Alpha Zero is one of the strongest chess playing programs ever created. It is a revolutionary new way of using machine learning and artificial intelligence to create a program that learns from its mistakes. It mastered the game of chess in just 4-hours. It did this by playing itself over and over again until it reached a chess rating of over 3500 ELO.
K-Chess is a chess playing program written in the late 1990’s. I came across it during an Internet search years ago and have been using it as my silicon chess sparring partner ever since. Unlike professional chess programs like Stockfish, Houdini or Komodo, K-Chess is a throwback to the weaker generation of chess-playing programs. One the one hand, It plays very strong tactically. On the other hand, it overlooks simple moves in way that closely resembles human play. Its endgame play is horrible too. With all those things considered, my guess is that K-Chess’ ELO is 2100-2200 USCF — sometimes higher, sometimes lower. It really depends on the position.
Inspired by Alpha Zero
In early 2018, I watched Chess.com’s analysis of Alpha Zero’s dominance over Stockfish, one of the world’s top chess programs. I was astounded how Alpha Zero sacrificed so many of its pawns to get an overwhelming positional advantage that turned into a crushing victory. This was a good lesson for me and all amateur players. We put way too much emphasis on guarding our pawns. I tend to protect every pawn that is threatened. A part of me can’t stand material loss no matter how small. That is a wrong instinct though. Sometimes we need to let a pawn go, particularly when it helps to activate our pieces or gain a development advantage. The following game is an example of that.
Takeaways from this game:
Fracturing your opponents pawn structure. Breaking up pawns opens up lines of attack to the enemy king. This isn’t always immediate though. Just like a house resting on a weak foundation, once the foundation starts to crumble, the house will soon fall. Notice I didn’t capture the loose pawns. Rather, I used them as an opportunity to develop pieces and push through an attack. At one point, K-Chess had triples pawns on the e-file. Alpha Zero did this to Stockfish in many of its games.
Occupying weak squares. You’ll notice how my knights infiltrated the light squares early on in the game. With my knights occupying e3 and f4, there was little Black could do to help himself.
Focusing on the uncastled king. There are some games where leaving the King on its starting square is perfectly reasonable. Not in this game though. The King was in constant danger from moves like Bh5. Castling queenside was not legal given my queen was controlling the d-file.
Seizing the initiative. Not capturing hanging pawns almost always leads to a lead in development. K-Chess was in trouble early when it did not castle or develop its pieces. The biggest development issue for K-Chess was its rooks, which never left their starting squares.
Allowing the bishop pair. Amateurs are always taught that keeping the bishop pair is important to maintain an advantage. But, like all motifs in chess, it really depends on the position. In this game, Black’s dark squared bishop was so bad and never saw the light of day. Under these circumstances, allowing Black to have the bishop pair was perfectly fine.
Applying constant pressure. Most of my moves in this game were purposeful. There were a handful of passive moves but not too many. Each move either improved a pieces’ scope or threatened to win material. Alpha Zero did this to Stockfish in amazing fashion. Watch the analysis, here.
Sacrificing the exchange. Rf7 was a killer blow. it took away Black’s only good piece and left him with no way to defend against the incoming attack.
Forcing his pieces to bad squares. Looking at the final position shows the level of control I had from start to finish. Pushing the knight to c8, keeping the bishop on f8 and never allowing the rook to move are all examples of this.