We have new number 1. engine - for solving

[Jouni]

Leptir TI MPV from 3.12. solves puzzles like mad! Just one example: it solves 105/110 in HTC suite with 1 minute and 6 cores.

[peter]

HTC (in latest version Vincent even left only 108 in the the formerly 220 positions big suite) is getting too easy for best solvers (in this one suite) for 60" and 6 threads, I'd say.

With 30" and 8t (of the16x4.3GHz-CPU) out of the 256, new MPV- Leptir solved 213 at a single run and comparable setting to before long time best RemsM-091224 (26 plies, MultiPV=4, Score 2000 for Rems, 500, which now is max. for Leptir), but almost a year old now, the former leader of the list yet still is unbeaten with 218/256.
StatTS Elo- wise even a little bigger difference compared to error bar because of better solution time indices:

Code: Select all

    Program                                    Elo   +/-  Matches  Score   Av.Op.   S.Pos.   MST1    MST2   RIndex

  1 RemsM-091224-8t-26-4-2000                : 3582    2  41827    61.9 %   3497   218/256    3.3s    7.3s   0.65
  2 LeptirM-251203-8t-26-4-500               : 3574    2  41397    60.9 %   3497   213/256    3.8s    8.2s   0.64

 178 Dragon1-MCTS-6t-MuPV4                   : 3306    2  30115    24.0 %   3507    42/256    7.5s   26.3s   0.06

MST1  : Mean solution time (solved positions only)
MST2  : Mean solution time (solved and unsolved positions)
RIndex: Score according to solution time ranking for each position

[Ajedrecista]

Hello Peter:

[...]

With 30" and 8t (of the16x4.3GHz-CPU) out of the 256, [...]
StatTS Elo- wise even a little bigger difference compared to error bar because of better solution time indices:

Code: Select all

    Program                                    Elo   +/-  Matches  Score   Av.Op.   S.Pos.   MST1    MST2   RIndex

  1 RemsM-091224-8t-26-4-2000                : 3582    2  41827    61.9 %   3497   218/256    3.3s    7.3s   0.65
  2 LeptirM-251203-8t-26-4-500               : 3574    2  41397    60.9 %   3497   213/256    3.8s    8.2s   0.64

 178 Dragon1-MCTS-6t-MuPV4                   : 3306    2  30115    24.0 %   3507    42/256    7.5s   26.3s   0.06

MST1  : Mean solution time (solved positions only)
MST2  : Mean solution time (solved and unsolved positions)
RIndex: Score according to solution time ranking for each position

I miss that the output of the tool does not include the maximum time given for each position, so it is not your fault. It is 30"/position because you kindly said it, but it should be included in the output of the tool, IMHO. Thank goodness that it can be estimated due to the own definition of MST1 and MST2, plus the amount of solved positions over the total:

Code: Select all

Tmax   := Maximum allowed time to solved each position

MST1   := Mean solution time (solved positions only)
MST2   := Mean solution time (solved and unsolved positions)

S.Pos. := Solved positions
U.Pos. := Unsolved positions

Pos.   := Positions

------------------------

Pos. = S.Pos. + U.Pos.

MST2 = ( S.Pos. * MST1 + U.Pos. * Tmax ) / ( S.Pos. + U.Pos. )
MST2 = [ S.Pos. * MST1 + ( Pos. - S.Pos. ) * Tmax ] / Pos.
MST2 = ( S.Pos. / Pos. ) * MST1 + [ 1 - ( S.Pos. / Pos. ) ] * Tmax

Tmax = [ MST2 - ( S.Pos. / Pos. ) * MST1 ] / [ 1 - ( S.Pos. / Pos. ) ]

------------------------

Extreme case:
S.Pos. = 0 => U.Pos. = Pos.
MST2 = Tmax
MST1 not available

Starting from MST1 and MST2 (rounded up to 0.1 seconds), I compute Tmax estimates for each engine up to 0.01 seconds:

Code: Select all

RemsM-091224-8t-26-4-2000  → Tmax = [  7.3 - ( 218 / 256 ) * 3.3 ] / [ 1 - ( 218 / 256 ) ] ~ 30.25 seconds
LeptirM-251203-8t-26-4-500 → Tmax = [  8.2 - ( 213 / 256 ) * 3.8 ] / [ 1 - ( 213 / 256 ) ] ~ 30.00 seconds
Dragon1-MCTS-6t-MuPV4      → Tmax = [ 26.3 - (  42 / 256 ) * 7.5 ] / [ 1 - (  42 / 256 ) ] ~ 29.99 seconds

One can guess that Tmax = 30" in this case, if the shared results accidentally did not include Tmax.

