Armageddon Opening set

[Nordlandia]

What about this opening.

1. e4 c6. 2. d4 Nf6.

It's approximate similar in eval as 2...Na6.

[Ajedrecista]

Hello again:

I also found several of your above lines (but not all). The reason most were rejected is that they didn't have my required minimum of thirty games in the Hiarcs database so we wouldn't have to rely solely on engine evals. The 1.e4 d5 2.exd5 c6 gambit comes closest, with 23 games, but its eval is at the very low end of the allowed range, it's probably not that close to losing. Several others have just a single digit number of games. The St. George defense has enough games, but it is normally just going to transpose to my Polish defense line 1.d4 b5 after 2.e4 Bb7 with ...a6 soon needed. The only line above that seems to fulfill all requirements is your Dutch defense line. I must admit that I never thought of it, because I assumed it would just transpose to the Leningrad Dutch, but now I know that 2...g6? is an incorrect move order, 2...Nf6 first is necessary. The King's gambit line violates the principle that White must have played no clearly inferior moves; obviously the King's Gambit is objectively a mistake. The Bucker variation gives an eval that is too clearly won (above 1.2) for this purpose.
Obviously, the problem is that very few lines that are on the cusp of losing will occur in practical play thirty times, especially after just two moves where there are not likely to be complex tactics (well, your Dutch example seems an exception!). Probably there are others that neither of us thought of. I suppose that if we ever needed a substantially larger set of openings, I could just raise the two move limit to three moves.

I got similar evals with Dutch Defence, 2. ... g6, but with 2.Nf3 and 2.g3 instead of 2.c4. I do not know if they are eligible or not due to transpositions. If that, you could have The Baker's Dozen; if not, you could try to find one more to have a special dozen, like in checkers (The Golden Dozen, which were formally barred). An Armageddon situation is often unpleasant, hence you could name it The Dirty Dozen, like the great film and opposed to The Golden Dozen.

Regards from Spain.

Ajedrecista.

[lkaufman]

I decided to revise my list a bit in view of the suggestions made here and other factors. Since many plausible lines have just a few games in the database, not enough for meaningful stats, I added the requirement that recent Lc0 (running for a minute on a good GPU with big net) shows a White win probability between 48% and 52%, reasoning that Lc0 better simulates human play in that it doesn't assume 40 plies of perfect play, while dropping the minimum number of required games in the database to just 4. In positions where the best third move for White was pretty clear, I looked at stats only after that move, requiring no more deviation between White wins and 50% of games played (which of course only has meaning with a decent sample size). I also used only SF evals after 40 ply (not more to make comparison fair), and set the range to 0.85 to 1.10. I kept the requirement of no bad moves by White and no harmful moves by Black, except in one special case as noted. The result is that I now have 12 Armageddon openings that are more clearly near the win/draw line both for human and engine play than in the original list. Here is the new list with comments:

1.e4 b6? Owen's defense, has the lowest eval, 0.91, but White won 51% of 577 games, the most popular of the ten.
1.d4 b5? Polish defense, has been played by strong players, but the 1.00 eval and 52% Lc0 win perc put it at the threshold.
1.c4 b5? (an inferior relative of the Benko gambit) gives enough comp to be on the list.
1.e4 Nf6 2.e5 Nb8?. Brooklin Defense. Right on the line.
1.e4 c6 2.d4 Qc7?. Not a "good" move, but has some point. Eval and results right around the win/draw line.
1.e4 c6 2.d4 Na6?. Intending ...Nc7 next, played by some top players. Also right on the line.
1.e4 e5 2.Nf3 Qe7? Gunderam Defense. Intending ...g6 and Bg7. Also right on the line.
1.e4 e5 2.Nf3 Bd6?. Good move in slightly different circumstances, here right on the win/draw line.
1.d4 f5 2.c4 g6?. Black should play ...Nf6 first; here 3.h4! puts Black on the precipice.
1.d3 Nc6 2.d4 Nb8?. Not a real opening, but equivalent to White getting a second free move. Not quite as bad as most on this list, but bad enough.
1.e4 e6 2.d4 Nf6?. Mediterranean Defense. An inferior relative of the Alekhine defense. Probably the worst for Black on the list, but near the line.
1.e4 d5 2.exd5 e5? Bohnke Gambit. Better than 2...e6? as White should take e.p. here, but can meet 2...e6? by 3.Bb5+.