manner of speaking. I did not understand this at first, because I had seen the program's earlier games and it was committing a few grave errors in a positional or strategic sense. Then I realized this is where the reversed sign of the mini-max would actually help out. If a program that is making "bad evaluations" all of sudden scores them "the opposite" of what it would normally, what precipates out of the search could be a series of good moves. This is a game that Seinfeld fans could appreciate, calling to mind the episode where George Castanza decides to "do the opposite" of whatever he would normally do, and the results of his various escapades improve substancially. Pawel was notified of this right away, and he was able to fix this error. Now Reinhard had some company to console with at least, meanwhile Gothic Vortex was able to deliver a mate in 13 moves.

1. d4 Ah6?? 2. Bxh6 gxh6 3. Nc3 Cf6 4. g3 Cc6?? 5. Bxc6 dxc6 6. Ng2 Qxd4 7. Qxd4 Bxd4 8. Cd1 Bxc3?? 9. Cd8+ Kg7 10. Ce8+ Kg6 11. Nf4+ Kg5 12. Af3+ Kf5 13. e4# 1-0

The real eye-opener in this round was the win by CapaGNU Modified over Zillions. The CapaGNU Modified program is suffering from several severe handicaps in this event. For one, it is playing a game it was not originally designed to play! Take a look at the figures below.

On the left, you can see the game that Jose Capablanca invented around 1924. This is what CapaGNU was designed to play. You can see in blue that there is an Archibishop, Bishop, and Queen all in a row on the queenside. These three pieces all have diagonal movement capabilities, and they are each pointed in the area of the pawn shown in red. That i-pawn is not defended by any of the black pieces in the starting configuration. This is the "Achille's Heel" of Capablanca's Chess, and perhaps the reason why the game never gathered a strong following. The undefended i-pawn would be under attack on the first move of the game after 1. d4, a simple pawn push. Furthermore, those three diagonal pieces impose an asymmetry of sorts. There is no "balance" from one side of the board to the other. Compare this to the right figure of Gothic Chess, which has two diagonal pieces on each half of the board (Bishop/Queen and Archbishop/Bishop) and all of the pawns are defended adequately.

So, you can imagine, the original CapaGNU might have some instructions governing the strengthening of its defenses around the weak i-pawn, or perhaps some code to look for ways to provoke an i-pawn weakness in the opponent's position. Any game-specific instructions that it utilizes would be a detriment in playing Gothic Chess. That being said, we present the upset from the third round.

1. Nc3 f5 2. e4 Nh6? 3. exf5 Nxf5 4. g4! Nd6 5. Cf3+ Cf6 6. g5 Cxf3 7. Qxf3+ Af7 8. Nh3 g6 9. Ni5! Kg8 10. h4 j6 11. Qj3 Nf5 12. i3

The game of Gothic Chess is very deep, so programs will still lose there way throughout much of the thread in the opening phases. Phantom attacks look good until the horizon is pushed beyond their tactical completion, sometimes after one has already committed forces to a battle which should be foresaken immediately. As such, we sometimes see pieces left in awkward dispositions.



