future of top engines:how much more elo?

[MikeB]

If a random moving engine played an infinite number of games against SF, some of the games would be wins for the random engine.
Obviously, a very tiny percentage.

But, just by chance, a win could happen early in the chain.

A random engine will not be zero Elo.
An engine that chooses the worst possible move after a careful search might be able to hit zero.

You mean it is not going to have the lowest elo.

I think the elo of random players should be less than 0 in the CCRL list if they accept the right engines.
unfortunately they test a lot of buggy engines.

I guess that many engines can beat the random player 500000-0 with one ply search(assuming one ply search is enough for them not to miss mate in the next move because they extend checks in the qsearch and also assuming one ply search is enough for them to avoid making a move that cause a repetition when they have the advantage or to avoid making stalemate because stalemate is evaluated by them as 0.00).

of course the random player may be lucky to get a position that it has mate in 2 but it is not going to happen often and even when it happens the probability of it to play the right move is small enough that I guess that the combination of these events is going to happen less than once in million games.

If one looks at the random engine games, one will see a random engine will get draws by buggy implementation of the draw rules. Most Elo systems in used by National or International Federations do incorporate a floor in their rating systems that no one has a negative rating. Also to Dann's point, a random engine will have a higher rating that an engine desired to play the worse possible moves such engine will be significantly lower that it would be most certainly be negative without a floor. All engines always have bugs, so to exclude any engine that has bugs would exclude all engines. Good luck with that suggestion. ;>)

[Uri Blass]

My own games against the random mover.
It seems that I can usually beat it even with one pawn.

I suspect that chess engines that do not play the opponent may perform worse than me.

my score against it is 4.5-0.5 so far when I play with one pawn against all the pieces and I gave different handicap in every game(starting with a different pawn).
Only in one game it captured my pawn when in 4 games my pawn promoted to queen and I won.

[pgn][Event "Computer chess game"]
[Site "URIBLASS-THINK"]
[Date "2019.07.23"]
[Round "?"]
[White "UriBlass"]
[Black "BrutusRND"]
[Result "1/2-1/2"]
[Time "05:51:51"]
[WhiteElo "2400"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/4P3/4K3 w kq - 0 1"]
[Termination "normal"]
[PlyCount "107"]
[WhiteType "human"]
[BlackType "program"]

1. e4 d5 {(a7a6) +40.72/1 0} 2. exd5 a6 {(d8d5) +39.89/1} 3. d6 Nd7 {(d8d6)
+39.94/1} 4. dxe7 h6 {(d8e7) +41.75/1} 5. Kd2 Rh7 {(d8e7) +40.05/1} 6. Kc3
Ngf6 {(d8e7) +40.31/1} 7. Kc2 Nh5 {(d8e7) +39.76/1} 8. Kc3 c6 {(d8e7)
+40.32/1} 9. Kc2 Qxe7 {(d8e7) +39.46/1} 10. Kc3 Ra7 {(b7b6) +40.55/1} 11.
Kc4 f5 {(b7b6) +41.17/1} 12. Kc3 Qe3+ {(b7b6) +40.69/1} 13. Kc4 Qe7 {(b7b6)
+42.20/1} 14. Kc3 c5 15. Kc4 Qh4+ {(b7b6) +41.22/1} 16. Kd3 Be7 {(b7b6)
+39.52/1} 17. Ke3 Ne5 {(b7b6) +40.57/1} 18. Kd2 Qe1+ {(b7b6) +41.37/1} 19.
Kxe1 f4 {(e5d3) +32.34/1 0} 20. Kf2 Nc4 {(e5g4) +32.06/1} 21. Ke2 b5
{(h5g3) +32.28/1} 22. Kd3 Bg5 {(c4e5) +32.06/1 0} 23. Ke4 Ne5 {(h5f6)
+32.14/1} 24. Kxe5 a5 {(g7g6) +27.17/1} 25. Kd5 Kd8 {(c5c4) +27.16/1} 26.
Kxc5 Rd7 {(b5b4) +26.02/1} 27. Kxb5 Bh4 {(g7g6) +26.66/1} 28. Kxa5 Re7
{(g7g6) +27.29/1} 29. Kb5 Rd7 {(g7g6) +26.34/1} 30. Kc4 Ke8 {(g7g6)
+25.95/1} 31. Kc5 Bg5 {(g7g6) +26.09/1} 32. Kc6 Bh4 {(g7g6) +26.40/1} 33.
Kc5 Rd5+ 34. Kxd5 Bf6 {(g7g6) +18.76/1} 35. Ke4 Be6 {(g7g6) +18.93/1} 36.
Kf3 Bd5+ {(g7g6) +19.00/1} 37. Ke2 Ba2 {(g7g6) +19.29/1} 38. Kf3 Kd7
{(g7g6) +18.77/1} 39. Ke4 f3 {(g7g6) +18.94/1} 40. Kxf3 Bd5+ {(a2d5)
+18.86/1 0} 41. Kg4 Bb2 {(g7g6) +17.82/1} 42. Kxh5 Ke8 {(d5f7) +14.88/1 0}
43. Kg6 Bb7 {(d5e4) +15.18/1 0} 44. Kxh7 Ba8 {(h6h5) +10.10/1} 45. Kg6 Ba1
{(a8e4) +9.85/1 0} 46. Kf5 Bb7 {(a8d5) +8.75/1 0} 47. Kg4 Be4 {(b7c8)
+9.89/1 0} 48. Kf4 Bg6 {(e4d5) +8.81/1 0} 49. Kg4 Bh5+ {(h6h5) +10.04/1}
50. Kxh5 Kd7 {(a1e5) +5.34/1 0} 51. Kg6 Ke6 {(a1e5) +5.32/1 0} 52. Kh5 g5
{(a1e5) +5.34/1 0} 53. Kxh6 Be5 {(g5g4) +6.20/1} 54. Kxg5 {Insufficient
material} 1/2-1/2
[/pgn]

[pgn][Event "Computer chess game"]
[Site "URIBLASS-THINK"]
[Date "2019.07.23"]
[Round "?"]
[White "UriBlass"]
[Black "BrutusRND"]
[Result "1-0"]
[Time "05:55:17"]
[WhiteElo "2400"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/P7/4K3 w kq - 0 1"]
[Termination "normal"]
[PlyCount "75"]
[WhiteType "human"]
[BlackType "program"]