Tweaking a little more, one can safely/hopefully assume that MST1 and MST2 outputs are rounded and not truncated. If that, the true values are within half of the accuracy/resolution (0.05" in this case). Then, lower and upper bounds can be computed for each engine and the final interval of Tmax would be [ max(lower bounds) , min(upper bounds) ]:

Code: Select all

-1 =< k_s < 1  // _s stands for 'solved'
-1 =<  k  < 1

( True MST1 ) = ( Reported MST1 ) + k_s * Accuracy / 2
( True MST2 ) = ( Reported MST2 ) +  k  * Accuracy / 2

------------------------

LTmax := Lower bound of Tmax => ( k = -1 ) AND ( k_s →  1 )
UTmax := Upper bound of Tmax => ( k →  1 ) AND ( k_s = -1 )

Computing LTmax and UTmax for each engine:

Code: Select all

RemsM-091224-8t-26-4-2000  → LTmax = [  7.25 - ( 218 / 256 ) * 3.35 ] / [ 1 - ( 218 / 256 ) ] ~ 29.62 seconds
RemsM-091224-8t-26-4-2000  → UTmax = [  7.35 - ( 218 / 256 ) * 3.25 ] / [ 1 - ( 218 / 256 ) ] ~ 30.87 seconds

LeptirM-251203-8t-26-4-500 → LTmax = [  8.15 - ( 213 / 256 ) * 3.85 ] / [ 1 - ( 213 / 256 ) ] = 29.45 seconds
LeptirM-251203-8t-26-4-500 → UTmax = [  8.25 - ( 213 / 256 ) * 3.75 ] / [ 1 - ( 213 / 256 ) ] ~ 30.54 seconds

Dragon1-MCTS-6t-MuPV4      → LTmax = [ 26.25 - (  42 / 256 ) * 7.55 ] / [ 1 - (  42 / 256 ) ] ~ 29.92 seconds
Dragon1-MCTS-6t-MuPV4      → UTmax = [ 26.35 - (  42 / 256 ) * 7.45 ] / [ 1 - (  42 / 256 ) ] ~ 30.06 seconds

------------------------

MAX(LTmax_i) = MAX( 29.62 , 29.45 , 29.92 ) = 29.92 seconds
MIN(UTmax_i) = MIN( 30.87 , 30.54 , 30.06 ) = 30.06 seconds

########################
29.92" =< Tmax =< 30.06"
########################

Tmax ~ 29.99" ± 0.07"

That greatly reinforces the initial estimate of 30" of this case, with the data of only three engines, when you must have at least 178. I hope no typos.

Regards from Spain.

Ajedrecista.

[peter]

That greatly reinforces the initial estimate of 30" of this case, with the data of only three engines, when you must have at least 178. I hope no typos.

No typo, Jesus, 178 runs are compared to each other in quoted list, I use to just copy the parts of interest for the certain question, here the two first ones and the last one listed, regards

[Eelco de Groot]

Hi guys, I do not think I found anything you did not know already but could not find it in the CCET topic as an alternate solution for instance, I realize it is a study not a problem but for giving points to computers should it not be included? Peter's Salai study, the new Omega with ten PV's only gives f4 a 0.00 but other new compile from Eduard, Leptir with 5 PV's then gives two solutions (I don't know how long it takes with just one PV to find f4, I have not tried).