12...Bxc3 13. dxc3 Ac4+ 14. Kg2 d6 15. Af3 Nc6 16. Bi2?? The start of a downward spiral for White. There was a need to initiate some counterplay with 16. Nxg6?! Nxh4+! 17. Axh4 Bxj1 18. Nxi7! to avoid the passive cascade of positions that soon follow. It is easy to anthropomorphize the play of CapaGNU Modified in moves to follow as being strategic in nature, reducing the scope of White's play, then engaging in swaps that left it clearly ahead. 16...Ne5! 17. Ae4 d5 18. Ad2 Axd2 19. Bxd2 d4! Black clears a path to d5 for the Queen, which White must deal with sooner or later. 20. Kf1 Nf3 21. Bf4 e5 22. Ke2? N3xh4!? This triggers a wave of blatant attrition, all of which favors Black heavily. CapaGNU Modified had play leading to many interesting paths, such as 22...exf4! which appears to be the strongest. 23. ixh4 exf4 24. Rjh1 dxc3 25. bxc3 Qe8+ To paraphrase Fermat: "I have discovered that 25...Ng3+ is an interesting side variation that this margin is too narrow to contain." 26. Kd2 Ng3 27. Bxc8 Nxh1 28. Qxh1 Rxc8 29. Qd5+ Personally, I would have taken care of the hanging Knight first with 29. Nh3, since ...Rd8+ is not any more fatal that the current disarray. 29...Kh8 30. Qd4+ Ki8 31. Re1 Qf8 32. Nxg6?! hxg6 White has the right idea, but makes an incorrect execution. 32. Nxh7 forces the King to take back with 32...Kxh7, allowing 33. Qd7+ which gives White time to unpin his own King from the Queen in the d-file since 34. Kc1 comes next "for free". 33. Qxa7 Rd8+ 34. Kc1 Qh8 35. Rd1? Rxd1+ White should have hung onto its only Rook. 36. Kxd1 Qxh4 37. Qxb7 Qxg5 38. Qxc7 f3 39. Qc6? Qg1+ The Zillions move here does not make too much sense. Why not 39. Qg3 to take Queens off the board? The game is still dead lost, but an ending with just one major piece would have to be the preferred way to drag things out. 40. Kd2 Qxf2+ 41. Kc1 Kj7 42. Qc4? Qe3+ 42. Qb7 would have avoided the mate in 9 that White walks into, and Black does play the move leading to the quickest mate. 43. Kb2 Rb8+ 44. Ka3 f2 Another mate in 6 for Black is 44...Qa7+ 45. Qa4 Qxa4+ 46. Kxa4 f2 47. c4 f1=Q 48. Ka5 Qg2 49. Ka6 Qa8#. 45. Qg4+ Ki8 46. Qb4 I think there is little point in mentioning that 46. Qxg6+ could have prolonged the game one move longer. 46... Rxb4 47. Kxb4 f1=Q 48. j4 Qb6+ 49. Ka4 Qfa6# 0-1

The game ends with a rare disambiguating Queen move made by CapaGNU Modified. The old DOS program conducted the late endgame stage with precision, never making an error once the mate was assured. After seeing this, I stopped allowing my contemporary world view of the artificial intelligence world to cloud my preconceived notions about the relative strengths of the various programs. This new "open minded" and "opinion free" version of the Gothic Chess programming world was immediately positively reinforced. My polite reverence for CapaGNU Modified was destroyed when Zillions thrashed it soundly from the other side of the board.

How could both of these programs play so well with Black and so horribly with White? I stopped looking for answers. In the game to follow, with 12. Nxh7+?? CapaGNU Modified cannot even be said to be attacking anything, and there is no explanation for the incredibly weak 13. Ae6+?? which should deserve a third '?' if there were an acceptable standard for such a designation. There are some tactical exchanges later that are interesting, where Zillions parts with the Queen but gets plenty of compensation for it. Then Black plays the odd-looking 31...Bg1! and is able to announce the mate in four on move 34 after about 3 seconds.

CapaGNU Modified vs. Zillions: 1. Af3 Nc6 2. Nc3 Nh6 3. Nh3 f6 4. d3 Ne5 5. Ae4 d5 6. Ag3 d4 7. Ne4 j5 8. Ni5 g6 9. Af4 j4 10. Ri1 Cg7 11. Bd2 Ci6 12. Nxh7+?? Axh7 13. Ae6+?? Bxe6 14. h3 Qd5 15. g4 Qb5 16. a4 Qxb2 17. Rb1 Qa2 18. Ra1 Qd5 19. Nxf6 Bxf6 20. Bxd5 Bxd5 21. Bf4 g5 22. Bg3 c6 23. c4 dxc3 24. Qb1 O-O-O 25. e4 Bf7 26. Ce2 Rxd3 27. Bxe5 Bxe5 28. Qxd3 Bxi1 29. Cg1 Bh2 30. Cf3 Cxi2 31. Re1 Ag6 32. Ke2 Rd8 33. Qxc3 Axe4 34. Rj1 Bg1! 35. Qe1 Bc4+ 36. Cd3 Axd3+ 37. Ke3 Axf2+ 38. Qxf2 Cg3# 0-1

In the later rounds, the games feature the higher seeds duking it out. Just when you thought the tactics in prior rounds were intense, the Master Class begins to engage and produce fireworks that have never been seen before.