1. a4 g6 {(b7b6) +39.76/1 0} 2. a5 h6 {(a7a6) +39.80/1} 3. a6 d6 {(b7b6)
+40.33/1 0} 4. axb7 Nf6 {(c8b7) +37.84/1} 5. bxa8=Q Nd5 {(c8a6) +23.15/1 0}
6. Qxd5 f5 {(e7e5) +20.93/1 0} 7. Qa8 Be6 {(c8a6) +20.91/1 0} 8. Qxa7 Bb3
{(c7c6) +21.10/1} 9. Qb7 e6 {(b3c4) +20.93/1 0} 10. Qxb3 Qh4+ {(d6d5)
+17.83/1} 11. Ke2 d5 {(h4e4) +17.36/1 0} 12. Qxb8+ Kf7 {(e8d7) +14.03/1}
13. Qxc7+ Be7 {(f8e7) +13.66/1 0} 14. Qc3 Qh5+ {(h8a8) +14.75/1 0} 15. Kf2
Qh2+ {(h8a8) +15.53/1 0} 16. Ke1 Qh1+ {(e6e5) +15.14/1} 17. Ke2 Bb4 {(h8a8)
+14.73/1 0} 18. Qxh8 Qg1 {(h1e1) +10.60/1 0} 19. Qxh6 Be1 {(g1e1) +9.44/1
0} 20. Qh7+ Kf8 {(f7f6) +8.39/1 0} 21. Qh8+ Ke7 {(f8e7) +8.39/1} 22. Qg7+
Kd6 {(e7d6) +8.44/1} 23. Qf8+ Kc7 {(d6e5) +8.71/1 0} 24. Qe7+ Kb8 {(c7b6)
+7.34/1} 25. Qxe6 Qc5 {(g1f2) +8.57/1 0} 26. Kxe1 Qg1+ {(c5c3) +3.38/1 0}
27. Ke2 Qc1 {(d5d4) +3.32/1} 28. Qxd5 Qb2+ {(c1c3) +2.25/1 0} 29. Ke3 Kc7
{(b2c3) +2.15/1 0} 30. Qf7+ Kc6 {(c7b8) +0.76/1} 31. Qxg6+ Kb7 {(c6c5)
-0.05/1} 32. Qxf5 Qb6+ {(b2c3) +0.17/1 0} 33. Kd3 Qb5+ {(b6b3) +0.16/1 0}
34. Qxb5+ Ka7 {(b7a8) -9.90/1} 35. Kd4 Ka8 36. Kc5 Ka7 37. Kc6 Ka8 38. Qb7#
1-0
[/pgn]

[pgn][Event "Computer chess game"]
[Site "URIBLASS-THINK"]
[Date "2019.07.23"]
[Round "?"]
[White "UriBlass"]
[Black "BrutusRND"]
[Result "1-0"]
[Time "05:58:12"]
[WhiteElo "2400"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/1P6/4K3 w kq - 0 1"]
[Termination "normal"]
[PlyCount "91"]
[WhiteType "human"]
[BlackType "program"]

1. b4 f5 {(e7e5) +39.89/1 0} 2. b5 b6 {(b7b6) +41.02/1} 3. Ke2 h6 {(a7a6)
+40.72/1} 4. Kf3 Na6 {(c7c6) +39.77/1 0} 5. bxa6 Rb8 {(c8a6) +36.18/1} 6.
Kf4 c5 {(e7e6) +36.81/1} 7. Kxf5 b5 {(c8a6) +36.92/1} 8. Ke5 b4 {(c8a6)
+36.73/1} 9. Kd5 Qb6 {(c8a6) +36.50/1} 10. Kc4 Qa5 {(c8a6) +40.38/1} 11.
Kd5 Rb7 {(c8a6) +37.80/1} 12. axb7 Kf7 {(c8b7) +35.44/1} 13. bxc8=Q Qb6
{(g8f6) +19.44/1 0} 14. Qxd7 Qe6+ {(g8f6) +18.11/1 0} 15. Kxc5 Kf6 {(e6d7)
+26.75/1} 16. Qd4+ Kf5 {(e6e5) +16.18/1 0} 17. Qxb4 h5 {(e6d6) +16.50/1 0}
18. Qb1+ Kf6 {(e6e4) +15.27/1 0} 19. Qf1+ Ke5 {(e6f5) +14.84/1 0} 20. Qxf8
g5 {(e6d5) +13.39/1 0} 21. Qg7+ Nf6 {(e6f6) +11.98/1 0} 22. Qxh8 a5 {(e6b6)
+9.46/1 0} 23. Qb8+ Ke4 {(e6d6) +9.61/1 0} 24. Qb1+ Kf3 {(e4e5) +8.20/1}
25. Qf1+ Kg3 {(f3e4) +8.22/1} 26. Qg1+ Kh3 {(g3f4) +8.07/1} 27. Qxg5 Qa6
{(e6d6) +7.72/1 0} 28. Qe3+ Kh2 {(h3g4) +6.14/1} 29. Qxe7 Nh7 {(h5h4)
+6.75/1 0} 30. Qxh7 Qe6 {(a6c8) +1.94/1} 31. Qxh5+ Kg3 {(h2g2) +1.11/1} 32.
Qg5+ Kh2 {(g3f2) +1.12/1} 33. Qf4+ Kg2 {(h2h1) +1.04/1 0} 34. Qd2+ Kf1
{(g2f3) +0.27/1} 35. Qf4+ Kg1 {(f1g1) +1.03/1 0} 36. Qd4+ Kh2 {(g1h2)
+1.03/1 0} 37. Qb2+ Kh1 {(h2g3) +1.11/1} 38. Qc1+ Kg2 {(h1g2) +1.11/1} 39.
Qd2+ Kf3 40. Qxa5 Qe1 {(f3e4) +0.36/1} 41. Qxe1 Kg2 {(f3f4) -9.69/1} 42.
Kd5 Kh3 {(g2h3) -9.87/1} 43. Qg1 Kh4 44. Ke5 Kh5 {(h4h5) -9.87/1} 45. Kf5
Kh4 {(h5h6) -9.87/1} 46. Qh2# 1-0
[/pgn]

[pgn][Event "Computer chess game"]
[Site "URIBLASS-THINK"]
[Date "2019.07.23"]
[Round "?"]
[White "UriBlass"]
[Black "BrutusRND"]
[Result "1-0"]
[Time "06:01:23"]
[WhiteElo "2400"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/2P5/4K3 w kq - 0 1"]
[Termination "normal"]
[PlyCount "89"]
[WhiteType "human"]
[BlackType "program"]

