Mastermind GM Larry Kaufman will introduce Dragon 3.3 by Komodo Chess.

October 4, 2023

[October 4, 2023]  We have released Dragon 3.3 at komodochess.com, an upgrade of the engine that won the World Computer Chess Championship in July 2022 after a tiebreak match with Lc0. Dragon uses NNUE (Neural Network Updated Efficiently) technology, originally developed for the game of shogi. Komodo has a great deal of chess knowledge in its evaluation. Training an NNUE network based on this evaluation was both an advantage and a challenge, requiring experimentation with architectures and data generation of billions of positions. The reinforcement learning phase for Dragon has already made great strides.

Dragon 3.3 is a huge strength improvement over Dragon 2, roughly 100 elo at blitz using normal openings, much more in MCTS mode or when playing Fischerandom (chess960). It is a significant improvement over Dragon 3, about 20 elo in all modes in blitz with normal openings, including an estimated 8 Elo since Dragon 3.2. The gains over Komodo 14.1, the last pre-dragon release (Nov 2020), are in the 300 to 400 Elo range on one thread, depending on mode and game type, all at CCRL blitz time control (2' + 1"). The improvement over Dragon 3.2 is due to a new larger net and to search enhancements and parameter tuning. The improvement from Komodo to Dragon is due to the embedded neural network providing much more accurate evaluation and also in some sense gaining an extra ply or so of search by seeing some tactics that a normal eval would not see. Our testing also shows that Dragon is especially strong in Fischerandom (960) chess. Dragon also won the 2021 TCEC Swiss and Swiss2 championships (normal chess). We believe that Dragon will play in a more human-like style than standard Komodo since it relies on learning what actually works in games rather than just on pre-assigned values for eval terms.

The Skill levels have been replaced by Elo values, made much more human-like, and are now intended to correspond to human Fide Rapid ratings or chess.com Rapid ratings, so for example an 2000 FIDE rated player should be closely matched with UCI Elo 2000 at 10' +5". In the grandmaster range this was confirmed by test games between Elo = 2600 setting and Grandmaster Alex Lenderman. They should in general be more accurate in Dragon 3.3 than in all earlier versions. The MultiPV play of standard mode is dramatically improved from the initial Dragon release, on top of the actual strength gain. Other features added since Komodo 13 include Armageddon mode, MCTS Optimism, Personalities, Auto-Skill, and Skill Time (to control speed of play when using Elo settings). Standard mode is usually the stronger mode objectively, while MCTS mode is usually better at setting problems for human opponents as it does not assume perfect defense, especially with multiple threads, and is also stronger than standard mode when looking at several lines of play at once (MultiPV).

Dragon 3.3 has scaled down the evals so that an eval of 1.00 will mean that the position is thought to be on the edge between a win and a draw.