Deep mate question

[D Sceviour]

Dennis links are very interesting indeed.

I must have something very wrong in Minic because it takes only 3 sec to find the good move, after that window is opened gradualy, but still quite fast, and only when the engine reach fully semi-opened window to search is slowing down and never finish although g2g7 is at that time the TT move at root ...

I don't get it ...

The analysis you posted initially does not show g2g7, so it is not clear what you are referring to when you say 3 sec to find g2-g7.

I am sure you have a good chess engine. However, mate finding is a special interest that can actually weaken a program if used too aggressively. Most game positions are not mating positions, but rather a struggle to pass a pawn, or queen a passed pawn.

However, we all like to see those mating announcements. Try adding checking extensions. If the move gives check then extend the depth by one ply.

In Schooner, I am considering adding a tactical or mate-finding flag to the user initialization. That is, this would give the user the opportunity to decide whether to slow the program down and look for mates. Currently, Schooner adds checking extensions only in analysis mode but in game mode there are no checking extensions. This seems to be the logical use of checking extensions but denies the user a choice. Which method would you like to see?

[xr_a_y]

Well g2g7 is not displayed in the analysis because in Minic I only disply at the end of the deepening loop. But if I print out the pv when bestmove change at root then g2-g7 is shown and window starts to grow. g2g7 then remains the TT move at root but the search never finish.

[xr_a_y]

Ok got it.

Prunings are enabled in minic only if !ismatescore(Bêta)

Without this restriction the mate in Shirov position is found in 8sec but i guess this would lose elo.