1. c4 f6 {(e7e5) +39.95/1 0} 2. c5 b5 {(e7e5) +39.09/1 0} 3. cxb6 e5
{(a7b6) +38.30/1} 4. b7 h6 {(f8b4) +40.96/1} 5. bxa8=Q Be7 {(b8c6) +24.06/1
0} 6. Qxb8 c5 {(a7a5) +21.86/1} 7. Qxa7 d5 {(f6f5) +21.79/1 0} 8. Qa4+ Kf8
{(c8d7) +21.19/1 0} 9. Qh4 Ke8 {(f6f5) +22.10/1} 10. Qh5+ Kf8 {(e8d7)
+21.13/1 0} 11. Qg6 c4 {(f6f5) +21.66/1} 12. Qc2 c3 {(f6f5) +22.62/1} 13.
Qxc3 Qe8 {(c8g4) +20.57/1 0} 14. Qc1 Bd7 {(g7g6) +20.78/1} 15. Qb1 Bc6
{(g7g6) +20.80/1} 16. Qd3 g6 {(e8a8) +20.97/1} 17. Qa6 Rh7 {(f6f5)
+21.20/1} 18. Qb6 Ba8 {(f6f5) +21.07/1} 19. Qe6 Qf7 {(e8c6) +20.70/1 0} 20.
Qc8+ Qe8 {(f7e8) +20.41/1 0} 21. Qa6 Qb5 {(f6f5) +20.98/1} 22. Qxb5 h5
{(f8g7) +10.50/1 0} 23. Qb8+ Bd8 {(f8f7) +7.08/1} 24. Qxd8+ Kg7 {(f8f7)
+3.88/1} 25. Qxa8 e4 {(d5d4) +4.12/1} 26. Qxd5 f5 {(f6f5) +2.84/1} 27. Qd7+
Kh6 {(g7h6) +2.91/1 0} 28. Qe8 g5 {(h7g7) +3.22/1 0} 29. Qxg8 Ra7 {(h7d7)
+0.60/1 0} 30. Qe6+ Kg7 {(h6g7) -0.96/1} 31. Qxf5 Ra1+ {(a7a1) -0.87/1} 32.
Ke2 Ra4 {(a1a2) -1.07/1 0} 33. Qxg5+ Kf7 {(g7f8) -3.02/1} 34. Qxh5+ Kg7
{(f7e6) -2.94/1 0} 35. Qg5+ Kf7 {(g7f8) -3.02/1} 36. Qd5+ Kg6 {(f7g6)
-3.02/1 0} 37. Qc6+ Kh5 {(g6f5) -8.20/1} 38. Qxa4 e3 {(e4e3) -9.87/1} 39.
Kxe3 Kg5 {(h5g6) -9.78/1} 40. Qf4+ Kh5 {(g5g6) -10.05/1} 41. Qg3 Kh6 42.
Kf4 Kh7 {(h6h7) -9.88/1} 43. Kf5 Kh8 {(h7h8) -9.88/1} 44. Kf6 Kh7 45. Qg7#
1-0
[/pgn]

[pgn][Event "Computer chess game"]
[Site "URIBLASS-THINK"]
[Date "2019.07.23"]
[Round "?"]
[White "UriBlass"]
[Black "BrutusRND"]
[Result "1-0"]
[Time "06:04:34"]
[WhiteElo "2400"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/3P4/4K3 w kq - 0 1"]
[Termination "normal"]
[PlyCount "63"]
[WhiteType "human"]
[BlackType "program"]

1. d4 f6 {(e7e6) +38.82/1 0} 2. d5 f5 {(d7d6) +40.91/1 0} 3. d6 a5 {(e7d6)
+38.91/1} 4. dxc7 Ra6 {(d8c7) +39.23/1} 5. cxb8=Q d5 {(a5a4) +27.18/1 0} 6.
Qe5 g6 {(e7e6) +27.08/1} 7. Qxh8 Ra7 {(g8f6) +21.79/1 0} 8. Qxg8 Qd6
{(e7e5) +18.65/1 0} 9. Qxh7 Qa3 {(e7e5) +18.69/1 0} 10. Qxg6+ Kd8 {(e8d8)
+16.00/1} 11. Qb6+ Kd7 {(d8d7) +11.47/1} 12. Qxa7 a4 {(e7e5) +13.10/1} 13.
Qb6 Ke8 {(e7e5) +12.71/1 0} 14. Qc7 Qe3+ {(a3e3) +11.52/1 0} 15. Kd1 b6
{(e3d3) +11.47/1 0} 16. Qxc8+ Kf7 17. Qxf5+ Ke8 {(f7g7) +6.18/1} 18. Qh5+
Kd8 {(e8d8) +6.12/1} 19. Qxd5+ Ke8 {(d8c8) +6.15/1} 20. Qc6+ Kf7 {(e8f7)
+5.07/1 0} 21. Qxa4 Qe1+ {(f8h6) +5.88/1 0} 22. Kxe1 b5 {(e7e5) -4.69/1}
23. Qxb5 Bg7 {(e7e6) -5.90/1} 24. Qf5+ Bf6 {(f7e8) -6.21/1} 25. Ke2 Kg8
{(e7e5) -5.82/1} 26. Qe6+ Kh7 {(g8h8) -6.37/1 0} 27. Kf3 Bc3 {(h7h8)
-6.34/1} 28. Qxe7+ Bg7 {(h7h8) -6.67/1} 29. Kg4 Kg8 {(h7h8) -6.71/1} 30.
Kf5 Bh6 {(g8h8) -6.71/1} 31. Kg6 Bg7 {(h6f4) -6.84/1 0} 32. Qxg7# 1-0
[/pgn]

[MikeB]

My own games against the random mover.
It seems that I can usually beat it even with one pawn.

I suspect that chess engines that do not play the opponent may perform worse than me.

my score against it is 4.5-0.5 so far when I play with one pawn against all the pieces and I gave different handicap in every game(starting with a different pawn).
Only in one game it captured my pawn when in 4 games my pawn promoted to queen and I won.

[pgn][Event "Computer chess game"]
[Site "URIBLASS-THINK"]
[Date "2019.07.23"]
[Round "?"]
[White "UriBlass"]
[Black "BrutusRND"]
[Result "1/2-1/2"]
[Time "05:51:51"]
[WhiteElo "2400"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/4P3/4K3 w kq - 0 1"]
[Termination "normal"]
[PlyCount "107"]
[WhiteType "human"]
[BlackType "program"]