An amazing demonstration of S.M.I.R.F's new tenacious time control code is showcased below. TSCP Gothic builds pressure from the opening, winning S.M.I.R.F's Queen for a Knight + Rook after 22. Ne6! Nxe6 23. Rc1 j5 24. Aj3 Nc7 25. Rxc5 dxc5. S.M.I.R.F battles back and minimizes the gains made by TSCP Gothic, all the while moving faster and faster. By the time TSCP Gothic makes its very late kingside castle (40. O-O) the balance had already shifted in its favor, but S.M.I.R.F dug in again and puts up an amazing defense given the speed at which it moved. TSCP Gothic starts to make wandering moves in the endgame despite completing 13 ply searches, failing to advance its pawns or put adequate pressure on S.M.I.R.F's King. As the operators blitz off moves, both flags were noticed down on the same move, mandating a drawn result.

TSCP Gothic vs. S.M.I.R.F.: 1. c4 Nh6 2. Nc3 c5 3. Nh3 Nc6 4. d3 d6 5. Bd2 Cc7 6. Nd5 Ca6 7. i3 g6 8. Qc1 Be6 9. Bi2 Bxd5 10. cxd5 Nb4 11. Nf4 Ca4 12. Cd1 Qb6 13. Ah3 f5 14. a3 Na6 15. Rb1 Nc7 16. b3 Ca6 17. b4 Ng4 18. Ai4 O-O-O 19. h3 Nf6 20. bxc5 Qxc5 21. Qb2 b5 22. Ne6! Nxe6 23. Rc1 j5 24. Aj3 Nc7 25. Rxc5 dxc5 26. Qc2 Axd5 27. Ah4 b4 28. axb4 cxb4 29. Cc1 b3 30. Qxb3 Axb3 31. Cxb3 j4 32. Af3 Nfd5 33. ixj4 Rxj4 34. e4 Nf4 35. exf5 Cb6 36. Cxb6+ axb6 37. h4 Nfe6 38. fxe6 Rxd3 39. Be3 b5 40. O-O Rc3 41. g3 b4 42. Rb1 b3 43. Ae4 Ri4 44. Bd2 Rc4 45. Ag5 Rxi2+ 46. Kxi2 Bf6 47. Axh7 Rc2 48. Be1 Re2 49. Axg6 b2 50. Axi7 Rxe6 51. Ag6 Be5 52. Ad3 Bh8 53. Bb4 Ri6+ 54. Kj1 Ra6 55. Bxe7 Ra1 56. Af5+ Kb7 57. Ae4+ Kc8 58. Bb4 Nb5 59. f3 Bg7 60. f4 Rxb1+ 61. Axb1 Kb7 62. Ba5 Ka7 63. Bd2 Kb6 64. Ki2 Kc7 65. g4 Kb7 66. Bb4 Kc7 67. Ad3 Kc6 68. Bd2 Nd6 69. Be3 Nc4 70. Ae4+ Kd7 71. Bc5 Ke8 72. Ag6+ Kd7 73. Ab1 Kc6 74. Bg1 Nd6 1/2-1/2

In this next game Max Gothic establishes the "broad pawn center" on d4-e4-f4 and Gothic Vortex plays in the early 20th century hypermodern style. After 6. e4 Qd8 we have an even game, and and interesting theoretical position. The pawn push 7. d5 is premature allowing Black interesting replies with either 7...Bg4! or 7...Bd4, both using Knights as forward staging areas for Bishop strikes against major pieces. Gothic Vortex finds 8...Nd4! and White takes the "poison pawn" on b7 with the Queen, completely turning over the initiative to Black. With 10. h3 the Black Bishop is hanging, but 10...Cd6! gets a fifth Black piece attacking. The Black Chancellor pries the White Queen out of the b7 hole, at the same time gaining momentum. After 12...Axe4 White still can't take the hanging Black Bishop, or else Ag3 checkmates. With 13...Be2+ we see the beginning of the end already. Black gets the White Archbishop for its own Knight + Bishop, always a favorable trade as the Archbishop is worth much more than the sum of its components. 18...e5!! was unexpected and a brilliant move, as 19. fxe5?? would be met with 19...Qh4!! 20. Bxh6 Ae4 and a mating attack against g3. Another amazing move was 22...Ab6!! which is played with the Knight on f5 hanging. The point of the move is supporting ...Bf2 with the trivial mate ...Bg3 to follow if the threat is not handled. The endgame mate is also rewarding, as 40...Rxg2+ is not apparent upon first glance.

