The chess engine Gaviota was certainly the surprise of the first stage of TCEC computer chess championship. With its current rating of just a little over 2700, it managed to hold a 3000+ engine and achieve stable results to outclass 2900 ELO opponents
Gaviota managed to qualify to the second stage of TCEC. It will play in group 2a together with Houdini, Stockfish, Hiarcs, Rybka, Critter, Junior, and Spike
Miguel Ballicora, the man behind the Gaviota engine, gave an exclusive interview for Chessdom.com just a day after the successful qualification.
Chessdom: Hello Miguel Ballicora and congratulations for making Gaviota go to the next round! Fantastic performance, we wish you only success in the future. Gaviota is certainly the surprise of the first stage of TCEC. Having a rating of just 2700 it fared well in a field of 8 engines over 3000+ ELO and 8 engines over 2900 ELO. Gaviota’s 4,0/7 was amazing, but was it a surprise for you?
Miguel Ballicora: Absolutely! I expected 2.5 to 3 points, and I was dreaming with a lucky 3.5 points, but 4 points was phenomenal. The rating of the version playing (0.86b3) is higher than 2700 (closer to 2800 according to a recently released CCRL ranking). Still, Gaviota clearly performed better than expected.
Chessdom: – Which of the games you would consider the best for Gaviota from the first stage games?
Miguel Ballicora: The last one against Shredder. It was a draw, and Gaviota missed a win at the end, but it attacked the multiple World Champion in both flanks, sustained a seemingly poor pawn structure with dynamic play, sacrificed the exchange for positional reasons, and dominated the board. All this happened while Shredder was thinking was better. Unfortunately, Gaviota missed a win with 50 … Kh7. Either Kf7 or Qd4 win, and it fact Gaviota had Qd4 in the principal variation but changed its mind suddenly (analysis of this position is here http://www.talkchess.com/forum/viewtopic.php?t=47263). Another good game was the victory against The Baron. Gaviota played really well the opening and took over the initiative really quickly.
Chessdom: – Did those seven games help you find things to improve in Gaviota? Can you mention any concrete example?
Miguel Ballicora: Yes, there are many instances in which I felt there was room for improvement. When you closely follow games of this nature, many problems become evident. In the first game against Equinox, Gaviota decided to sacrifice a pawn to equalize with dynamic play. This seems to be ok, but Gaviota did not realize there was a simpler way. It needs more knowledge to understand the drawishness of certain endgame positions. Against Scorpio, it had some draw chances but thanks to wrong scores it misevaluated the opponents plans in the endgame.
Chessdom: – Among the other engines, can you point out some that has also surprised you as a programmer?
Miguel Ballicora: I was surprised by Komodo’s style. I have not had a chance to look at it closely before, but here I observed a couple of games in which it showed an impressive positional understanding (against Nemo and Spike). I was actually impressed by Spike’s performance against Gaviota. It was a really good treatment of a Volga-like gambit position.
Chessdom: – How do you evaluate the chances of Gaviota in the second stage? Who do you consider favorite in TCEC?
Miguel Ballicora: Gaviota should be just happy to be there! I will be really glad if it can score few draws and it gives good fights. The second stage is brutally strong, so Gaviota’s chances are nil.
The favorite is clearly Houdini, but I expect others to be ready to surprise it. However, in the superfinal, with so many games, it will be hard to beat. There are more chances to be knocked out in stage 3, when fewer games are played.