8man Endgame Tablebases ELO gain?

[hgm]

I dont see how 8 man would be less useful than 7 man?

The point is that the value of all chess pieces happens to be odd. And the sum of an een number of odd terms must be even. So with an even number of men you are talking about material balances worth 2 or 0. The former are usually easily won even without EGT, the latter easily drawn. It are the advantages of 1 where most of the contest is that can swing the game result.

[yurikvelo]

It are the advantages of 1 where most of the contest is that can swing the game result.

such hard 7-men position before conversion always were 8-men.
There is higher chance to get "winnable 1" when probing Syzygy-8.
And higher chance to miss "winnable 1" with higlhly selective pruning + LMR before 8->7 conversion

Engines started probing (tbhits > 0) very early

most likely 100% of this hits either easy win/loss (eval > 3 without Syzygy) or dead draw (0.00) and do not influence ELO.
Such enourmous selective depth happens with extensions (checks or captures), combined with sparse LMR.

Very unlikely that this TBHits happen on position with uncertain eval like +0.7 ... +2, when assigning +150 or 0.00 will help to secure win or hold draw.

[dangi12012]

It are the advantages of 1 where most of the contest is that can swing the game result.

such hard 7-men position before conversion always were 8-men.
There is higher chance to get "winnable 1" when probing Syzygy-8.
And higher chance to miss "winnable 1" with higlhly selective pruning + LMR before 8->7 conversion

Engines started probing (tbhits > 0) very early

most likely 100% of this hits either easy win/loss (eval > 3 without Syzygy) or dead draw (0.00) and do not influence ELO.
Such enourmous selective depth happens with extensions (checks or captures), combined with sparse LMR.

Very unlikely that this TBHits happen on position with uncertain eval like +0.7 ... +2, when assigning +150 or 0.00 will help to secure win or hold draw.

I dont know if this is true. I can see in taskmanager disk activity after a few moves - and i have 6 piece tablebase installed.
Maybe someone can corroborate

[yurikvelo]

I dont know if this is true. I can see in taskmanager disk activity after a few moves - and i have 6 piece tablebase installed.
Maybe someone can corroborate

count is very low, 2k tbhits is nothing from 20-men FEN.
reaching 6-pos with 20 pieces on board is possible only due to search extensions (combined with LMR). Extensions only cover captures and checks, to ensure that after piece trading you don't lose material. This trading can easily reach 6-men, and mostly it is evaluated correctly (close to 0.00 or >3).

Access to Syzygy won't influence your eval.

Set up position on board and make 2 runs with fixed amount of nodes or time ('go nodes 1000000000' for example)
1) with Syzygy, to reach few kilo TBHits
2) without Syzygy (clean engine restart)

depth/eval and bestmove will be the same all the time (make 10 or 20 such different positions)

Some tree leafs which got TBHits will be assigned as +150 or 0.00 rather than +4.2 or +0.4, but this won't influence cp/depth/bestmove