How worthwhile is an aspiration window?

[micron]

Recently I implemented an aspiration window for iterative deepening at the root, but am finding it hard to detect any improvement.
If I turn off PVS, the new aspiration window code gives a clear benefit (around 25 Elo).
With PVS enabled, the new code's effect on Elo is too small to measure easily.

Does this mean that PVS already harvests most of the potential benefits from playing tricks on alpha and beta, so that there is little to be gained from an aspiration window?

At least two well known engines use an aspiration window, whereas others do not.

Robert P.

[zamar]

Aspiration search is closely related to time management and to the shape of the search tree, so possible advantage is very engine dependent.

I've never heard though that aspiration search could hurt (when using big enough aspiration window)

[Daniel Shawul]

Your search should be stable otherwise it could be a loss. For games like reversi and go, where there is no defined queiscent search, scores fluctuate too much from iteration to iteration. And I was forced to drop aspiration search all in all. It just doesn't make sense if there is a big odd-even effect. Luckily for chess that is not the case.

[bob]

Recently I implemented an aspiration window for iterative deepening at the root, but am finding it hard to detect any improvement.
If I turn off PVS, the new aspiration window code gives a clear benefit (around 25 Elo).
With PVS enabled, the new code's effect on Elo is too small to measure easily.

Does this mean that PVS already harvests most of the potential benefits from playing tricks on alpha and beta, so that there is little to be gained from an aspiration window?

At least two well known engines use an aspiration window, whereas others do not.

Robert P.

It is a minor improvement. In some positions it will help significantly. In most, it makes little difference.