Max Gothic vs. Gothic Vortex: 1. d4 Nh6 2. c4 g6 3. Na3 d5 4. cxd5 Qxd5 5. f4 Nc6 6. e4 Qd8 7. d5? Bg4! 8. Qb3 Nd4! 9. Qxb7 Af6 10. h3 Cd6! 11. Qb4 Cxe4 12. Cxe4 Axe4 13. Af2 Be2+ 14. Kg1 Axd5 15. Qe1 Bg7 16. Kh2 O-O 17. Axd4 Bxd4 18. Qxe2 e5!! 19. g3 e4 20. Ng2 Re8 21. Ki1 Nf5 22. g4 Ab6!! 23. gxf5 Bf2 24. Kh2 Ad4 25. Qg4 h5 26. Be3 hxg4 27. Bxd4 g3+ 28. Ki1 Qxd4 29. Rb1 gxf5 30. Nb5 Qd3 31. Na3 Rab8 32. b3 Bc5 33. Re1 Bxa3 34. Re3 Bd6 35. i3 Qd2 36. Rxg3 Bxf4 37. Nxf4 Qxf4 38. Kh2 Rbd8 39. Bxe4 Rd2+ 40. Bg2 Rxg2+ 41. Kxg2 Re2+ 42. Kh1 Qxg3 43. h4 Qh2# 0-1

TSCP Gothic gets revenge for allowing the draw in the previous round. From a quiet "Four Knights" opening we see S.M.I.R.F. commit the first inaccuracy with the unnecessary 5. a4 pawn push. This loss of time allows Black to get in the first attacking volley with 6...Nf5, threatening the Bishop on e3 while revealing Black's Bishop on i7. After 7. Bxi7 Axi7 we see the unusual "fianchettoed Archbishop" which proves to be deadly later. With 10. Ng5 White launches an unprepared attack against the uncastled Black King, and the curious looking 10...Aj6 is the proper way to hold the weak h7 square. After 11. Kf2, Black gets the easier game, although 14...Kxe8 may take some by surprise when 14...Qxe8 was possible. After 16...h6 TSCP Gothic was reporting a large score, and it mopped up S.M.I.R.F. in short order thereafter.

S.M.I.R.F. vs. TSCP Gothic: 1. Nc3 Nc6 2. Nh3 Nh6 3. d3 i6 4. g3 Bi7 5. a4 d6 6. Be3 Nf5 7. Bxi7 Axi7 8. Bxc6 bxc6 9. f4 d5 10. Ng5 Aj6 11. Kf2 e5 12. e3 d4 13. exd4 exf4 14. Cxe8+ Kxe8 15. Qe2+ Kf8 16. Nge4 h6 17. Qd2 fxg3+ 18. hxg3 Nxg3 19. Nxg3 Qxd4+ 20. Kg2 Ah4+ 21. Kf1 Axg3+ 22. Kg2 Qg4 23. Ac5+ Ad6+ 24. Kh1 Qh4+ 25. Kg1 Axc5+ 26. Kf1 Ad4 27. Ne4 Bh3+ 28. ixh3 Qxh3+ 29. Qg2 Qxj1+ 30. Qg1 Qxg1# 0-1

Gothic Vortex vs. Chess V: 1. d4 g6 2. Nh3 Nh6 3. Nc3 Bg7 4. g3 Af6 5. Bg5 The first act of aggression. White provokes Chess V's Archbishop, enticing it will the "hanging" d4 pawn. Notice in the leftmost postion below, Black's Archbishop is the only piece defending the h7 square. Gothic Vortex exploits this in the extreme. 5...Axd4? 6. Ni5! Kg8 If 6...0-0? 7. Bxh6 and Black cannot recapture with the Bishop since it is defending the Archbishop on d4. If 7...ixh6 White wins easily with 8. Nxj7!!, exploiting the fact that a pawn push to e3 would reveal a Queen's check if the King recaptures the Knight with 8...Kxj7?? and at the same time the e3 pawn threatens the Archbishop on d4. After 8...Rf8 White still can take advantge of the Archbishop needing protection and play 9. Nxh6+! winning easily. Black should have played 5...Ag8 to have avoided this whole mess.

