Internal Iterative Deepening makes the search more robust against move changes, by never starting a deep, completely uninformed search. I don't know if you already do that.


Well, I'm not implementing IID, but I feel I should take a look at it (including the variation you suggested) :)
My next ...

The suspect thing is that 1... b8c6 seems to be in the TT with score -20 and depth 9, as shown when you do the search without clearing TT after 1. e2e4. Until d=9 you live from this single hash hit, and the search only uses a single node, the root. (You reset the node count between iterations?) I ...

It is a bit hard to see what exactly goes on, because you don't print complete PVs but only the first move.

The results you show strike me as strange. Normally when you have searched e2e4 to 10 ply, the replies to it should have been searched to 9 ply, and stored like that in the TT. It seems from ...

I'm playing with TT and after some frustrating moments I've reached few reasonable results.
Unfortunately not everything works as I expected. :roll: :D

One of my tests to check the correctness of TT is the following:
1) Search at fixed depth (above the search output)

DEPTH SCORE TIME NODES MOVE ...

hyatt wrote:Maybe you mean you "see the light through the trees?"

yeah, it fits perfectly!

The point is that beta is always strictly larger than alpha. So val van only be larger than beta if it is also larger than alpha. So there is no need to make the >= beta test if val is not > alpha.

Thank you both.

I'm trying different algoritms in a more simplistic zero-sum game and I can ...

After studying basics chess engines (mainly TSCP and VICE) I'm trying to write my own chess engine for pure fun. :cry:
My doubt today (actually not just one :roll: ) is about Alphabeta.

From Bruce Moreland's archives (and almost everywhere in the net) I've found the following code:

int AlphaBeta ...