Extremely fair opening set "EFOS"

[User923005]

Attached file the file "extremely fair opening set" which is a set of EPD positions which should create extremely fair results in contests.
Both a deep computer evaluation and actual outcome in games played show that the moves are fairly chosen.
Some effort was also spent to ensure that the positions are not dominated by draws.
All of the positions are frequently played in actual games.
All of the positions have several possible choices for the move to make that are reasonable.

Using the set for a contest requires a GUI that can use a set of EPD positions as starting positions for contests, such as Arena.

[User923005]

The result set was created via this query criteria:
SELECT Epd
FROM Epd e where ( Opening IS NOT NULL ) AND ( ce between -15 and 25 ) AND (NOT (ce = 0 AND pv like '%+%')) AND abs(round(coef * 444.0,0)) <= 30 and games >= 32 AND len(pm) >= 28 and ((white_wins + black_wins) > draws*1.99) order by Opening, acd, acs, Epd

and where coef is computed as:
UPDATE Epd SET coef = (white_wins - black_wins)* 1.0/(1e-20+white_wins+black_wins+draws)
UPDATE Epd SET coef = -coef WHERE Epd like '% b %'

[lucasart]

Interesting stuff!

How deep are these lines ?
Would it be possible to include the move number at the end of the FEN ?

[User923005]

I do not have the move number. I assume that the chess engine is tracking half move clock. If there is a draw coming for that reason, I hope that the engine can detect it. Otherwise I would have to analyze the same position over and over and I do not want to do that.
The lines average about 35 plies.
The effort spent analyzing each line will be a function of the number of times it has been played.
Even lines with only 27 games are analyzed for five minutes (a bare minimum).
Some lines require a novelty in order to get the equal score, so a few database systems with relatively solid analysis may not see them as nearly equal, and I do not specify the correct move to make even though I know it.
Some of the positions will have been analyzed for many hours or even a few days, if they are very frequently played.
I aim for one second per time it has been played (but positions with millions of games do not get millions of seconds) and I have a minimum of 5 minutes.
If, after 5 minutes, on a high end machine I have not reached 30 plies, I will reanalyze again to a greater depth.

Here is an interesting example position:
r2q1bnr/pp1k2p1/2n1p1p1/3pP2p/3N2PP/8/PPP2P2/R1BQKB1R w KQ -

see how long it takes you to get to 30 plies.

This position is also interesting because Nxc6 has good outcomes in real games, but some top engines refuse to play it, preferring instead Be3.

[lucasart]

OK, thank you. 35 plies is way too deep for me. I'm more interested in varied and very shallow openings (let's say 16 plies = 8 moves max).

[velmarin]

Hey, sorry if I confused.
Maybe the wrong file
Most positions have only one movement, two, three, ect.
second epd, just 1.Cc3.
rnbqkbnr/pppppppp/8/8/8/2N5/PPPPPPPP/R1BQKBNR b KQkq - 0 1
example, position 12:
rnbqkbnr/ppp1pppp/8/8/4p3/2N5/PPPP1PPP/R1BQKBNR w KQkq - 0 1

[User923005]

35 plies is the depth of analysis, not the depth of the positions.
The depth of the positions is varied from position to position.