7. Nxj7 This diagram is shown above in the middle position. Such a position has nothing remotely similar in our chess "pattern recognition database" so it looks very unusual. The point is the same as mentioned above: the pawn push e3 would reveal the Queen's influence over h7 while hitting the Archbishop. 7...Nxj7 8. e3 Axc3 Or 8...Ae6 9. Bxi7 Ac4+ Cd3 with interesting play, but still favoring White. 9. bxc3 Bxc3 10. Cd3 Bxa1 11. Bxi7 Ri8 12. Qxj7 As a friend of mine used to say, it's not over, but "...the fat lady is warming up her vocal chords." 12...Rxi7 13. Qxi7 Bh8 14. Qj8 b6 Black is "threatening" ...Ba6 to pin the Chancellor. 15. Af3!! Kg7 16. Qj4+ Announcing mate from here. 16...Kg8 17. Ag5 Cg7 18. Ah6+ Kf8 19. Axg7+ Bxg7 20. Qj8+ Bh8 21. Qxh8# 1-0

S.M.I.R.F. programmer Reinhard Scharnagl sends in a new version of the program, and it plays much better than previous versions. This was the most exciting game of the event so far.

Gothic Vortex vs. S.M.I.R.F.: 1. c4 g6 2. Nh3 d6 3. Nc3 a5 There is no reason for this move to be played by Black here. 4. g3 Nc6 5. Bd5 And Gothic Vortex is not above reproach either. This is what I have identfied as the classical "Bishop lunge" problem. Gothic Vortex assigns mobility bonus points for Bishops that have more legal moves, so it sometimes places them where an enemy pawn can just chase them away (in this case 5...e6 does the trick.)

5...Nh6 6. d3 Nd4! Black gets very good Knight placement with that move. 7. Bg5?! Bf6! White does another "Bishop lunge" and Black handles the position perfectly. By diffusing the detonator of the dark squared Bishop, Black will no longer have to worry about pawn structure threats such as Bxh6 ixh6 or "passive defenses" of h6 with ...Bg7. 8. Bxf6 Axf6 9. Ne4 Ah5 10. Neg5 Af6 11. f4 Ng4! 12. Bxf7!? Nf5! White is potentially asking for trouble with 12. Bxf7!?, and S.M.I.R.F. completes its total infiltration with the Knights. 13. e4 Nfe3+ 14. Axe3 Nxe3+ 15. Cxe3 Cg7 White traded the Archbishop for two Knights + 1 pawn and gets a very safe King, extending the combination to recover the Archbishop for its two Knights, gaining another pawn, but losing the Bishop. Overall, Gothic Vortex will be minus a Bishop for two pawns and a solid King position. 16. Ni5 j6 17. Nixh7+ Axh7 18. Nxh7+ Cxh7 19. O-O Cxf7

The dust has settled now, and the difference in material is 10 pawns vs. 8 pawns and 1 Bishop. This position is shown above on the far left. 20. f5 gxf5 21. exf5 c6 22. g4 Qb6 23. Qe2 Rg8 24. Rag1 a4 Both sides are seeking activity with the major pieces now. S.M.I.R.F. is looking to open up the a-file. 25. Rg3 Ra5 26. Qd2 a3 27. b3 And with this move, Gothic Vortex has sealed the a-file off. Re5 28. Cc2 Qa6 29. b4 This is shown above in the middle diagram. 29...Qa4? According to Gothic Vortex, this is the final turning point of the game. White quickly gets a strong endgame with no defects after some further trades. 30. d4! Qxc2 31. Qxc2 Bxf5 32. Qf2 Bxg4 33. Qxf7+ Kxf7 34. dxe5 dxe5 35. Rf1+ Ke6 36. Rfg1 e4 37. Rxg4 Rxg4 38. Rxg4 Ke5 As you can see from the above diagram on the far right, this game is over. Fortunately the S.M.I.R.F. program does not resign and we are able to witness a truly entertaining conclusion.

