Houdini 2.0c - Mate in 1 bug

[Arpad Rusz]

Using Chessbase 11 in the following position Houdini finds all three mate in one moves (1.Qf8#, 1.Qg8#, 1.Qf6#):
[D]3Q3R/pp3pk1/6b1/4K3/8/8/PPq5/8 w - - 0 1

But try it by removing the Pf7:
[D]3Q3R/pp4k1/6b1/4K3/8/8/PPq5/8 w - - 0 1

You will get no analysis at all...

This is more interesting:
[D]2Q2R2/pp2k1p1/8/4KP2/8/8/PP5r/8 w - - 0 1
Houdini finds the two checkmating moves (1.Qd8#; 1.Qe8#)

If we try to remove one of them by defending d8 or e8 with a black Bishop we get the following:
[D]2Q2R2/pp2k1p1/8/b3KP2/8/8/PP5r/8 w - - 0 1
Houdini finds the 1.Qe8#! Bravo!

But try this:
[D]2Q2R2/pp2k1p1/8/4KP1b/8/8/PP5r/8 w - - 0 1
You will get nothing in the analysis window... It only shows Depth=1/15 but no moves.

Try removing the useless White Pawn from f5:
[D]2Q2R2/pp2k1p1/8/4K2b/8/8/PP5r/8 w - - 0 1
Bingo! Now it shows the 1.Qd8#!

If you defend again the e6 square (e.g. with a Knight):
[D]2Q2R2/pp2k1p1/8/4K2b/3N4/8/PP5r/8 w - - 0 1
Guess what happens?

[zullil]

What do you get if you run the engine from a command line, without any GUI?

[Arpad Rusz]

I am not used with running an engine from command line. What should I type to get the analysis for a FEN? (I guess the FEN alone is not enough.)
Anyway, this is not a Chessbase problem, Houdini 1.5 doesn't have these problems at all.

[Houdini]

Hello Arpad,

This is a Chessbase/Fritz bug. Houdini 2 outputs a single line of analysis showing the mate. The GUI doesn't show the output from Houdini.
This has been discussed before at the Rybka Forum, please start reading from this post downwards.

Robert

[noctiferus]

yes, in Shredder everything is OK

[MM]

I am not used with running an engine from command line. What should I type to get the analysis for a FEN? (I guess the FEN alone is not enough.)
Anyway, this is not a Chessbase problem, Houdini 1.5 doesn't have these problems at all.

Hello, you are absolutely right. Every engine (but Houdini 2.0) included Houdini 1.5a, show all the mate lines, at least for the 1st position without f7 pawn.

So it can't be a chessbase problem.

Regards

[Arpad Rusz]

I know that Houdini actually sees those mates and the problem is that sometimes the communication with the GUI is broken. Chessbase is buggy, it has always been that. But you know that waiting for them to fix this is hopeless. So maybe you should do it yourself.
What I don't understand: Why are the outputs from Houdini different for the position with, or without attacking the e6 square?

[MikeGL]

Hi Arpad,

I think similar question, running UCI engine in command prompt, was already asked here in talkchess, but I can't find it now.

Anyway for Houdini1.5a, I just made a temporary copy in my desktop for testing, not sure if that is possible with Houdini2. Then I go to command prompt and typed:

Houdini1.5a
uci
position fen (copy your fen from talkchess and right-click mouse in command prompt and paste fen without my parenthesis of course)
setoption MultiPV 5
go

(make sure you are in correct path where Houdini 2 is present, it could be c:\Program Files\Houdini2 in your case)

I don't see textfile log or debug option in Houdini1.5 so prepare for a fast scroll in your command prompt as houdini will scroll it fast for its output. Try to press your keyboard button [pause/break] so you can pause the screen and read the output.

[Houdini]

I know that Houdini actually sees those mates and the problem is that sometimes the communication with the GUI is broken. Chessbase is buggy, it has always been that. But you know that waiting for them to fix this is hopeless. So maybe you should do it yourself.
What I don't understand: Why are the outputs from Houdini different for the position with, or without attacking the e6 square?

Yes, I'll probably have to make another work-around for the Chessbase GUIs. Like sending a second line of output with a higher search depth so that Fritz actually shows it...

To test the different positions, just run the engine in a command window by double-clicking on it in the Windows Explorer.
Type or copy/paste the UCI instructions:

Code: Select all

position fen 3Q3R/pp4k1/6b1/4K3/8/8/PPq5/8 w - - 0
go

produces:

Code: Select all

info multipv 1 depth 1 seldepth 16 score mate 1 time 2 nodes 5410 nps 2705000 tbhits 0 hashfull 0 pv d8g8

Robert

[Arpad Rusz]

For me, it's more a funny bug than a serious one.

Thanks! Actually, I have tried something similar with the command line but without the word "FEN". Naturally, the engine started to analyse the initial position...