Prophet @ ACCA 2006 Closed

Last modified: 11/23/06

Prophet participated in The 2006 First Annual ACCA Americas' Computer Chess Championships on October 28th and 29th, 2006. The tournament was -physically- hosted by High Point University, and -virtually- hosted by the Internet Chess Club (ICC).

This tournament was unique in that it was a hybrid tournament. Authors were encouraged to be onsite, but it wasn't strictly necessary. All games were played over ICC, freeing up the authors and spectators to chat, either online or in-person for those of us that were onsite. All three of us (out of 16). :) The lack of onsite participants was a bit disappointing... it would've been nice to meet some of the people I've been chatting with over the Internet all these years. Bob Hyatt (Crafty fame) was going to be there, but was called for jury duty the week before and couldn't make it. Despite that, the three of us that were there (Charles Roberson/Telepath, Brian Richardson/Tinker and myself) had some great fellowship. It was a really good time. Never before had I been able to hold a conversation with a person IN-PERSON about computer chess, so that was really a highlight for me.

My expectations going into this tournament were, as usual, not so high. In Prophet's last tournament just two months ago (The 2006 First Annual ACCA Presidents Tournament), Prophet 2.0.delta.1 finished 16/18 with only 2/6 points. However, I knew the current version (2.0.epsilon.1) was quite a bit stronger. I had added some pruning, including late move reductions, improved the aspiration window bounds, and added a principal variation search.. My hope was to come out with 2 points out of 5. (Yes, my expectations are a bit low.)

I'm a patzer at chess, so I can't give any in-depth analysis. If someone wants to provide some, feel free. However, I'll give my impressions of each game below.

Round 1 - Prophet vs Amyan [PGN] [LOG]

I really don't know too much about this engine, other than it was stronger than Prophet. Prophet thought things were going ok up until move 40 or so.

The following was the position after 40. Nc5. Prophet's black bishop is bad, its rooks aren't contributing, and it looks like the kingside is vulnerable.



Prophet pondered on 40. ... Bh3 and saw that it was in big trouble.
12& -227 40174 14381879 Kh1 Rxe3 Qd2 Qe7 Be4 Rxe4 fxe4 Rxc5 Qf4 Rc8 exd5 Ne4
12. -227 42859 15286565 Kh1 Rxe3 Qd2 Qe7 Be4 Rxe4 fxe4 Rxc5 Qf4 Rc8 exd5 Ne4
Indeed, it was over. The kingside defences fell and Prophet was mated with 62. ... Qxe2#. I attribute this loss to lousy king safety. Given that Amyan went on to win the tournament, I don't feel so bad. :)

Round 2 - Matheus Chess vs Prophet [PGN] [LOG]

This program finished one spot above Prophet in the ACCA President's Tournament, so I thought I had a shot here. But, my hopes were starting to fade by move 19...



Prophet saw that it was going to lose a pawn:
12. -85 72123 24565358 a6 Qa1 Bf4 Ne3d1 b4 g3 Bd6 Rxa6 Rxa6 Qxa6 h4 gxh4 bxc3 bxc3
Things stayed pretty bleak for a while, getting down to around -1.5 pawns. But then, Matheus blundered with 27. Bxa6?? .



Prophet saw that it was out of the woods, and even winning!
12& 133 56410 20521234 Qb6 g3 Rxa6 Rxa6 Bxe3+ Nxe3 Nxa6 Qa4+ Kd8 Nc4 Qb7 Ke2 Bd3+ Kf2 Bxc4 Qxc4 Things steadily improved for Prophet, UNTIL...



Here, Prophet played :
13. 241 2 166 Ke7
The only problem with that move is that it allowed 53. Qg5+, a threefold rep! Argh!!!! It looked like Prophet threw away a won game due to some kind of rep bug (more on this later).

Round 3 - Prophet vs Gaviota [PGN] [LOG]

Another program I knew very little about. I didn't even know enough about Gaviota to know whether I should beat it or not. Prophet played some kind of post-modernist opening, with bishops on b2 and g2 and no pawns fighting for the center. Well, not initially anyway. Once out of book that changed as Prophet's pawn tables couldn't be denied some center squares.

This game was rather unremarkable. That was good! I'm pretty sure Prophet was just outsearching Gaviota, which was running on a machine Charles provided. I don't remember the exact specs, but I think the hardware was decent. Prophet's assessment of things slowly and steadily increased, until it thought it was at +1 at move 27.



To my patzer eyes, the position looked pretty good. The bishop on g2 was bearing down on the isolated d5 pawn, the queen/rook were supporting the c5 pawn, and it was a pawn up. Ok, so the a1 rook is misplaced, but you can't have everything.

Things kept improving, until move 60. At this point Prophet had it in the bag - it was a queen up (or soon would be).



But then, the gods of computer chess once again frowned upon poor Prophet and forced it to play:
15. 919 14842 5809734 Kf6
NNNOOOOO!!!!! That allowed 60. ... Qf2+ ... A THREEFOLD REP!!! There was much wailing and gnashing of teeth on my part. How unlucky could I be? Two rep bugs in three tournament games?

That night, I traced the problem down. It turns out that Prophet did not evaluate 60. Kf6 as a draw, because it had only seen that position once before. Then, it got a hash hit (notice the one move PV) and returned the score, all the way through to iteration 15. It never finished resolving Kf6 in iteration 16. When time ran out, the search bailed, and it resorted to the last completed search. I decided to change the search to evaluate any position seen even once as a draw.

