Dink Heckler wrote on 09/10/23 at 20:28:44:

In championship play the programmer usually is also the operator. The programmer who does not code in draw offers and acceptances always misses dinner.

Shredder offers me draws all the time. Unless there are no pawns left I decline, and usually go on to lose. Maybe this excessive draw tendency is the opposite of contempt: See, I am a friendly machine! Do you want to play again?

I cannot suggest any tuning tips; I'm not a programmer. But I could suggest that there may be something in any good engine's programming that might lead to this. The engine has no "soul," no "feelings". It will always defer to the strongest line. But, unlike you who perhaps has an emotional investment in winning, for the engine a draw is just one other choice in a heirarchy of choices. For example, "Always win. If no win, draw. If losing, continue anyway and see if an opponent's blunder revitalizes win/draw chances." [IMHO, the real reason they never offer to resign. I read somewhere, I don't remember where exactly, that Fischer had a game that came down to literally just 2 kings on the board but he still insisted on having his one more move.]

Not sure this is the right palce to post this query.
But since this happened with a Alekhine defence game I decided to post it here.
Essentially, I have been doing some engine matches with diferet engine/programs and noticed that in several cases the engines agree on a draw when the position is starting to warm up. I understabd that from a human point of view "warming up" might be drawish. But the game that I copied below, looked to be very interesting, since Leela played 18...fxe6. I was looking at the game and saw the different search lines each engine was "suggesting". Also strange that the draw was agree when S16 has 0.00 evaluation but leela has -1.something.
Also, i had a similar game where leela plyed the "safest-looking 18...Rg8" and the game continued until move 35 with evaluations of - but close to zero (although in my opinion the draw was also agrre to early).
This also in constrast with games that end with mate when half the game is played with evaluations of +5 for both engines.

Any suggestins on how to tune the engine?

Thank you,
[Event "DESKTOP-SUDTRNN, 30min/40+15min/20+(15m"]
[Site "DESKTOP-SUDTRNN"]
[Date "2023.09.09"]
[Round "1"]
[White "Stockfish 16"]
[Black "Lc0 v0.30.0"]
[Result "1/2-1/2"]
[ECO "B04"]
[Annotator "-0.11;-0.39"]
[PlyCount "46"]
[TimeControl "40/1800:20/900:900+3"]

{Intel(R) Core(TM) i7-9750H CPU @ 2.60GHz 2592 MHz W=48.0 plies; 3 574kN/s
B=25.7 plies; 22kN/s} 1. e4 Nf6 2. e5 Nd5 3. d4 d6 4. Nf3 g6 5. Bc4 Nb6 6. Bb3
Bg7 7. Ng5 e6 8. Qf3 Qe7 9. Ne4 dxe5 10. Bg5 Qb4+ 11. c3 Qa5 12. Nf6+ Kf8 13.
d5 e4 14. Qg3 {Both last book move} Na6 {-0.39/26 169} 15. dxe6 {-0.11/42 86}
Bxe6 {-0.39/30 36} 16. Bxe6 {-0.05/40 40} h6 {-0.40/21 0} 17. Bh4 {-0.05/42 49}
g5 {-0.38/19 0} 18. Nh5 {-0.06/42 58} fxe6 {-0.47/20 0 (Rg8)} 19. f4 {0.00/48
39} exf3 {-0.45/19 0} 20. O-O {0.00/55 44} Ke7 {-0.44/24 203} 21. Nxg7 {
0.00/49 42 (Nd2)} Qc5+ {-1.14/29 151} 22. Rf2 {0.00/52 46} gxh4 {-1.14/31 42}
23. Qg6 {0.00/57 99} Qe3 {-1.10/31 55 (Qe5) Draw accepted} 1/2-1/2