How to evaluate whether a position will result in a draw vs a decisive outcome?

Question

Suppose 2 positions both have 'equal' evaluations according to the engine, but the first position has strong, easy-to-spot moves which lead to a drawn endgame, whereas the second game is extremely volatile with many (e.g. at least hundreds of thousands) of possible continuations for both players. It may reasonably be assumed the first game has a higher likelihood of ending in a draw, as compared to the second.

Does any engine exist that can estimate the likelihood of a position resulting in a win/loss as opposed to a draw?

Answer 1

Here is a link to an engine that tells you Expected CP loss and % chance of blundering, depending on your ELO. I don't know how accurate the engine is, but it is effectively telling you how sharp a position is.

If I understand correctly, sharpness is the volatility you are referring to. If a position is not sharp, then both players' moves will be low-risk and usually straightforward. I imagine in this scenario a draw is more likely. If a position is very sharp, then even though the engine evaluation may be 0, many possible moves can change the evaluation by a few points. I imagine in this scenario a win/loss is more likely.

Hope that helps :)

Answer 2

good answer of odinchess. here's an illustration of the site:

1. This pawnless rook vs rook endgame is completely drawn, not just theoretically but also practically. Even new players should realise it is drawn: k7/r7/8/8/8/4K3/4R3/8 w - - 1 1.

   [FEN "k7/r7/8/8/8/4K3/4R3/8 w - - 1 1"]
   

There are so few ways to make a bad move. Expectedly, the blunder chance is 0% even at an elo of 1000.

2. This basic king vs king and pawn endgame 8/3k4/8/4K3/3P4/8/8/8 b - - 1 1 is completely (theoretically) drawn, but it's hard to keep the draw if you're completely new. In contrast to earlier, there are several bad (well you can't lose, so 'bad' here refers to 'not winning' instead of 'losing') moves here.

   [FEN "8/3k4/8/4K3/3P4/8/8/8 b - - 1 1"]
   

The blunder chance at 1000 elo is expectedly high at 73%.

---

Ways the site doesn't seem to work:

1. Unbelievably, the blunder chance of the pawn endgame above at 2048 elo is 15%. Apparently, Gotham Chess is serious that e was rated 1900 without knowing much endgames.

2. Unbelievably, the blunder chance of the recent Ian Nepomniachtchi vs Magnus Carlsen game where Nepo makes a Bobby Fischer (vs Spassky)-like blunder is actually 0% at an elo of 1625 and 2% at an elo of 1000: 3rb1k1/1Bp2pp1/4p3/4P2p/r1P3nP/1N4P1/P4P2/R3R1K1 w - - 1 27

   [FEN "3rb1k1/1Bp2pp1/4p3/4P2p/r1P3nP/1N4P1/P4P2/R3R1K1 w - - 1 27"]