37 41:13 +6.97 1.f4 Lxf4 2.Kd7 Ld2 3.Kc6 La5 4.b4 cxb4 5.Lb3+ Kf8 6.Kd5 Ke8 7.Ke6 h5 8.Kf5 Ke7 9.Kg5 h4 10.Kxh4 Kd6 11.f4 Kc5 12.f5 Kxb5 13.f6 Kc6 14.Kh5 (2.525.986.242) 1021
37 41:13 +6.86 1.Ld1 h5 2.f4 Lf6 3.Lxh5 Kg7 4.Kd7 Lc3 5.Kc6 La5 6.a3 Kh7 7.f5 Kg7 8.b4 cxb4 9.axb4 Lxb4 10.Kxb6 Kf6 11.Lg4 Ke7 12.Kc6 Lc3 13.f3 Kd8 14.b6 (2.525.986.242) 1021
37 41:13 +1.56 1.Kd7 Ld2 2.Ke6 Kg7 3.a4 h5 4.Ld3 La5 5.f4 h4 6.Lf1 Lc3 7.f5 Kf8 8.f3 Ke8 9.f6 Kf8 10.f4 Ke8 11.Kf5 Kf7 12.Lc4+ Kf8 13.Kg4 Lxf6 14.a5 (2.525.986.242) 1021
37 41:13 +1.28 1.a4 Ld2 2.Ke7 h5 3.Ld3 Kg7 4.Ke6 Lb4 5.f4 Ld2 6.Ke7 h4 7.f5 Lg5+ 8.Ke6 Ld2 9.Lf1 Lc3 10.Ke7 Lf6+ 11.Kd7 Lc3 12.Ke6 Kf8 13.Lh3 Ke8 14.Lg4 (2.525.986.242) 1021
37 41:13 +1.20 1.Ld3 Kg7 2.Kd7 Ld2 3.Ke6 h5 4.a4 La5 5.Le2 Lb4 6.Lf1 Ld2 7.f4 h4 8.f5 Lc3 9.Lh3 Kf8 10.f3 Kg7 11.Ke7 Lf6+ 12.Kd6 Lc3 13.Ke6 Kf8 14.f6 (2.525.986.242) 1021

depth 38 for Bd1 takes longer to resolve, now at 62 minutes. It is not a complete transposition (that would go faster perhaps) or almost transposition because of h5 as Black's move in first move pair..

[Eelco de Groot]

In case 1.Bd1 is not a correct solution, there could be a Fail Low going on so I will wait a little more...

[Eelco de Groot]

Now it is resolved at depth 38:

[d]4K1k1/8/1p5p/1Pp3b1/8/1P3P2/P1B2P2/8 w - -

Engine: Leptir TI MPV (1024 MB)
gemaakt door E.N and the Stockfish developers (see A

38 100:45 +7.25 1.Ld1 Ld2 2.f4 Kg7 3.Kd7 Kg6 4.Kc6 La5
5.Lg4 Kf7 6.a3 Kf6 7.Kc7 Kg7 8.f5 Kf6
9.b4 cxb4 10.axb4 Lxb4 11.Kxb6 h5
12.Lxh5 Kxf5 13.f3 Ke6 14.Kc6 (5.730.697.888) 947

38 100:45 +7.25 1.f4 Lf6 2.Ld1 Lc3 3.Lg4 Kg7 4.Kd7 Kf7
5.Kc6 La5 6.a3 Ke8 7.f5 Ke7 8.Kc7 h5
9.Lxh5 Kf6 10.b4 cxb4 11.axb4 Lxb4
12.Kxb6 Kxf5 13.Kc6 Kg5 14.b6 (5.730.697.888) 947

38 100:45 +1.13 1.Kd7 Ld2 2.Ke6 Kg7 3.a4 h5 4.Ld3 La5
5.f4 h4 6.Lf1 Lc3 7.f5 Kf8 8.f3 La5
9.Lh3 Lc3 10.Lg4 Le1 11.Kf6 Lc3+
12.Kg5 Le1 13.f6 Kf7 14.Lh5+ (5.730.697.888) 947

38 100:45 +1.09 1.a4 Ld2 2.Ke7 h5 3.Ld3 h4 4.Lf1 Kg7
5.Ke6 La5 6.f4 Lc3 7.f5 Kf8 8.f6 La5
9.Kf5 Ld2 10.Lh3 Le1 11.f4 La5
12.Kg5 Le1 13.Le6 Ke8 14.Kh5 (5.730.697.888) 947

38 100:45 +1.08 1.Ld3 Kg7 2.Kd7 Ld2 3.Ke6 h5 4.a4 Lb4
5.f4 h4 6.Lf1 Lc3 7.f5 Kf8 8.f6 La5
9.Kf5 Ld2 10.Lh3 Le1 11.f4 Lc3
12.Lf1 Le1 13.Kg5 Kf7 14.Lc4+ (5.730.697.888) 947

[peter]

Hi Eelco!

