FEN depth toughness • page 1/1 • Game Analysis • lichess.org

Evaluate this FEN code and see how deep you can go.
It took very long time to reach depth 24.

r2r2k1/2q1bp2/p1b1p2Q/3pP2P/1p2P3/3B4/PPP1N1P1/1K1R3R b - - 0 21

https://lichess.org/analysis/standard/r2r2k1/2q1bp2/p1b1p2Q/3pP2P/1p2P3/3B4/PPP1N1P1/1K1R3R_b_-_-_0_21

Why does it get stuck at Depth 14 and Depth 22?

Evaluate this FEN code and see how deep you can go. It took very long time to reach depth 24. r2r2k1/2q1bp2/p1b1p2Q/3pP2P/1p2P3/3B4/PPP1N1P1/1K1R3R b - - 0 21 https://lichess.org/analysis/standard/r2r2k1/2q1bp2/p1b1p2Q/3pP2P/1p2P3/3B4/PPP1N1P1/1K1R3R_b_-_-_0_21 Why does it get stuck at Depth 14 and Depth 22?

tpr

What is the point? Black is losing. Calculating deeper does not change that.

Toscani

The point is to know more, to understand more, to share more.

According to AI, this fen is causing a branching explosion. The heavy threats are forcing the engine to calculate deep check sequences. It said without sufficient hash memory, the engine was re-evaluating the exact same sub-trees millions of times, creating a massive bottleneck. Well it was not true, because I tested again this morning, with stockfish 19 (default settings), directly in the terminal, for 5 seconds.

position fen r2r2k1/2q1bp2/p1b1p2Q/3pP2P/1p2P3/3B4/PPP1N1P1/1K1R3R b - - 0 21
go movetime 5000
...
info depth 19 seldepth 45 multipv 1 score cp -312 nodes 2040784 nps 408156 hashfull 691 tbhits 0 time 5000 pv d5e4 h1h3 c7e5 h3g3 e5g3 e2g3 e4d3 c2d3 c6a4 b2b3 a4b5 d3d4 d8d5 h6f4 a8f8 h5h6 e7h4
bestmove d5e4 ponder h1h3

In the terminal in did not jam at a depth. So it must be in the code I used.
I increased the time to see what depth it could reach in 30 seconds (30,000 milliseconds):

go movetime 30000
...
info depth 31 seldepth 60 multipv 1 score cp -392 nodes 16571770 nps 552392 hashfull 991 tbhits 0 time 30000 pv d5e4 h1h3 c7e5 h3g3 e5g3 e2g3 e4d3 c2d3 c6b5 d3d4 d8d5 h6f4 a8f8 g3e4 b5e2 f4g3 g8h7 d1c1 d5f5 g3e3 e2g4 c1c7 e7d8 c7b7 f5h5 a2a4 h5d5 b7b4 h7g7
bestmove d5e4 ponder h1h3

What is the point of engine self-play with depth settings, if they are going to jam on a FEN code and never reach the depth requested. I've learned my lesson, and now I use time per move. Each move will have a different depth and nodes ...

I should of noted the number of nodes. It would of explained the hang effect ... if it was actually analyzing more or just looping like the AI was trying to say.

My project code is trying to find alternative best moves, using 3 settings, depth, time, and nodes.
What I forgot in the code was to use a timed fall-back.

The point is to know more, to understand more, to share more. According to AI, this fen is causing a branching explosion. The heavy threats are forcing the engine to calculate deep check sequences. It said without sufficient hash memory, the engine was re-evaluating the exact same sub-trees millions of times, creating a massive bottleneck. Well it was not true, because I tested again this morning, with stockfish 19 (default settings), directly in the terminal, for 5 seconds. position fen r2r2k1/2q1bp2/p1b1p2Q/3pP2P/1p2P3/3B4/PPP1N1P1/1K1R3R b - - 0 21 go movetime 5000 ... info depth 19 seldepth 45 multipv 1 score cp -312 nodes 2040784 nps 408156 hashfull 691 tbhits 0 time 5000 pv d5e4 h1h3 c7e5 h3g3 e5g3 e2g3 e4d3 c2d3 c6a4 b2b3 a4b5 d3d4 d8d5 h6f4 a8f8 h5h6 e7h4 bestmove d5e4 ponder h1h3 In the terminal in did not jam at a depth. So it must be in the code I used. I increased the time to see what depth it could reach in 30 seconds (30,000 milliseconds): go movetime 30000 ... info depth 31 seldepth 60 multipv 1 score cp -392 nodes 16571770 nps 552392 hashfull 991 tbhits 0 time 30000 pv d5e4 h1h3 c7e5 h3g3 e5g3 e2g3 e4d3 c2d3 c6b5 d3d4 d8d5 h6f4 a8f8 g3e4 b5e2 f4g3 g8h7 d1c1 d5f5 g3e3 e2g4 c1c7 e7d8 c7b7 f5h5 a2a4 h5d5 b7b4 h7g7 bestmove d5e4 ponder h1h3 What is the point of engine self-play with depth settings, if they are going to jam on a FEN code and never reach the depth requested. I've learned my lesson, and now I use time per move. Each move will have a different depth and nodes ... I should of noted the number of nodes. It would of explained the hang effect ... if it was actually analyzing more or just looping like the AI was trying to say. My project code is trying to find alternative best moves, using 3 settings, depth, time, and nodes. What I forgot in the code was to use a timed fall-back.

Toscani

edited

Imagine discovering a correlation between chess ratings and the gap between depth and seldepth.

I'll have to use fixed nodes for this project.

A massive gap between a nominal depth for example of 29 and a selective depth (seldepth 84) would be a real indicator of extreme tactical complexity. Well I've go a new project idea for that. I'll be measure some played games at different ratings. Imagine discovering a correlation between chess ratings and the gap between depth and seldepth. I'll have to use fixed nodes for this project.

plopsky

Then I tuned down the parameters a bit (number of lines, search depth) to see how it behaved later but it went fairly smooth till depth 26 or so, so again nothing special about this position at depth 22.

I guess it is just a typical position with a lot of possibilities, and thus your compute requirements grow fairly predictably, but due to the exact configuration you use as well as the hardware you may suddenly hit a resource constraint slowing things down a lot?!

I am not sure if this helps, but I just tried it out on a typical laptop and with default settings stockfish in lichess did not show any significant slowdown at specifically move 14 or 22. In fact it very quickly went to move 15, and then it auto stopped but I could of course click the + to go deeper. Then I tuned down the parameters a bit (number of lines, search depth) to see how it behaved later but it went fairly smooth till depth 26 or so, so again nothing special about this position at depth 22. I guess it is just a typical position with a lot of possibilities, and thus your compute requirements grow fairly predictably, but due to the exact configuration you use as well as the hardware you may suddenly hit a resource constraint slowing things down a lot?!