So, day 1 ends. Prophet is at +1, half way to my goal of two points. Not great, but not too bad. I'm a bit torqued over giving up half a point (twice), but that's life.

Round 4 - Telepath vs Prophet [PGN] [LOG]

Up to this point Telepath was not having a good tournament. Charles was running Telepath on a 64 bit Gentoo Linux box, but failed to TEST it before the tournament. That got him in round 1 vs Amateur. A timing bug caused Telepath to make very quick moves, so it went down fast. To be fair, Amateur is a very strong program and would probably have won anyway, but Telepath didn't even have a chance. Charles stated 64 bit problems were still giving him trouble in rounds 2 and 3. I thought that by this point (the next day) he would probably have all that worked out and would be looking for blood. And, Telepath did beat Prophet in the ACCA tournament two months ago.

Prophet thought it was slightly down coming out of book, all the way up until move 15 when Telepath played 15. Nh4.



At this point, Prophet got "out of the hole" and started evaluating its position as slightly positive. Things stayed pretty stable for a while, with Prophet thinking it had a small positional advantage. By 21. a4, however, Prophet saw its advantage was translating into extra material.



12. 118 54222 19615954 e4 Nf5 Nb6d5 Nd4 a5 Kh2 b6 Ba3 Qe8 Rc6 Rc8 Re1c1 Rxc6 Nxc6

Charles stated Telepath was evaluating the position as +0.7 or so for Telepath, so I was cautiously optimistic.

Telepath played 25. h5, leading to the following position.



Prophet saw this:
11& 252 43250 15525192 Nxh5 Ne3 b6 Nxd5 Rxd5 Be3 Rxc1 Rxc1 Nf6 Qc3 Be2 Qc8 Rd1+ Kh2

So, things were looking pretty good. Or were they? Charles stated Telepath _still_ saw the position as +.0.7(ish) for white. By this point I was believing Prophet, and sure enough it went on to mate Telepath at move 66. Charles concluded he had gone too far with some pruning he added (turned on) that morning. I felt bad that his troubles were continuing, but I was glad to have the win anyway.

At this point I've met my goal of +2, and I was satisfied. I expected to be paired against Romi or Symbolic in round 5, either of which would probably trounce Prophet.

Round 5 - Prophet vs Symbolic [PGN] [LOG]

Alright, so it would be Symbolic. Symbolic had a good first day and I knew it was a solid program; much better than Prophet. I fully expected to lose, but hoped for even half a point. Prophet has beaten Symbolic, but the record favors Symbolic... especially since Steven bought his new hardware.

Prophet stayed in book through nine moves. After its first search at move 10 it played 10. Qxd4, resulting in the following position:



Both Symbolic and Prophet agreed that Prophet had a very slight advantage. I thought at this point that I had been very fortunate in this tournament to not have come out of book even once down on material, especially considering the little effort I've put into the thing. So anyway, Prophet had a fighting chance, and that's all I could hope for.

To my surprise, Prophet maintained its slight advantage, even built on it! After move 27. Rb6 Prophet was getting close to +1.



12. 80 50396 19024096 Rb6 Qd7 Qb2 Qc8 Rb1 Qf5 Qa2 Qc8 e4 Rd8 Qb2 dxe4 Bxe4

Could it be??? Could Prophet be so lucky? I started to wonder if it could win! I was literally on the edge of my seat. My heart was racing... a very exciting point in the tournament for me. By the time we got to 29. Rb1, I thought things were looking pretty good. I was a bit concerned about Symbolic's c and d pawns, but I thought Prophet was ok.



12. 180 54026 20884496 Rb1 Ra8 Rb1b5 Ra8d8 Rxb7 Qxg3+ fxg3 Rxd2 a6 Re8 Rb8 Rd1+ Kf2 Re8d8

Prophet was starting to get low on time. And, I realized that this might get down to just a few men, and Edwards was the man for tablebases before Nalimov's were around. Prophet doesn't have tablebases, so I was becoming concerned it would lose afterall. What a rollercoaster!

Prophet's score briefly went over +2, then hovered around 1.6-1.7 for a long time. That is, before 74. ... Qc2+.



Prophet's score fell back to almost even, the queens came off the board, and a tablebase hit by Symbolic started to look more and more likely. But, quite by luck I'm sure, Prophet managed to get in a won position... it would promote a pawn. Symbolic resigned at move 97.

I felt very fortunate to have that win. It was quite unexpected.

Final thoughts

I had a blast at this event. It was Prophet's best tournament finish ever; it finished with three points, in a four way tie for fourth place. It ended in seventh by buchholz, but not bad at all. Prophet was very lucky, finishing above several stronger programs, but that can happen in short tournaments.

There was a bit of controversy regarding Charles's decision to enter Amyan, but his intentions were good. I agree that he should've gotten Antonio's blessing, but technically he did nothing wrong in my view, and it's not worth getting bent out of shape over. I thought some of the postings on CCC were a bit over the top. I hope we can just call it a lesson learned and move on- life's too short.

To other chess programmers - I hope that next year we have another tournament, that I'm able to attend, and that you will too.

Many, many thanks to Charles for organizing this. It's easy to criticize, but I think he deserves a big pat on the back for a job well done.
Back to the main page