1.Bd1 is a correct solution as well as author's one 1.f4 is, here I have copy- pasted some of the lines stored in my problem- database since quite a while, haven't checked the DTM- values all with latest engine- versions very lately anymore (e.g. with Huntsman), but they yet should be more or less exactly right as for that (distances to mate) too. 1.Bd1 starts at end of second row in reader- displayed .pgn, just telling so to make it easier to find it directly on screen also, yet I'd rather copy- paste it into GUI for better readability:

[pgn][Event "1st Prize, Sachova skladba"]
[Site "?"]
[Date "2011.04.??"]
[Round "?"]
[White "Salai, Ladislav Jr."]
[Black "Eret90, Major Dual 1.Bd1"]
[Result "1-0"]
[Annotator "3.Kc6 MH"]
[WhiteFideId "-1"]
[WhiteFideId "-1"]
[SetUp "1"]
[FEN "4K1k1/8/1p5p/1Pp3b1/8/1P3P2/P1B2P2/8 w - - 0 1"]
[PlyCount "59"]
[GameId "442446275578"]
[EventDate "2013.??.??"]
[EventType "game"]
[SourceTitle "pure42"]

1. f4 $1 (1. Kd7 $2 Bd2 2. Kc6 (2. f4 h5 3. Kc6 Ba5 4. Bd1 h4 5. b4 Bxb4 6. Kxb6 c4 7. a4 h3) 2... Ba5 3. b4 $1 cxb4 4. Bb3+ Kg7 5. Kd5 Kf6 6. Ke4 h5 $1 7. Kf4 Kg6 8. Ke5 h4 9. Kf4 Kh5) (1. Bd1 $5 Kg7 (1... Bh4 $5 2. Kd7 Bxf2 3. Kc6 h5 4. Kxb6 c4+ 5. Ka5 c3 6. Kb4 Be1 7. Kc4 h4 8. f4 Kf7 9. Kd3 Ke6 10. Bg4+ Kd5 11. f5 c2 12. Kxc2 Ke5 13. a4 Kf6 14. b6 h3 15. Bxh3 Bg3 16. b7 Kg5 17. Kc3 Kh4 18. Kc4 Be5 19. a5 Kg5 20. b4 Bb8 21. a6 Kf4 22. f6 Kg3 23. f7 Kxh3 24. f8=Q Ba7 25. b8=Q Bxb8 26. Qxb8 Kg4 27. a7 Kg5 28. Qg3+ Kf5 29. a8=Q Kf6 30. Qf8+ Ke6 31. Qgd6#) (1... Bd2 $5 2. f4 (2. Kd7 $2 {ZUS 1.Kd7}) 2... Bxf4 3. Kd7 Bd2 4. Kc6 Ba5 5. b4 cxb4 6. Bb3+ {ZUS 1.f4 + 1 Zug}) (1... Bf4 $5 2. Kd7 Kf7 3. Kc6 Ke6 4. Kxb6) 2. Kd7 Bd2 3. f4 Kg6 4. Bg4 h5 5. f5+ Kf6 6. Bxh5 Kxf5 7. Kc6 Ba5 8. a3 Ke5 9. Bg4 Kf4 10. f3 Ke5 11. b4 cxb4 12. axb4 Bxb4 13. Kxb6 Kd6 14. f4 Ke7 15. f5 Bc3 16. Kb7 Kd6 17. b6 Bd4 18. Ka6 Ke7 19. b7 Be5 20. Ka7 Bd4+ 21. Ka8 Be5 22. b8=Q Bxb8 23. Kxb8 Kf7 24. Kb7 Kf6 25. Kc6 Ke5 26. Kc7 Kf4 27. f6 Kxg4 28. f7 Kf3 29. f8=Q+ Ke3 30. Qf1 Kd4 31. Qe2 Kc3 