One of the best ways to improve in Horde Chess is to review your games and learn from your mistakes. Lichess offers a free computer analysis for any game to help players perform such analysis. For that, it uses a version of the Stockfish chess engine internally, adapted to analyze Horde Chess.
This analysis, however, is far from perfect. The user Lbluffuout noticed a few times when the computer analysis was wrong and misleading. When getting help from a computer for analysis, it is crucial to notice that blindspots like those exist and explore why the computer says its evaluation.
Exhibit A: Was the bishop sacrificed for nothing?
In the game Rayholt vs. mindhunter0, move 8, Stockfish evaluates the position as +0.1, meaning a balanced position.
mindhunter decided to take advantage of the temporary weakness of the kingside pawns and went with 8…B:f4!. Stockfish drops the evaluation to +2.3, meaning a substantial advantage for white, and recommended recapturing with 9.g:f4. After playing Stockfish’s recommended moves 9…. h:g5 10. h:g5
We can see that black has an opportunity to intrude on white’s pawns over the h file, and white has too little counterplay on the queenside. I played this position against local Stockfish and only after I positioned the queen on h7 and took R:h2 it admitted black stands better.
Exhibit B: A matter of depth
In the following position, from my game against paHorde:
ׁHere, we can see that White is in serious trouble. There is no way to defend against 53… R:e6 54.f:e6 Q:e6, which leads to a clear win since the king can handle the pawns on one side while the queen gets the pawns on the other side. Illustration:
There is no way for White to advance his pawns without exposing them to capture.
However, the default “analyze game” result gives this position scoring of -1.7 only, suggesting everything is still open for White. This might lead to the wrong conclusion that white made some mistakes down the road which led to his defeat. In that case, turning on local analysis corrects the evaluation and within 2-3 seconds Stockfish gives this position a solid -7.7
Exhibit C: ּSacrifice a pawn for… what exactly?
In my game against Tonne1983, we got into this position:
Tonne1983 played 44. b4, which might not be the best move in this position, but Stockfish recommendation was to play 44. h5?! instead. This bold move sacrifices a pawn, but I scratched my head trying to understand what is the compensation. In this case, too, trying local analysis quickly made this odd suggestion disappear.
Proceed with Care
Using Stockfish can reveal the game’s key moments, and suggest moves you didn’t think about before. However, one should not blindly trust its assessments without double-checking its evaluation and plans. Turning on local evaluation and playing several moves along Stockfish’s top suggestion may be particularly helpful.