1. e4 d5 {(a7a6) +40.72/1 0} 2. exd5 a6 {(d8d5) +39.89/1} 3. d6 Nd7 {(d8d6)
+39.94/1} 4. dxe7 h6 {(d8e7) +41.75/1} 5. Kd2 Rh7 {(d8e7) +40.05/1} 6. Kc3
Ngf6 {(d8e7) +40.31/1} 7. Kc2 Nh5 {(d8e7) +39.76/1} 8. Kc3 c6 {(d8e7)
+40.32/1} 9. Kc2 Qxe7 {(d8e7) +39.46/1} 10. Kc3 Ra7 {(b7b6) +40.55/1} 11.
Kc4 f5 {(b7b6) +41.17/1} 12. Kc3 Qe3+ {(b7b6) +40.69/1} 13. Kc4 Qe7 {(b7b6)
+42.20/1} 14. Kc3 c5 15. Kc4 Qh4+ {(b7b6) +41.22/1} 16. Kd3 Be7 {(b7b6)
+39.52/1} 17. Ke3 Ne5 {(b7b6) +40.57/1} 18. Kd2 Qe1+ {(b7b6) +41.37/1} 19.
Kxe1 f4 {(e5d3) +32.34/1 0} 20. Kf2 Nc4 {(e5g4) +32.06/1} 21. Ke2 b5
{(h5g3) +32.28/1} 22. Kd3 Bg5 {(c4e5) +32.06/1 0} 23. Ke4 Ne5 {(h5f6)
+32.14/1} 24. Kxe5 a5 {(g7g6) +27.17/1} 25. Kd5 Kd8 {(c5c4) +27.16/1} 26.
Kxc5 Rd7 {(b5b4) +26.02/1} 27. Kxb5 Bh4 {(g7g6) +26.66/1} 28. Kxa5 Re7
{(g7g6) +27.29/1} 29. Kb5 Rd7 {(g7g6) +26.34/1} 30. Kc4 Ke8 {(g7g6)
+25.95/1} 31. Kc5 Bg5 {(g7g6) +26.09/1} 32. Kc6 Bh4 {(g7g6) +26.40/1} 33.
Kc5 Rd5+ 34. Kxd5 Bf6 {(g7g6) +18.76/1} 35. Ke4 Be6 {(g7g6) +18.93/1} 36.
Kf3 Bd5+ {(g7g6) +19.00/1} 37. Ke2 Ba2 {(g7g6) +19.29/1} 38. Kf3 Kd7
{(g7g6) +18.77/1} 39. Ke4 f3 {(g7g6) +18.94/1} 40. Kxf3 Bd5+ {(a2d5)
+18.86/1 0} 41. Kg4 Bb2 {(g7g6) +17.82/1} 42. Kxh5 Ke8 {(d5f7) +14.88/1 0}
43. Kg6 Bb7 {(d5e4) +15.18/1 0} 44. Kxh7 Ba8 {(h6h5) +10.10/1} 45. Kg6 Ba1
{(a8e4) +9.85/1 0} 46. Kf5 Bb7 {(a8d5) +8.75/1 0} 47. Kg4 Be4 {(b7c8)
+9.89/1 0} 48. Kf4 Bg6 {(e4d5) +8.81/1 0} 49. Kg4 Bh5+ {(h6h5) +10.04/1}
50. Kxh5 Kd7 {(a1e5) +5.34/1 0} 51. Kg6 Ke6 {(a1e5) +5.32/1 0} 52. Kh5 g5
{(a1e5) +5.34/1 0} 53. Kxh6 Be5 {(g5g4) +6.20/1} 54. Kxg5 {Insufficient
material} 1/2-1/2
[/pgn]

[pgn][Event "Computer chess game"]
[Site "URIBLASS-THINK"]
[Date "2019.07.23"]
[Round "?"]
[White "UriBlass"]
[Black "BrutusRND"]
[Result "1-0"]
[Time "05:55:17"]
[WhiteElo "2400"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/P7/4K3 w kq - 0 1"]
[Termination "normal"]
[PlyCount "75"]
[WhiteType "human"]
[BlackType "program"]

1. a4 g6 {(b7b6) +39.76/1 0} 2. a5 h6 {(a7a6) +39.80/1} 3. a6 d6 {(b7b6)
+40.33/1 0} 4. axb7 Nf6 {(c8b7) +37.84/1} 5. bxa8=Q Nd5 {(c8a6) +23.15/1 0}
6. Qxd5 f5 {(e7e5) +20.93/1 0} 7. Qa8 Be6 {(c8a6) +20.91/1 0} 8. Qxa7 Bb3
{(c7c6) +21.10/1} 9. Qb7 e6 {(b3c4) +20.93/1 0} 10. Qxb3 Qh4+ {(d6d5)
+17.83/1} 11. Ke2 d5 {(h4e4) +17.36/1 0} 12. Qxb8+ Kf7 {(e8d7) +14.03/1}
13. Qxc7+ Be7 {(f8e7) +13.66/1 0} 14. Qc3 Qh5+ {(h8a8) +14.75/1 0} 15. Kf2
Qh2+ {(h8a8) +15.53/1 0} 16. Ke1 Qh1+ {(e6e5) +15.14/1} 17. Ke2 Bb4 {(h8a8)
+14.73/1 0} 18. Qxh8 Qg1 {(h1e1) +10.60/1 0} 19. Qxh6 Be1 {(g1e1) +9.44/1
0} 20. Qh7+ Kf8 {(f7f6) +8.39/1 0} 21. Qh8+ Ke7 {(f8e7) +8.39/1} 22. Qg7+
Kd6 {(e7d6) +8.44/1} 23. Qf8+ Kc7 {(d6e5) +8.71/1 0} 24. Qe7+ Kb8 {(c7b6)
+7.34/1} 25. Qxe6 Qc5 {(g1f2) +8.57/1 0} 26. Kxe1 Qg1+ {(c5c3) +3.38/1 0}
27. Ke2 Qc1 {(d5d4) +3.32/1} 28. Qxd5 Qb2+ {(c1c3) +2.25/1 0} 29. Ke3 Kc7
{(b2c3) +2.15/1 0} 30. Qf7+ Kc6 {(c7b8) +0.76/1} 31. Qxg6+ Kb7 {(c6c5)
-0.05/1} 32. Qxf5 Qb6+ {(b2c3) +0.17/1 0} 33. Kd3 Qb5+ {(b6b3) +0.16/1 0}
34. Qxb5+ Ka7 {(b7a8) -9.90/1} 35. Kd4 Ka8 36. Kc5 Ka7 37. Kc6 Ka8 38. Qb7#
1-0
[/pgn]

[pgn][Event "Computer chess game"]
[Site "URIBLASS-THINK"]
[Date "2019.07.23"]
[Round "?"]
[White "UriBlass"]
[Black "BrutusRND"]
[Result "1-0"]
[Time "05:58:12"]
[WhiteElo "2400"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/1P6/4K3 w kq - 0 1"]
[Termination "normal"]
[PlyCount "91"]
[WhiteType "human"]
[BlackType "program"]