39. Rg7 e6 40. Rxi7 e3 41. Rg7 Kd4 42. h4 b5 43. cxb5 e2 One might accuse Gothic Vortex of creating some deliberate drama by allowing such a move, but the ending is totally under control.

44. Rg1 cxb5 45. h5 Ke4 46. h6 Ke3 47. h7 Kf2 48. h8=C Kxg1 49. Ch1+ Kg2 50. Ch2+ Kg3 51. Cxe2+ Kg4 52. Cd4+ Kg5 53. Cxb5+ Kg6 54. Cxa3 Kf6 55. b5 Ke5 56. b6 Kd6 57. b7 Kc7 58. Cb5+ Kb8 59. Cb6 j5 60. Cd7+ Ka7 61. b8=C# 1-0

Rarely do we get to see an ending with two Chancellors, and we have an exceptionally rare ending here. Gothic Vortex checkmates S.M.I.R.F. on the same move that pawn is promoted to the second Chancellor.

After 12 rounds, only two programs were still undefeated: Gothic Vortex and TSCP Gothic. Some of the drama of this final pairing had been diluted by the round 5 draw between TSCP Gothic and S.M.I.R.F., leaving TSCP Gothic behind in the score 11.5 points to 12 for Gothic Vortex. In order for TSCP Gothic to win the tournament, it would need to win both games against Gothic Vortex. From what I had witnessed over the course of the tournament, TSCP Gothic had very real chances to do this. From its screen output, TSCP Gothic was averaging about 525,000 nodes per second throughout most of the tournament. Gothic Vortex was closer to 350,000 nodes per second and was occasionally outsearched by one full ply of nominal depth. From the history of computer chess tournaments, this would invariably favor the deeper searching program. The question was, did the Gothic Vortex search extensions beyond the nominal depth in the principal variation play a larger role in determing the outcome? We would soon find out.

Gothic Vortex vs. TSCP Gothic: 1. d4 Nh6 2. Nh3 Nc6 3. d5 Ne5 4. Nc3 d6 5. g3 Bd7 6. Ng5 Nhg4 7. f4 Ng6 8. h3 Nf6 9. Be3 a5 10. Bf3 a4? 11. Af2 This leads to the position on the far left below. The first a-pawn push by Black was the result of White stacking the Bishop and Archbishop along the a7-g1 diagonal, attacking it twice. The second pawn push is not warranted, and the third pawn push, which is coming on the next move, sets Black significantly behind in development.

11...a3 12. Qc1 h6 13. Ni4 axb2 14. Qxb2 b6 15. O-O j5 16. Nh2 A quick look at the middle diagram above and you can see White's development advantage. White has six pieces in play and has castled. Black has only three pieces in play and is nowhere near securing a castled position for his own King.

16...i5? 17. Bd4 i4? Black is just pushing pawns aimlessly, very unlike TSCP Gothic. Such pawn pushes are usually a sign of the "horizon effect". In these scenarios, a "decisive" move leads to a loss in material, so any move that can delay the negative score is said to "push the move beyond the horizon" of the search. In reality, the sequence leading to the bad score is still waiting to happen, it is just that the program is unaware of it. 18. hxi4 jxi4 19. Nxi4 Ri8 It looks like TSCP Gothic is banking heavily on getting some counterplay against White's i2 square. 20. j3 Qc8 And Black piles on another hit towards i2, but White has an answer... 21. Cg2! Ra6 22. g4 Bi7 23. e3 c5 Black continues to move everywhere but the center of the board. 24. dxc6 e.p. Bxc6 25. Ne4 Nxe4 26. Bxe4 Qb7

As I review the game, I get the sense that TSCP Gothic is just "marking time". Black is engaging in simple trades, avoiding complications, and just getting from one move to the next without an accumulation of small positional gains.

27. Bxc6 Qxc6 28. Ch2 Ai6 29. Ad3 Qb7 This is shown in the leftmost diagram above. One of the first things to catch your eye is the disposition of the h-file. Gothic Vortex has stacked a Chancellor and Rook in this semi-open file, and with a Queen and Bishop on the a1-h8 diagonal, White is aiming to undermine the backward g7 pawn that is supporting h6. White's pieces are well placed and the King is sheltered. Black's pieces are in a state of disarray and the King is becoming more vulnerable.