32. Kc6 Kb3 33. Qd2 Ka3 34. Kc5 Kb3 35. Kb5 Ka3 36. Kc4 Ka4 37. Qa2#) (1. a4 $2 Bd2 (1... c4 $2 2. a5 bxa5 (2... Bf4 {#15}) 3. bxc4 Bh4 4. c5 Bxf2 5. c6 Bb6 6. Kd7 Kg7 7. c7 Bxc7 8. Kxc7 a4 9. Bxa4 Kg6 10. b6 Kf5 11. b7 Kf4 12. b8=Q Kg3 13. Qb2 h5 14. Qd4 Kxf3 15. Bd1+ Kg3 16. Bxh5 Kg2 17. Qh4 Kg1 18. Qg3+ Kf1 19. Qh2 Ke1 20. Qe2#) 2. f4 Bb4 (2... Bxf4 $2 3. a5 bxa5 4. b6 h5 5. Be4 Kg7 6. Ke7 h4 7. Bg2 Bh2 8. Ke6 Kf8 9. Bh3 Ke8 10. Kd5 Ke7 11. Bg4 Kd8 12. f3 Bg1 13. Kc4 Bf2 14. f4 Bd4 15. Bh3 Be3 16. f5 $18) 3. Ke7 (3. Kd7 Kf7 4. Kc6 Ba5 5. Bd1 Kf6 6. Bg4 h5 7. Bh3 Ke7 $11) (3. f5 Kg7 4. Ke7 c4+ 5. Ke6 cxb3 6. f6+ Kf8 7. Bxb3 h5 8. Kf5 Bc3 9. f7 h4 10. Kf4 Be1 11. Kf3 Ke7 12. Bc4 Kf8 13. Kg2 Ke7 14. f3 Kf8 15. Kh3 Ke7 16. Ba2 Kf8 17. f4 Ke7 18. Bc4 Kf8 19. Be6 Ke7 20. Kg4 Bb4 21. Bc4 Be1 22. Bb3 Kf8 23. Bd5 Ke7 24. Ba2 Kf8 25. Bb3 Ke7 $11) 3... c4+ 4. Kf6 Bc3+ 5. Kg6 cxb3 6. Bxb3+ Kf8 7. Kxh6 Bd2 8. Kg5 Ke7 9. Bc4 Kd6 10. Kf5 Kc5 11. Be2 Kd6 12. Bh5 Ke7 13. Ke5 Bc3+ 14. Kd5 Kf6 15. Bg4 Be1 16. f3 Bd2 17. Bc8 Ke7 18. Kc6 Ba5 19. f5 Kf7 20. Be6+ Ke7 21. f4 Kf6 22. Kd7 Be1 23. Kc6 Ba5 24. Kd7 Be1 25. Kc7 Ba5 $11) (1. Bf5 $2 h5) 1... Bxf4 (1... Bf6 2. Kd7 (2. a4 $6 Bc3 3. f5 Kg7 4. Ke7 Bf6+ 5. Ke6 Bc3 6. Bd1 Kf8 7. f6 Be1 8. Bh5 Bc3 9. Bg6 Bb4 10. Kd7 Ba5 11. Bh5 Be1 12. Kc6 Ba5 13. Kc7 Kg8 14. f7+ Kf8 15. f4 Ke7 16. Kc6 Kf8 17. f5 Kg7 18. f6+ Kf8 19. Kc7 Bc3 20. Kxb6 Bd4 21. Kc6 Bxf6 22. Kxc5 Be5 23. b6 Ke7 24. Kc6 Bd4 25. a5 Bg7 26. f8=Q+ Kxf8 27. a6 Ke7 28. a7 Kf6 29. a8=Q Kg5 30. Qg8 Kf4 31. Qxg7 Ke3 32. Kd5 Kf2 33. Kd4 Kf1 34. Ke3 Ke1 35. Qg1#) 2... Bc3 (2... Bd4 $5 3. Kc6 Bxf2 (3... h5 4. a4 Bxf2 5. Kxb6 c4+ {ZU}) 4. a4 h5 5. Kxb6 c4+ 6. Kc6 (6. Ka6 $6 c3 7. b6 Bg3 8. a5 h4 9. b7 Bxf4 10. Kb5 h3 11. a6 h2 12. a7 h1=Q 13. b8=Q+ Bxb8 14. axb8=Q+ Kg7 15. Qc7+ Kg8 16. Qc4+ Kf8 17. Qc5+ Kf7 18. b4 Qb7+ 19. Kc4 Qa6+ 20. b5 Qa2+ 21. Kxc3 Qa1+ 22. Kb3 Qa5 23. Qc6 Qd2 24. Qc4+ Ke8 25. Bg6+ Kd8 26. Qd3+ Qxd3+ 27. Bxd3 Kd7 28. Kc4 Kd6 29. Bc2 Ke5 30. b6 Kd6 31. Kb5 Ke7 32. b7 Kf6 33. b8=Q Kg5 