1. b4 f5 {(e7e5) +39.89/1 0} 2. b5 b6 {(b7b6) +41.02/1} 3. Ke2 h6 {(a7a6)
+40.72/1} 4. Kf3 Na6 {(c7c6) +39.77/1 0} 5. bxa6 Rb8 {(c8a6) +36.18/1} 6.
Kf4 c5 {(e7e6) +36.81/1} 7. Kxf5 b5 {(c8a6) +36.92/1} 8. Ke5 b4 {(c8a6)
+36.73/1} 9. Kd5 Qb6 {(c8a6) +36.50/1} 10. Kc4 Qa5 {(c8a6) +40.38/1} 11.
Kd5 Rb7 {(c8a6) +37.80/1} 12. axb7 Kf7 {(c8b7) +35.44/1} 13. bxc8=Q Qb6
{(g8f6) +19.44/1 0} 14. Qxd7 Qe6+ {(g8f6) +18.11/1 0} 15. Kxc5 Kf6 {(e6d7)
+26.75/1} 16. Qd4+ Kf5 {(e6e5) +16.18/1 0} 17. Qxb4 h5 {(e6d6) +16.50/1 0}
18. Qb1+ Kf6 {(e6e4) +15.27/1 0} 19. Qf1+ Ke5 {(e6f5) +14.84/1 0} 20. Qxf8
g5 {(e6d5) +13.39/1 0} 21. Qg7+ Nf6 {(e6f6) +11.98/1 0} 22. Qxh8 a5 {(e6b6)
+9.46/1 0} 23. Qb8+ Ke4 {(e6d6) +9.61/1 0} 24. Qb1+ Kf3 {(e4e5) +8.20/1}
25. Qf1+ Kg3 {(f3e4) +8.22/1} 26. Qg1+ Kh3 {(g3f4) +8.07/1} 27. Qxg5 Qa6
{(e6d6) +7.72/1 0} 28. Qe3+ Kh2 {(h3g4) +6.14/1} 29. Qxe7 Nh7 {(h5h4)
+6.75/1 0} 30. Qxh7 Qe6 {(a6c8) +1.94/1} 31. Qxh5+ Kg3 {(h2g2) +1.11/1} 32.
Qg5+ Kh2 {(g3f2) +1.12/1} 33. Qf4+ Kg2 {(h2h1) +1.04/1 0} 34. Qd2+ Kf1
{(g2f3) +0.27/1} 35. Qf4+ Kg1 {(f1g1) +1.03/1 0} 36. Qd4+ Kh2 {(g1h2)
+1.03/1 0} 37. Qb2+ Kh1 {(h2g3) +1.11/1} 38. Qc1+ Kg2 {(h1g2) +1.11/1} 39.
Qd2+ Kf3 40. Qxa5 Qe1 {(f3e4) +0.36/1} 41. Qxe1 Kg2 {(f3f4) -9.69/1} 42.
Kd5 Kh3 {(g2h3) -9.87/1} 43. Qg1 Kh4 44. Ke5 Kh5 {(h4h5) -9.87/1} 45. Kf5
Kh4 {(h5h6) -9.87/1} 46. Qh2# 1-0
[/pgn]

[pgn][Event "Computer chess game"]
[Site "URIBLASS-THINK"]
[Date "2019.07.23"]
[Round "?"]
[White "UriBlass"]
[Black "BrutusRND"]
[Result "1-0"]
[Time "06:01:23"]
[WhiteElo "2400"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/2P5/4K3 w kq - 0 1"]
[Termination "normal"]
[PlyCount "89"]
[WhiteType "human"]
[BlackType "program"]

1. c4 f6 {(e7e5) +39.95/1 0} 2. c5 b5 {(e7e5) +39.09/1 0} 3. cxb6 e5
{(a7b6) +38.30/1} 4. b7 h6 {(f8b4) +40.96/1} 5. bxa8=Q Be7 {(b8c6) +24.06/1
0} 6. Qxb8 c5 {(a7a5) +21.86/1} 7. Qxa7 d5 {(f6f5) +21.79/1 0} 8. Qa4+ Kf8
{(c8d7) +21.19/1 0} 9. Qh4 Ke8 {(f6f5) +22.10/1} 10. Qh5+ Kf8 {(e8d7)
+21.13/1 0} 11. Qg6 c4 {(f6f5) +21.66/1} 12. Qc2 c3 {(f6f5) +22.62/1} 13.
Qxc3 Qe8 {(c8g4) +20.57/1 0} 14. Qc1 Bd7 {(g7g6) +20.78/1} 15. Qb1 Bc6
{(g7g6) +20.80/1} 16. Qd3 g6 {(e8a8) +20.97/1} 17. Qa6 Rh7 {(f6f5)
+21.20/1} 18. Qb6 Ba8 {(f6f5) +21.07/1} 19. Qe6 Qf7 {(e8c6) +20.70/1 0} 20.
Qc8+ Qe8 {(f7e8) +20.41/1 0} 21. Qa6 Qb5 {(f6f5) +20.98/1} 22. Qxb5 h5
{(f8g7) +10.50/1 0} 23. Qb8+ Bd8 {(f8f7) +7.08/1} 24. Qxd8+ Kg7 {(f8f7)
+3.88/1} 25. Qxa8 e4 {(d5d4) +4.12/1} 26. Qxd5 f5 {(f6f5) +2.84/1} 27. Qd7+
Kh6 {(g7h6) +2.91/1 0} 28. Qe8 g5 {(h7g7) +3.22/1 0} 29. Qxg8 Ra7 {(h7d7)
+0.60/1 0} 30. Qe6+ Kg7 {(h6g7) -0.96/1} 31. Qxf5 Ra1+ {(a7a1) -0.87/1} 32.
Ke2 Ra4 {(a1a2) -1.07/1 0} 33. Qxg5+ Kf7 {(g7f8) -3.02/1} 34. Qxh5+ Kg7
{(f7e6) -2.94/1 0} 35. Qg5+ Kf7 {(g7f8) -3.02/1} 36. Qd5+ Kg6 {(f7g6)
-3.02/1 0} 37. Qc6+ Kh5 {(g6f5) -8.20/1} 38. Qxa4 e3 {(e4e3) -9.87/1} 39.
Kxe3 Kg5 {(h5g6) -9.78/1} 40. Qf4+ Kh5 {(g5g6) -10.05/1} 41. Qg3 Kh6 42.
Kf4 Kh7 {(h6h7) -9.88/1} 43. Kf5 Kh8 {(h7h8) -9.88/1} 44. Kf6 Kh7 45. Qg7#
1-0
[/pgn]

[pgn][Event "Computer chess game"]
[Site "URIBLASS-THINK"]
[Date "2019.07.23"]
[Round "?"]
[White "UriBlass"]
[Black "BrutusRND"]
[Result "1-0"]
[Time "06:04:34"]
[WhiteElo "2400"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/3P4/4K3 w kq - 0 1"]
[Termination "normal"]
[PlyCount "63"]
[WhiteType "human"]
[BlackType "program"]