30. f5 Ne5 31. Bxe5 dxe5 32. Rhd1 Kg8 33. Ac5! Just a few moves later, but worth a second look. This leads to the middle diagram above. White has made some noticeable progress since the previous diagram. A kingside defender has been removed from g6 (the Knight), the d-file has been opened and seized by the Rook on d1, Black has been saddled with doubled pawns in the e-file which will soon result in a loss of one of them, and White is gaining attacking momentum. Notice that Black cannot capture the Archbishop after the last bold move of Gothic Vortex. If 33...bxc5? then the Queen is lost with 34. Qxb7. Since the Archbishop is threatening both the Queen and the Rook, TSCP Gothic must retreat its Queen into the corner to defend the Rook. 33...Qa8 34. Ad7 Aj4 35. Ch1 Cd8 36. Axe5 Qxh1+ 37. Kxh1 Ce8 The e-pawn has been won, and Black's Queen was traded for the White Chancellor. 38. Qb3 Ai3+ The deadly Archbishop should never be underestimated. White only has one legal move to get out of the Black Archbishop's near-fatal checking auras. 39. Ki1 Cf8 40. Ag6! White exploits the Queen's pin against the f7 pawn in the extreme! h5 41. gxh5 Ra3! The Rook threatens White's Queen, but this is not just "desperate measures". If the Queen takes the Rook, then the pawn pin is undone and the Archbishop would be captured. 42. h6!! Punch and counterpunch! This is shown in the rightmost diagram above. Now the hanging White Queen cannot be taken, since 42...Rxb3?? 43. h7+!! Cxh7 44. Rd8+ Cf8 45. Rxf8# mates! 42...gxh6 43. Axf8 Rxb3 44. Rd8 Rxe3 45. Ag6+! Kg7 46. Axi8+ And now the final King hunt begins. 46...Kf6 47. Rg8 Ke5 48. Ag7+ Af6 49. Axf6+ Kxf6 50. Ng3 Ke5 In this simplified ending TSCP Gothic is searching 2 plies deeper than Gothic Vortex now, but the end is near. 51. Nh5 Rc3 52. Re1+ Kxf5 53. Ng3+ Rxg3 54. Rxg3 1-0 White had a 12 minute time advantage and two Rooks and four Pawns to Black's Bishop and four pawns.

We took our lunch break, and I was generally upbeat. There was no way for Gothic Vortex to finish in anything but first place, but I would prefer to finish off with at least a draw against TSCP Gothic, and a win would be even nicer. Even though TSCP Gothic did not make a stellar showing in the last game, it is still a very strong program. You can't rate the quality of a win in one game, then make an infinite extrapolation about the superiority of Program A vs. Program B in all other possible games. This was the lesson learned from some of the previous rounds, the most shocking example was Zillions losing to CapaGNU Modified. Also playing to some of my fears, I saw TSCP Gothic reach search speeds of 700,000 nodes per second at the end of the game, and I seemed to recall that was the speed of the Deep Thought chess program at one point in time. Michel Langeveld and TSCP Gothic were obviously going to go far in the world of Gothic Chess, and we had to take them seriously.

TSCP Gothic vs. Gothic Vortex: 1. c4 f5 2. d3 Nh6 3. Nc3 Nc6 4. i3 g6 5. Nh3 Bf6 6. Nd5 Be5 Black is suffering from the "bishop lunge" problem again, unfortunately. White is playing a respectable opening, and I am wondering if a push of the king's pawn will send the Bishop retreating again. 7. Bj3 Ng4 8. Ndf4 Cf6 9. Ng5 Bd4 10. e3 Bc5 I am too distracted by the crowd at this point. The game was being projected onto the big screen monitor at the CompUSA department store, so customers milling about were first just standing in the doorway watching, then filtering in as the other staff allowed it. I remember looking at this position and thinking it looked confusing to me at first glance. See the diagrams below. 11. Nxh7+ Axh7 12. Nxg6+ Axg6 13. Bxg6 Cxg6

We have an interesting theoretical position. Is an Archbishop + two pawns worth more than two Knights and a Bishop? From the 8x8 chess world we know that a Queen is worth 3 minor pieces, so the question is, is an Archbishop + two pawns worth as much as a Queen? We will just have to watch this game unfold and find out.