34. Qe5+ Kh4 35. Qf4+ Kh3 36. Bf5+ Kg2 37. Be4+ Kg1 38. Qg3+ Kf1 39. Bd3#) 6... c3 7. Kd5 h4 8. Kc4 h3 (8... Kf7 $5 9. Kxc3 h3 10. Kc4 h2 11. Be4 Bb6 12. b4 Ke6 13. a5 Bf2 14. b6 Kd7 15. Kb5 Bg3 16. a6 Bxf4 17. a7 h1=Q 18. Bxh1 Ke7 19. Be4 Bd6 20. b7 Bc7 21. b8=Q Bxb8 22. axb8=Q Kf6 23. Qf8+ Kg5 24. Qf5+ Kh4 25. Qg6 Kh3 26. Bf3 Kh4 27. Qg4#) 9. Kxc3 Kg7 10. Kc4 Kh6 11. a5 Kh5 12. Be4 Kg4 13. b6 Kxf4 14. Bd5 Ke5 15. b7 Ba7 16. b4 Kd6 17. b5 Kc7 18. b6+ Bxb6 19. axb6+ Kb8 20. Bh1 h2 21. Kd5 Kxb7 22. Kc5+ Kc8 23. Kd6 Kd8 24. b7 Ke8 25. Bd5 Kf8 26. b8=Q+ Kg7 27. Qg8+ Kf6 28. Be4 h1=R 29. Qg6#) 3. Kc6 Ba5 4. b4 cxb4 5. Bb3+ Kg7 6. Kd5 Kf6 (6... h5 $5 7. f5 h4 8. Ke4 Kf6 9. Kf4 h3 10. Kg3 Kxf5 11. f3 Ke5 12. Kxh3 Kf4 13. Kg2 Ke5 14. Kg3 Kd4 15. f4 Kc5 16. f5 Kxb5 17. f6 Kc6 18. f7 b5 19. f8=Q Bc7+ 20. Kg4 Bh2 21. Qf6+ (21. Qe8+ $5 Kb6 22. Qg6+ (22. Qe6+ $5 Kc5 23. Qf5+ Kd6 24. Qd5+ (24. Qf2 $5 Be5 25. Qb6+ Ke7 26. Kf5 Bh2 27. Qc5+ Bd6 28. Qa7+ Kd8 29. Ke6 Bc7 30. Qa8+ Bb8 31. Qxb8#) 24... Ke7 25. Kf5 Bc7 26. Qc5+ Kd8 27. Be6 Be5 28. Qc8+ Ke7 29. Qd7+ Kf8 30. Qf7#) 22... Ka5 23. Qc6 Be5 24. Qb7 Bd4 25. Kf5 Bc5 26. Ke6 Bf2 27. Kd5 Bc5 28. Kc6 Bb6 29. Qxb6#) 21... Kc7 22. Qa6 Kd7 23. Kf5 Bg3 24. Be6+ Kc7 25. Bd5 Kd7 26. Qc6+ Ke7 27. Be6 Kf8 28. Qd7 b3 29. Qf7#) 7. Ke4 h5 8. f5 h4 9. Kf4 h3 10. Kg3 Kxf5 11. f3 Ke5 12. Kxh3 Kf4 13. Kg2 Ke3 14. Kg3 Kd4 15. f4 Kc5 16. f5 Kxb5 17. f6 Kc6 18. f7 b5 19. f8=Q Bc7+ 20. Kg4 Bd6 21. Qa8+ Kb6 22. Bd5 b3 23. Qc6+ Ka5 24. Qxd6 bxa2 25. Bxa2 Ka4 26. Qc5 b4 27. Bd5 Ka3 28. Qc2 b3 29. Qxb3#) 2. Kd7 Bd2 3. Kc6 Ba5 4. b4 cxb4 (4... Bxb4 $6 5. Kxb6 Kf7 6. a4 Ke6 7. a5 Kd5 8. a6 c4 9. a7 Bc5+ 10. Ka6 Bxa7 11. Kxa7 Kd4 12. b6 Kc3 13. Bg6 h5 14. b7 h4 15. b8=Q h3 16. Qh8+ Kb3 17. Qxh3+ Ka2 18. Bf7 Kb2 19. Bxc4 Kb1 20. Qd3+ Kc1 21. Qc3+ Kb1 22. Qd2 Ka1 23. Qc1#) 5. Bb3+ Kg7 6. Kd5 Kf6 7. Ke4 h5 (7... Kg5 $6 8. Ke5 h5 9. f4+ Kh6 10. f5 Kg7 11. f6+ Kf8 12. Kf5 h4 13. Kg4 h3 14. Kxh3 Ke8 15. Kg4 Kd8 16. f7 Ke7 17. Kf5 Kf8 18. Ke6 Kg7 19. Ke7 Kg6 20. f8=Q Kg5 21. Qf3 Kg6 