1. d4 f6 {(e7e6) +38.82/1 0} 2. d5 f5 {(d7d6) +40.91/1 0} 3. d6 a5 {(e7d6)
+38.91/1} 4. dxc7 Ra6 {(d8c7) +39.23/1} 5. cxb8=Q d5 {(a5a4) +27.18/1 0} 6.
Qe5 g6 {(e7e6) +27.08/1} 7. Qxh8 Ra7 {(g8f6) +21.79/1 0} 8. Qxg8 Qd6
{(e7e5) +18.65/1 0} 9. Qxh7 Qa3 {(e7e5) +18.69/1 0} 10. Qxg6+ Kd8 {(e8d8)
+16.00/1} 11. Qb6+ Kd7 {(d8d7) +11.47/1} 12. Qxa7 a4 {(e7e5) +13.10/1} 13.
Qb6 Ke8 {(e7e5) +12.71/1 0} 14. Qc7 Qe3+ {(a3e3) +11.52/1 0} 15. Kd1 b6
{(e3d3) +11.47/1 0} 16. Qxc8+ Kf7 17. Qxf5+ Ke8 {(f7g7) +6.18/1} 18. Qh5+
Kd8 {(e8d8) +6.12/1} 19. Qxd5+ Ke8 {(d8c8) +6.15/1} 20. Qc6+ Kf7 {(e8f7)
+5.07/1 0} 21. Qxa4 Qe1+ {(f8h6) +5.88/1 0} 22. Kxe1 b5 {(e7e5) -4.69/1}
23. Qxb5 Bg7 {(e7e6) -5.90/1} 24. Qf5+ Bf6 {(f7e8) -6.21/1} 25. Ke2 Kg8
{(e7e5) -5.82/1} 26. Qe6+ Kh7 {(g8h8) -6.37/1 0} 27. Kf3 Bc3 {(h7h8)
-6.34/1} 28. Qxe7+ Bg7 {(h7h8) -6.67/1} 29. Kg4 Kg8 {(h7h8) -6.71/1} 30.
Kf5 Bh6 {(g8h8) -6.71/1} 31. Kg6 Bg7 {(h6f4) -6.84/1 0} 32. Qxg7# 1-0
[/pgn]

haha - makes you wonder how they ever got a draw...

[Uri Blass]

If a random moving engine played an infinite number of games against SF, some of the games would be wins for the random engine.
Obviously, a very tiny percentage.

But, just by chance, a win could happen early in the chain.

A random engine will not be zero Elo.
An engine that chooses the worst possible move after a careful search might be able to hit zero.

You mean it is not going to have the lowest elo.

I think the elo of random players should be less than 0 in the CCRL list if they accept the right engines.
unfortunately they test a lot of buggy engines.

I guess that many engines can beat the random player 500000-0 with one ply search(assuming one ply search is enough for them not to miss mate in the next move because they extend checks in the qsearch and also assuming one ply search is enough for them to avoid making a move that cause a repetition when they have the advantage or to avoid making stalemate because stalemate is evaluated by them as 0.00).

of course the random player may be lucky to get a position that it has mate in 2 but it is not going to happen often and even when it happens the probability of it to play the right move is small enough that I guess that the combination of these events is going to happen less than once in million games.

If one looks at the random engine games, one will see a random engine will get draws by buggy implementation of the draw rules. Most Elo systems in used by National or International Federations do incorporate a floor in their rating systems that no one has a negative rating. Also to Dann's point, a random engine will have a higher rating that an engine desired to play the worse possible moves such engine will be significantly lower that it would be most certainly be negative without a floor. All engines always have bugs, so to exclude any engine that has bugs would exclude all engines. Good luck with that suggestion. ;>)

Not all engines have bugs that cause them to lose points by stupid repetition or stupid stalemates like the following games

stockfish at depth 1 beat the random engine in every game that I tried but unfortunately the CCRL does not test it as a reference.

[pgn][Event "CCRL 40/4"]
[Site "CCRL"]
[Date "2012.08.15"]
[Round "173.2.428"]
[White "MicroChess 1976"]
[Black "Brutus RND"]
[Result "1/2-1/2"]
[BlackElo "259"]
[ECO "B06"]
[Opening "Modern"]
[Variation "1.e4 g6"]
[Termination "normal"]
[PlyCount "56"]
[WhiteType "human"]
[BlackType "human"]

1. e4 g6 2. Qg4 h6 3. d4 Nf6 4. Qf3 Ng8 5. Bb5 a5 6. e5 f5 7. Bd2 Na6 8.
Bxa5 Kf7 9. Bc4+ e6 10. Nd2 Ba3 11. bxa3 Qf8 12. a4 Qa3 13. Qxa3 b6 14. Bc3
d6 15. exd6 c6 16. Rb1 Nc7 17. dxc7 Ra6 18. Qd6 Rxa4 19. Qxc6 Ne7 20. Qxa4
Bb7 21. Rxb6 Rc8 22. Bxe6+ Kf8 23. Rxb7 g5 24. Ngf3 Ng8 25. Qa3+ Ke8 26.
Qa4+ Kf8 27. Qa3+ Ke8 28. Qa4+ Kf8 1/2-1/2
[/pgn]

[pgn][Event "CCRL 40/4"]
[Site "CCRL"]
[Date "2012.08.09"]
[Round "173.3.535"]
[White "Brutus RND"]
[Black "CPP1"]
[Result "1/2-1/2"]
[BlackElo "459"]
[ECO "A05"]
[Opening "Reti Opening"]
[Variation "1...Nf6 2.e3"]
[WhiteElo "259"]
[Termination "normal"]
[PlyCount "112"]
[WhiteType "human"]
[BlackType "human"]

1. Nf3 Nf6 2. e3 c5 3. b3 g6 4. Bb2 Bg7 5. d4 cxd4 6. Nxd4 Nc6 7. Be2 Nxd4
8. O-O Nc6 9. Qd5 Nxd5 10. Bd4 a5 11. Ba6 bxa6 12. Bc3 Nxc3 13. f3 Nb5 14.
Na3 Nxa3 15. Rab1 Nxc2 16. Rbc1 Nxe3 17. a3 Na7 18. Rc2 Nxc2 19. Rd1 Nxa3
20. Rd4 Bxd4+ 21. Kh1 Nc6 22. h3 Bb7 23. b4 axb4 24. f4 a5 25. h4 Bb6 26.
Kh2 Bd4 27. Kh1 Bf6 28. g3 h5 29. Kg1 Bd4+ 30. Kh1 Bc3 31. Kh2 Bd4 32. Kg2
Nb8+ 33. Kh3 Be4 34. Kh2 e6 35. Kh3 Bf5+ 36. Kh2 Bf6 37. Kh1 Ke7 38. Kg1
Ra6 39. Kh2 Nc6 40. Kg1 Ra7 41. g4 hxg4 42. h5 gxh5 43. Kh2 h4 44. Kg2 d5
45. Kg1 d4 46. Kf2 h3 47. Kg1 Rd7 48. Kf2 Re8 49. Ke2 h2 50. Kd1 h1=Q+ 51.
Ke2 Qg2+ 52. Ke1 Qe4+ 53. Kf2 Qxf4+ 54. Kg2 Qe4+ 55. Kf1 Qf3+ 56. Kg1 g3
{Stalemate} 1/2-1/2
[/pgn]

some games of CCRL have also wrong draw result

[pgn][Event "CCRL 40/4"]
[Site "CCRL"]
[Date "2012.08.09"]
[Round "173.3.579"]
[White "Brutus RND"]
[Black "LaMoSca 0.10"]
[Result "1/2-1/2"]
[BlackElo "2200"]
[ECO "A17"]
[Opening "English"]
[Variation "Anglo-Queen's Indian, 4.g3 Bb7 5.Bg2 Be7 6.O-O O-O 7.Re1"]
[WhiteElo "259"]
[Termination "normal"]
[PlyCount "103"]
[WhiteType "human"]
[BlackType "program"]

