I played a few games against the crazyhouse computer. My impression is that while tactically it is very strong it plays weird moves and has bad impressions about material. I played both games with white, because my repertoire against 1.e4 and pretty bad and I wanted to see how the computer would fare in non-tactical crazyhouse positions. (If such thing exists, lol)
en.lichess.org/lLIZVFPB/white#1
In the first game I was clearly winning after move 16, but the computer thinks that 4 minor pieces vs queen is around equal. In addition, blacks king is weak, white is safe and sound pawn structure, and black has trouble getting out his queenside.
Around move 20 engine thought it was winning, which is of course absurd. I think the engine should reduce the value of queens and increase development/king safety in evaluations.
Then on move 25 I played Nxf2. Whoops! This is the kind of machine trickery that they are good at. Instead Rg3 is winning. I had a variation in analysis and it seems that the eval is okay for a while and then shows completely lost for black. Well, what do you know.
After that it seems okay for white but the engine unleashed a flurry of tactics to crush me.
Second game: en.lichess.org/C3dlL4c3/white#0
I played a solid line, and at move 14 the engine suggested @Bc8! which didn't occur to me. What an idea!
Anyway after @Bh6 I felt the position was quite alright. The engine played some nonsensical bishop moves, ensnaring my king. Though I should be winning, I had to be careful for some knight checkmates (e.g. @Nf3, Bxf3, @Ng2#). At the end computer had no clue about what to do after sacrificing his own knight for nothing.
In conclusion, the engine is very strong in open positions (as shown in game #1) and comes up with resources even when lost (@e2!...@f2!) however in closed positions it seems quite weak. I think 2500 crazyhouse is reasonable is short time controls but in longer time controls it is more or less easy to beat.
en.lichess.org/lLIZVFPB/white#1
In the first game I was clearly winning after move 16, but the computer thinks that 4 minor pieces vs queen is around equal. In addition, blacks king is weak, white is safe and sound pawn structure, and black has trouble getting out his queenside.
Around move 20 engine thought it was winning, which is of course absurd. I think the engine should reduce the value of queens and increase development/king safety in evaluations.
Then on move 25 I played Nxf2. Whoops! This is the kind of machine trickery that they are good at. Instead Rg3 is winning. I had a variation in analysis and it seems that the eval is okay for a while and then shows completely lost for black. Well, what do you know.
After that it seems okay for white but the engine unleashed a flurry of tactics to crush me.
Second game: en.lichess.org/C3dlL4c3/white#0
I played a solid line, and at move 14 the engine suggested @Bc8! which didn't occur to me. What an idea!
Anyway after @Bh6 I felt the position was quite alright. The engine played some nonsensical bishop moves, ensnaring my king. Though I should be winning, I had to be careful for some knight checkmates (e.g. @Nf3, Bxf3, @Ng2#). At the end computer had no clue about what to do after sacrificing his own knight for nothing.
In conclusion, the engine is very strong in open positions (as shown in game #1) and comes up with resources even when lost (@e2!...@f2!) however in closed positions it seems quite weak. I think 2500 crazyhouse is reasonable is short time controls but in longer time controls it is more or less easy to beat.