22. Qf6+ Kh7 23. Bc2+ Kg8 24. Qf8#) 8. Kf4 Kg6 (8... h4 $5 9. f3 h3 10. Kg3 Ke5 11. Kxh3 Kf4 12. Kg2 Ke3 13. Kg3 Kd4 14. f4 Kc5 15. f5 Kxb5 16. f6 Kc6 17. f7 b5 18. Kg4 Bd8 19. f8=Q Bc7 20. Qa8+ Kc5 21. Qb7 Bh2 22. Qa7+ Kc6 23. Kf5 Kd6 24. Qb6+ Ke7 25. Be6 Bd6 26. Qa7+ Bc7 27. Qxc7+ Ke8 28. Qd7+ Kf8 29. Qf7#) 9. Ke5 h4 10. Kf4 Kh5 11. f3 $1 h3 12. Kg3 Kg5 13. Kxh3 Kf4 14. Kg2 Ke3 15. Kg3 Kd4 16. f4 Kc5 17. f5 Kxb5 18. f6 Kc6 19. f7 b5 20. f8=Q Bc7+ 21. Kg4 Bh2 22. Qe8+ Kb6 23. Qg6+ (23. Qe6+ $5 Kc5 24. Qf5+ Kc6 25. Qc2+ Kb6 26. Qxh2 Kb7 27. Qd6 Kc8 28. Qc6+ Kd8 29. Qb7 Ke8 30. Qc7 Kf8 31. Qf7#) 23... Ka5 24. Qc6 Bg1 25. Kf3 Ba7 26. Ke4 Bf2 27. Kd3 Be1 28. Bc4 bxc4+ 29. Kxc4 b3 30. Qb5# 1-0
[/pgn]

[Jouni]

In my 6 testsuites (about 700 positions) Leptir is better than Rems in 5. Note, that this even with default setting mv=3! Also Leptir has 10-20% faster nps.

[peter]

With the 1024 I once in a while described together with download- link here

https://forum.computerschach.de/cgi-bin ... #pid176682

, it looks like this, as for direct comparison between Rems and Leptir (T4.3 means 1 thread (out of 16) = 4.3GHz, in this one list I have some older runs with 3.5GHz- CPU still too, 15" is TC single thread, cc means concurrencies at single run, percentage of MEA- points comes from that relative to max. of 13810 in this one suite, error bar is about 1%), both engines ran with their default settings:

Code: Select all

Letpir251203T4.3    15" 1t 16cc : 10363=75.03% Found 866
Rems091224T4.3      15" 1t 16cc :  9880=71.54% Found 837

So here the higher nps- numbers probably help more than they do in EloStatTS, where the differences in time to solution are compared not only engine- but position- and solution- wise and the higher TC for the (on average) more "difficult" (as for hardware- time necessary till best move stays stable in output) 256 (compared to the 1024) gives a more setting- dependent result, as you will have seen on your own too, trying especially different Random Op. values, regards