1. Nf3 Nf6 2. c4 b6 3. g3 Bb7 4. Bg2 e6 5. O-O Be7 6. Nc3 O-O 7. Re1 d5 8.
h4 Qc8 9. Qb3 Nbd7 10. c5 Qe8 11. Qa3 Qd8 12. Nh2 Nxc5 13. Bh3 Nce4 14. Kg2
Bxa3 15. b4 d4 16. h5 Bxc1 17. Nd1 Ng5+ 18. e4 Bxd2 19. Nf3 Bxe4 20. Kh1
Bxe1 21. h6 Nxh3 22. g4 Bxf3+ 23. Kh2 Qd6+ 24. Kxh3 g5 25. a4 a6 26. Nb2
Bxg4+ 27. Kg2 Bc3 28. Rd1 Bf3+ 29. Kg1 Ng4 30. Rd3 Qh2+ 31. Kf1 Bg2+ 32.
Ke2 Bxb2 33. f3 Ne5 34. a5 Bxf3+ 35. Kf1 Nxd3 36. axb6 cxb6 37. b5 f5 38.
bxa6 Ne5 39. a7 Be4 40. Ke1 Bc3+ 41. Kd1 d3 42. Kc1 Bd4 43. Kb1 d2+ 44. Ka2
f4 45. Kb3 Rxa7 46. Kb4 f3 47. Kb3 Ra5 48. Kb4 Ra6 49. Kb3 Ra5 50. Kb4 Ra6
51. Kb5 Ra5+ 52. Kb4 1/2-1/2
[/pgn]

[MikeB]

<snip>
[pgn][Event "CCRL 40/4"]
[Site "CCRL"]
[Date "2012.08.09"]
[Round "173.3.535"]
[White "Brutus RND"]
[Black "CPP1"]
[Result "1/2-1/2"]
[BlackElo "459"]
[ECO "A05"]
[Opening "Reti Opening"]
[Variation "1...Nf6 2.e3"]
[WhiteElo "259"]
[Termination "normal"]
[PlyCount "112"]
[WhiteType "human"]
[BlackType "human"]

1. Nf3 Nf6 2. e3 c5 3. b3 g6 4. Bb2 Bg7 5. d4 cxd4 6. Nxd4 Nc6 7. Be2 Nxd4
8. O-O Nc6 9. Qd5 Nxd5 10. Bd4 a5 11. Ba6 bxa6 12. Bc3 Nxc3 13. f3 Nb5 14.
Na3 Nxa3 15. Rab1 Nxc2 16. Rbc1 Nxe3 17. a3 Na7 18. Rc2 Nxc2 19. Rd1 Nxa3
20. Rd4 Bxd4+ 21. Kh1 Nc6 22. h3 Bb7 23. b4 axb4 24. f4 a5 25. h4 Bb6 26.
Kh2 Bd4 27. Kh1 Bf6 28. g3 h5 29. Kg1 Bd4+ 30. Kh1 Bc3 31. Kh2 Bd4 32. Kg2
Nb8+ 33. Kh3 Be4 34. Kh2 e6 35. Kh3 Bf5+ 36. Kh2 Bf6 37. Kh1 Ke7 38. Kg1
Ra6 39. Kh2 Nc6 40. Kg1 Ra7 41. g4 hxg4 42. h5 gxh5 43. Kh2 h4 44. Kg2 d5
45. Kg1 d4 46. Kf2 h3 47. Kg1 Rd7 48. Kf2 Re8 49. Ke2 h2 50. Kd1 h1=Q+ 51.
Ke2 Qg2+ 52. Ke1 Qe4+ 53. Kf2 Qxf4+ 54. Kg2 Qe4+ 55. Kf1 Qf3+ 56. Kg1 g3
{Stalemate} 1/2-1/2
[/pgn]

some games of CCRL have also wrong draw result

[pgn][Event "CCRL 40/4"]
[Site "CCRL"]
[Date "2012.08.09"]
[Round "173.3.579"]
[White "Brutus RND"]
[Black "LaMoSca 0.10"]
[Result "1/2-1/2"]
[BlackElo "2200"]
[ECO "A17"]
[Opening "English"]
[Variation "Anglo-Queen's Indian, 4.g3 Bb7 5.Bg2 Be7 6.O-O O-O 7.Re1"]
[WhiteElo "259"]
[Termination "normal"]
[PlyCount "103"]
[WhiteType "human"]
[BlackType "program"]

1. Nf3 Nf6 2. c4 b6 3. g3 Bb7 4. Bg2 e6 5. O-O Be7 6. Nc3 O-O 7. Re1 d5 8.
h4 Qc8 9. Qb3 Nbd7 10. c5 Qe8 11. Qa3 Qd8 12. Nh2 Nxc5 13. Bh3 Nce4 14. Kg2
Bxa3 15. b4 d4 16. h5 Bxc1 17. Nd1 Ng5+ 18. e4 Bxd2 19. Nf3 Bxe4 20. Kh1
Bxe1 21. h6 Nxh3 22. g4 Bxf3+ 23. Kh2 Qd6+ 24. Kxh3 g5 25. a4 a6 26. Nb2
Bxg4+ 27. Kg2 Bc3 28. Rd1 Bf3+ 29. Kg1 Ng4 30. Rd3 Qh2+ 31. Kf1 Bg2+ 32.
Ke2 Bxb2 33. f3 Ne5 34. a5 Bxf3+ 35. Kf1 Nxd3 36. axb6 cxb6 37. b5 f5 38.
bxa6 Ne5 39. a7 Be4 40. Ke1 Bc3+ 41. Kd1 d3 42. Kc1 Bd4 43. Kb1 d2+ 44. Ka2
f4 45. Kb3 Rxa7 46. Kb4 f3 47. Kb3 Ra5 48. Kb4 Ra6 49. Kb3 Ra5 50. Kb4 Ra6
51. Kb5 Ra5+ 52. Kb4 1/2-1/2
[/pgn]

You must be getting old Uri ;>) . The last two games you posted were drawn - one by stalemate and the one below it by 3-fold repetition. Unless you meant some other game.

[Ovyron]

A random engine will not be zero Elo.
An engine that chooses the worst possible move after a careful search might be able to hit zero.

