Game adjudication in rating lists

Graham wrote:

The GUI will not adjudicate a game as drawn until it is over 60 moves in length, and the last 10 moves for both sides were below the 0.80 threshold.
Of course, this doesn't apply where any 3-fold repetitions or tablebases adjudications happen earlier.

The tablebases adjudication only kicks in once there are 5 pieces left on the board (at present), or when it sees a mate either 50 or less than 50 moves away.

A game is adjudicated as lost by an engine once there are four consecutive moves giving a 4.75 or higher disadvantage for it.

50 move draws are not ignored.

There is no minimum game length set.

The above is taken from Graham's tournament at Talkchess

There is a something to wish for IMO and I would like to know how other rating lists do Game adjudication, maybe we can all profit from your long experience.

It's not about experience, but testing under your HW+TC conditions. Run a test without adjudication and then another with whatever adjudication rules you want to apply (same openings, 1 thread preferably). Compare.

Ozymandias wrote:

It's not about experience, but testing under your HW+TC conditions. Run a test without adjudication and then another with whatever adjudication rules you want to apply (same openings, 1 thread preferably). Compare.

From what I have seen Stefan plays games till mate, no draws by adjudication except from table bases.

Like to know the adjudication rules Dio and Ipman are using

cute chess, draw adjucastion, move number 60, move count 5, score 5 cp. The score applies to both engines, movecount is restarted when there is a pawn move or a capture move, I still have games with a length of more than 600 moves.

My lists has only tablebase adjudication.

If my opinion on this is desired, I can also make a statement about the extent to which "unbalanced opening" supposedly lead to a lower draw rate in the games or not, or lead to more accurate results in the rating lists. It's just my personal experience.

2 likes, so do tell  

At CEGT 40/4 we have been using opening guidelines that contain the most balanced UHO openings for at least a year (not freely available). The draw rate is no higher than with very unbalanced opening handicaps. I assume that in a tournament with engines that have roughly the same skill level, the draw rate will always be very high.
In our tests we always test e.g. 10 matches against opponents who e.g. 250 Elo weaker and 250 stronger than the engine we are currently testing. This is intended to counteract a “pool effect”. It is also important to use engines with a lot of games played so far as “anchor engines” as test opponents during the tests.
We (CEGT) could be wrong, just our opinion..

The openings contain more than 16,000 positions that are played randomly, ideally 500 different openings are played in 1,000 games, all opening specifications have been previously evaluated by Komodo and include games by grandmasters or IM.