Always full principal variation

[cdani]

Seeing a long time control game between engines or using one to analyze a position, it happens often that the variation shown is only one move or pretty short.

So I have added a new UCI option in Andscacs to show full pv that tries to recover it from hash, most times with success.

I have put it the name of "AlwaysFullPv". Maybe someone thinks on a better name or some engine has a different name for this.

It works when it's thinking without time (analysis mode, fixed depth, ...) or it has more than 20 seconds to do the move and current depth is > 15.

It's not available yet, it will be when I publish a new version in some months.

Anyway I comment it to suggest other engine developers to add such idea to their engines, as I think most people finds annoying to see such short pvs.

[elcabesa]

I decided some time ago to retrieve the PV during the search and I'm now very happy with it.
what I'm not totally happy with is that my search still doesn't follow the PV searched at the previous step but retrieve it from the HASH. I have noticed that my engine, when doing iterative deepening, doesn't always start searching the last PV. this happens when the PV contain repetitions.

I'll try to fix it.

[velmarin]

Ivanhoe has in your code
the normal mode and mode analysis in addition to Monte Carlo.
You might want to take a look.

Code: Select all

Option: AlwaysAnalyze
This option allows GUI compatibility for some and propels the companion of MultiPV mode when playing a game.

Option: TryPVInAnalysis
This option demands an expansion of the PV in analysis mode. If this it not on the PV can be truncated from hash hits.

Option: FixedAgeAnalysis
This option delimits the AGE count in the hash table when in analysis. This is useful when applying forward and backward analysis as in the contrary, the AGE is incremented upon every position