The referenced video already explores that, and you're right, if you make Stockfish sort the moves from best to worse, and make the worst one every time, it'll achieve the lowest possible elo. It'll be interesting if someone can come up with a weaker playing algorithm.

If you give this weakest opponent 0 elo, then Random has 269.8 elo.

Note: Some of these algorithms were based on downloading a huge lichess database and studying how likely were pieces to survive in a given square (noted as *). Others where moves are sorted alphabetically by ascii, an they're given an Interval from 0 to 1, then a number is used as seed, and depending where it lands on the interval a move is picked (noted as **).

The bottom of the elo system for tested algorithms looks like this:

269.8 elo = Random Mover
268.8 elo = Survivalist (moves pieces where they're most likely to survive, by captured/not captured ratio*)
267.9 elo = Interval Pi -3**
267.7 elo = Opposite Color (as white tries to put pieces in black squares, as black tries to put pieces in white squares)
267.5 elo = Same Color (as white tries to put pieces in white squares, as black tries to put pieces in black squares)
267.5 elo = Interval Euler's number -2**
265.8 elo = Sym 180 (tries to create a board where white and black's pieces would be symmetrical if rotated 180 degrees, ignoring piece value)
246.4 elo = Interval Binary Pi** (either plays first alphabetical move or last alphabetical move depending on Pi's depicition in binary)
231.8 elo = Huddle (Minimize the number of moves own King would need to reach the square of another friendly piece)
213.3 elo = No I Insist (Suicide personality that purposely avoids mating and stalemate, makes moves that give away material, and goes into positions where opponent is forced to capture material)
189.7 elo = First Move (Makes first possible move according to board coordinates)
184.9 elo = Rare (moves pieces towards squares they're least likely to end the game in)*
174.6 elo = Dangerous (moves pieces towards squares they're least likely to survive)*
173.2 elo = Generous (makes moves that give away material. It's interesting that it performs worse than No I Insist)
164.2 elo = Popular (moves pieces towards squares they're played to the most)*
159.9 elo = Safe (moves pieces where they're most likely to survive)*
127.1 elo = Alphabetical (Interval 0**. Note that when sorting by alphabetical ascii, all uppercase letters appear before lowercase letters, so this one would move major pieces unless they can't move)
104.6 elo = Pacifist (random mover that avoids mating, checking, capturing, and if it's forced to capture, captures piece of worst value)
000.0 elo = Worstfish (plays Stockfish's worst scoring move)

Under this scale, Stockfish limited to 1000000 nodes has 2500.6 elo.

So if Stockfish 1000000 nodes beats Worstfirst 100% of the time, then elo doesn't have a way to show a difference higher than 2500.6 elo.

[Guenther]

If a random moving engine played an infinite number of games against SF, some of the games would be wins for the random engine.
Obviously, a very tiny percentage.

But, just by chance, a win could happen early in the chain.

A random engine will not be zero Elo.
An engine that chooses the worst possible move after a careful search might be able to hit zero.

Old stuff about possible rating difference to worst players.
I thought though we agreed on 'negative Elo' for worst movers and scaled plain random to zero?

http://talkchess.com/forum3/viewtopic.p ... 20#p699420

[Guenther]

If a random moving engine played an infinite number of games against SF, some of the games would be wins for the random engine.
Obviously, a very tiny percentage.

But, just by chance, a win could happen early in the chain.

A random engine will not be zero Elo.
An engine that chooses the worst possible move after a careful search might be able to hit zero.

Old stuff about possible rating difference to worst players.
I thought though we agreed on 'negative Elo' for worst movers and scaled plain random to zero?

http://talkchess.com/forum3/viewtopic.p ... 20#p699420

(One has to check the whole thread, especially because the first posted results were wrong due to some bugs
in AndscacsWorst)

[Zenmastur]

Obviously, statistics don't handle "100%" or "0%" conditions very well. Which is why the +0.5 helps out. There's no rigorous theory however with the +0.5 methodology, but it does allow you to avoid the "division by zero" and carry on.

It let's you carry on, but that doesn't mean the results are anywhere near correct for an "always winner":

Win% elo diff
0.9 382
0.99 798
0.999 1200
0.9999 1600
0.99999 2000
0.999999 2400

So you could look at a 500 game match and say that they are 1600 elo stronger. Then after playing 500,000 games you find your original estimate was off by 800+ elo!

If a player can be good enough to always force a win, there is no limit on elo. So the question becomes one of how well can the weaker programs force draws.

Well, yes. That's why I prefer the acyclic graphs with topological sorting. There is a problem with topological sorts of acyclic graphs on the opposite end, with only one game played.

If after 500 games, the player is still undefeated, the acyclic / topological sort methodology would assume "infinite" elo. After 500,000 games, the player finally loses and you learn that the Elo isn't "infinity", but in fact, is "only" +2400 elo.

Obviously, the acyclic + topological sort methodology is poorly behaved in the case of 1-win / zero losses / zero draws. While the "+0.5" method handles this case reasonably.

----------

How do you want to handle your division by zero? Do you want to under-estimate, or do you want to over-estimate? Because "infinite" is clearly the wrong answer, but +0.5 is clearly an underestimate in most cases. Personally speaking, I think there is elegance in calling it "infinite" Elo in the face of the evidence.

My statistics teacher many decades ago said that you should simply carry forth any experiment until you notice at least 10-successes and 10-failures, that is... if you want to have a valid statistical result. If you don't get 10+ successes and 10+ failures on any proportion, the math simply won't work out very well.

If you can't get the minimum number of 10, then you have to just live with weird / wonky math. In any case, the solution to these cases is to get more data. You can't squeeze blood out of a stone. Garbage-in / garbage out. Etc. etc. Whatever metaphor you want, if you don't have enough win/loss data, then you simply can't calculate any good statistics.

Well by using SF set to depth 1 and Ruyrandom and Cerebellum_Light_Poly.bin set to 5 full moves I got a score of 0, 9985, 15. Openings were set so each engine got to play both sides of the opening so it was balanced.

Code: Select all

Score of RuyRandom-sse3 vs stockfish_17121709_x64_bmi2: 0 - 9985 - 15 [0.001]
Elo difference: -1249.8 +/- 96.6, LOS: 0.0 %, DrawRatio: 0.1 %

10000 of 10000 games finished.

Now all I need to to be able to play SF against itself with one engine set to depth 1 and the other set to depth “X”. Where “X” is 10 to start with and can be adjusted until we get a measurable difference in score.

Does anyone know how to do that. I'm not good with cutechess or cutechess_cli as I haven't used either until now.

Regards,

Zenmastur

[Zenmastur]

Of the 15 draws, 2 were 3-move repetition and the 13 were stalemates. I guess a 6 or 7 ply search will solve both of these problems and make SF much stronger!

Regards,

Zenmastur