Greetings, I am experimenting with fixed depth search, which means no iterative deepening, no Q search, and I am not dealing with hash moves, just transpositions.
OK so a hit is defined as a hash key match and remaining depth <= stored depth.
So I am comparing results when I am only using exact matches on the search depth, to those where depth <= stored depth. Is it possible with a properly functioning system that changing the condition from == to <= should result in a small increase in leaf nodes? I am noticing this with a small percentage of the WAC positions I am testing with. No change in PV and eval, just a small increase in nodes. This seems counter intuitive to me.
Unusual Hash Probe results?
Moderators: hgm, Rebel, chrisw
-
- Posts: 814
- Joined: Sat May 09, 2009 4:51 pm
- Location: Toronto
-
- Posts: 128
- Joined: Sat Sep 23, 2006 7:10 pm
- Location: Prague
Re: Unusual Hash Probe results?
Could you provide me some of those positions in question? I would like to run the same test and let's see if I observe this effect.
-
- Posts: 814
- Joined: Sat May 09, 2009 4:51 pm
- Location: Toronto
Re: Unusual Hash Probe results?
Well. I ended up doing some reworking in my engine code for another reason, so I am not in a position to try and replicate the situation. Nor am I 100% certain which position I was testing. I should be able to try again let you know on the weekend.tvrzsky wrote:Could you provide me some of those positions in question? I would like to run the same test and let's see if I observe this effect.
In the meantime, I have been informed by an expert in the field that such a scenario is possible, it does not necessarily mean a problem