14. h3 Nh6 15. d4 Bb6 16. j4 d5 17. cxd5 Qxd5 18. Af3 With the exception of 16. j4, White is really pushing Black around. Gothic Vortex has been doing an awful lot of retreaing! See the lower left diagram below.

18...Qd6 19. Ah5! Cg8 20. Af4! Qf6 21. Ad5 Three consecutive Archbishop moves by TSCP Gothic an three consecutive retreats for Gothic Vortex as White pretty much dictates the action. But was the last move by White a strategical error? Did White force once move too many? See the middle diagram above. 21...Qxi3 22. Axg8 Nxg8! 23. Qh5 Be6 24. b3 Nf6! 25. Qi5+ Qxi5 26. jxi5 O-O-O! Black surrenders the Chancellor for an Archbishop + Pawn, capturing to preserve castling priviledges to either side, then gaining a mere tempo with 24...Nf6! which chased the White Queen for one brief moment. But look at the rightmost position above now, after queens have been traded and Black castled kingside. White now has no pieces in play, while Black has all four minor pieces developed and is castled. White's iniative has evaporated and Black is positioned well in this middlegame. The material difference is Chancellor + Pawn vs. Bishop + two Knights. 27. Bb2 j6! 28. ixj6 ixj6!

The stage is being set for some Grandmaster class engame play. Black will push the passed j-pawn down the board, coordinating the efforts of the minor pieces to overwhelm the Chancellor and disrupt White's defensive fallback postitions.

29. Cf3 Rdg8 30. Rj2? Ba5! A curious choice for White, which allows Black to latch onto an excellent transient diagonal for some later manuevering to a better post while at the same time instituting the mild annoyance of disallowing a late kingside castling. 31. Rc1 Bb4 32. Cf4 Nd8 33. Cd3 Bd6 34. Ri2 j5 35. Bc3 Nc6 36. Be1 j4 37. Rj2 Rgi8 Both sides are groping for the thread of the game, needing more search time to resolve the best way to proceed. Each program has about 35 minutes to finish the game from here. See the leftmost diagram below.

White is going to have some serious problems if it can't make better use of the Bishop and connect the Rooks to rally against the j-pawn. 38. Bc3 Bi1! Black strikes at the heart of White's vulnerabilities, paving the way for the final onslaught via an incursion into the second rank. 39. Rj3 Ri2 40. Rj1 j3 41. g3 Bd5 42. Cf4 e6 43. Rd1 Bj2 44. Rc1 Bg5 45. Ce2 Bg2+! 46. Ke1 Bxh3 47. Cg1 Rh2 48. f4 Bi3! There is no need to be hasty removing the Rook. 49. Ci1 This is shown in the middle diagram above. The masses are huddled in the corner, and all play revolves around the pawn promotion. 49...Bxj1 50. Cxj1 Ri2 51. b4 j2 52. b5 Ng4!! Allowing the Knight and Pawn to drop, with check even, to finally get the Chancellor off of the board. 53. bxc6 Nh2 54. cxb7+ See the far right position shown above.

Each program is searching very deeply now, and White is battling hard to keep the game from ending. Should Black miscue and take the pawn, the exposure of the King would become a factor. The safe play is to hide behind this dead pawn, and use it as a traitor to the White camp. 54...Kb8 55. e4 Nxj1 56. Ba5 Bh4!! Even cornered and obviously lost, White is still finding provacative play to extend them game in the event the opponent makes the slightest inaccuracy. Black counters with a move you would never have expected to see in a million years! Such is the "elegance" of brute force. 57. Bxc7+ Kxb7 White supplies Black the correct time to remove the pawn, then the Bishop, as the play winds down. 58. Rb1+ Kxc7 59. Rc1+ Kd7 60. Rc2 Rxc2 61. exf5 Rxa2 62. fxe6+ Ke8 63. gxh4 Nh2 64. Kd1 j1=Q# 0-1

A well-earned victory for Gothic Vortex in a game that was very interesting to observe.

For more Gothic Chess, see Gothic Chess Miniatures and Tactical Artwork!