Idea of reduction

[cdani]

Hi!
When you search a position in hash table, and you have an exact score, maybe it's possible to reduce the depth in some cases, for example if you have a previous better move that meets some minimums.
Someone had tried this?
Thanks.

[ZirconiumX]

Hi!
When you search a position in hash table, and you have an exact score, maybe it's possible to reduce the depth in some cases, for example if you have a previous better move that meets some minimums.
Someone had tried this?
Thanks.

Just return the score found if it's deeper than your current entry. If it isn't then use it as the first move to search.

No fancy reductions needed here.

Matthew:out

[cdani]

Just return the score found if it's deeper than your current entry. If it isn't then use it as the first move to search.

No fancy reductions needed here.

Matthew:out

It's the standard and what I'm doing now. But may be someone tried to also reduce depending on the score.

[bob]

Just return the score found if it's deeper than your current entry. If it isn't then use it as the first move to search.

No fancy reductions needed here.

Matthew:out

It's the standard and what I'm doing now. But may be someone tried to also reduce depending on the score.

Why would you reduce? You are not going to do any searching at this node, so there is nothing to reduce.

[cdani]

Why would you reduce? You are not going to do any searching at this node, so there is nothing to reduce.

The idea is when you have an score but it's not of enough depth to return immediately, so you are going to search, but for example compare this score with of the best move of the previous alpha_beta call (the one that called the actual alpha_beta), you see that its worst.

* iterate 1st move.
* new alpha of 20.
* iterate next move.
* found this position on hash, with exact value of -20, so reduce.

Or something like this.

[bob]

Why would you reduce? You are not going to do any searching at this node, so there is nothing to reduce.

The idea is when you have an score but it's not of enough depth to return immediately, so you are going to search, but for example compare this score with of the best move of the previous alpha_beta call (the one that called the actual alpha_beta), you see that its worst.

* iterate 1st move.
* new alpha of 20.
* iterate next move.
* found this position on hash, with exact value of -20, so reduce.

Or something like this.

Finding exact scores is so rare, I doubt this would ever trigger. You can certainly try it, but there are likely better indicators to trigger a reduction than that score...

[cdani]

I understand. Thanks!