PV-Nodes likely to fail low

[Vizvezdenec]

Do not decrease reduction at pv-nodes which are likely to fail low.

Passed STC
https://tests.stockfishchess.org/tests/view/6023a5fa7f517a561bc49638
LLR: 2.95 (-2.94,2.94) {-0.25,1.25}
Total: 70288 W: 6443 L: 6223 D: 57622
Ptnml(0-2): 239, 5022, 24436, 5174, 273

passed LTC
https://tests.stockfishchess.org/tests/view/6023f2617f517a561bc49661
LLR: 2.94 (-2.94,2.94) {0.25,1.25}
Total: 105656 W: 4048 L: 3748 D: 97860
Ptnml(0-2): 67, 3312, 45761, 3630, 58

Idea of this patch can be described as following - if PvNode was researched in non-pv search and value this search got was much lower than alpha we can assume that this PvNode is likely gonna fail low so we can reduce more at it.
bench 3766422

[FauziAkram]

congrats!!!

[snicolet]

Congrats too, way to go!! :-)

[snicolet]

Merged via 76daa88, thanks :-)

[joergoster]

According to the comment, this should be

&& ttValue < alpha - 200 - 100 * depth

alpha is the lower bound.

100 * depth also looks potentially risky ...

[Vizvezdenec]

Yeah, I've already noticed this.
Probably we should rewrite the comments.
Well, it passed SPRTs so if anything it doesn't seem to be that risky. We have confidence of 99,9% that it's not a regression if anything...
It's not the first time I write something and then fuck up some sign and it works better than what I wanted - what can be done with it? Probably nothing.

[joergoster]

It becomes riskier at very high depths. Probably nothing fishtest can detect.

Even though it passed SPRT, I'm not happy it was committed as is. Sorry.

[Vizvezdenec]

At best what it does is making reductions change by 2 - at higher depths it effect becomes more and more subtle. What it means is that it will have no effect as soon as we increase depth by 2.
Which is of course something, but it's by no means massive.
The biggest effect (as any of this patches) is at lower depths - well, and we can see it thru fishtest. Sure, search is recursive and thus is has effect anywhere, but if anything this is how we always did things.
For example, in null move pruning we have static eval condition having - 28 * depth or smth, at higher depths it can also be pretty huge, but no one I think is against it being as it is, right?
Also I must say that fishtest LTC is like depth 25~ or even more in endgames, and I don't think that having +2500 is that much different to having +5000 at TCEC conditions, not gonna lie.

[ddugovic]

I apologize for conflating matters, but I have a modified engine which without NNUE supports exotic (more than 32 pieces) material combinations; and while that in itself is not supported, the other day I observed an infinite recursion in qsearch. I'm not asking for support but am curious whether there are legal positions which cause a regression.

[Vizvezdenec]

qsearch has nothing to do with this patch, write an issue.