Author: Heiner Marxen
Date: 04:26:29 07/26/01
Go up one level in this thread
On July 26, 2001 at 02:42:26, Uri Blass wrote:
>On July 25, 2001 at 21:22:05, Uri Blass wrote:
>
>>On July 25, 2001 at 16:59:56, Heiner Marxen wrote:
>>
>>>On July 24, 2001 at 13:25:39, Uri Blass wrote:
>>>
>>>>On July 24, 2001 at 09:48:29, Ian Kennedy wrote:
>>>>
>>>>>In a recent test game against my computer (which was Black), the following
>>>>>postion arose:
>>>>>
>>>>>[D] 4Nr2/2QR1pkp/4p1p1/8/1p6/r3q1NP/6PK/8 b - - 0 0
>>>>>
>>>>>I'd gone for a cheapo draw attempt hoping to fool it into playing Rxe8. No such
>>>>>luck, it played Kh8 and went on to win.
>>>>>
>>>>>When I went back to the position after 31. ... Rxe8 to see if there was a draw
>>>>>to be had, I was surprised to find its a very forced win for White. But neither
>>>>>Crafty or Fritz can see the mate all in one go, even though this line I worked
>>>>>out in Fritz has 15 checks out of 20 moves before the final mate:
>>>>>
>>>>>(31... Rxe8?? 32. Rxf7+ Kh6 33. Rxh7+ Kg5 34. h4+ Kg4
>>>>>35. Qc4+ Qf4 36. Qe2+ Rf3 37. gxf3+ Qxf3 38. Qe1 Qf4 39. Qd1+ Qf3 40. Qg1 Qf4
>>>>>41. Qg2 Qf3 42. Qh3+ Kf4 43. Rf7+ Ke5 44. Rxf3 Kd6 45. Rd3+ Kc6 46. Qg2+ Kc7
>>>>>47. Qc2+ Kb8 48. Qc5 Rf8 49. Qxb4+ Kc7 50. Qc5+ Kb7 51. Rd7+ Ka6 52. Ra7# {
>>>>>(Fritz 5)})
>>>>>
>>>>>Yet Crafty on my PC is only showing +3.65 after an hour or so. Maybe there is
>>>>>some inaccuracy in the above variation?
IMO it is a bit too deep for todays hardware/programs.
>>>>The first move for black that is not obviously forced is 38...Qf4
>>>
>>>Yes, I agree.
>>>
>>>>Here are some non trivial mate to find before you can claim mate in 21
>>>>
>>>>1)After 38...e5
>>>>2)After 38...g5
>>>>3)After 38...Rc8
>>>>4)After 40...Rd8
>>>>5)After 40...Re7
>>>>6)After 44...Re7
>>>>7)After 45...Ke7
>>>>
>>>>I do not know if there is a forced mate in 21 but you have to work in order to
>>>>prove it by a tree and check every legal move of black in every position in
>>>>order to see if the computer find a forced mate after it.
>>>>
>>>>Uri
>>>
>>>I did investigate this position a bit more. I started with crafty 18.10
>>>before 38.Qe1. I did forget to enable EGTBs, and are not sure how much
>>>difference that would have made (I have only the 4-piece EGTBs). Anyhow...
>>>
>>>Crafty v18.10
>>>
>>>White(1): ponder off
>>>pondering disabled.
>>>White(1): hash 192M
>>>hash table memory = 192M bytes.
>>>White(1): setboard 4r3/7R/4p1p1/8/1p4kP/5qN1/4Q2K/8 w - -
>>>
>>> +---+---+---+---+---+---+---+---+
>>> 8 | | | | | *R| | | |
>>> +---+---+---+---+---+---+---+---+
>>> 7 | | | | | | | | R |
>>> +---+---+---+---+---+---+---+---+
>>> 6 | | | | | *P| | *P| |
>>> +---+---+---+---+---+---+---+---+
>>> 5 | | | | | | | | |
>>> +---+---+---+---+---+---+---+---+
>>> 4 | | *P| | | | | *K| P |
>>> +---+---+---+---+---+---+---+---+
>>> 3 | | | | | | *Q| N | |
>>> +---+---+---+---+---+---+---+---+
>>> 2 | | | | | Q | | | K |
>>> +---+---+---+---+---+---+---+---+
>>> 1 | | | | | | | | |
>>> +---+---+---+---+---+---+---+---+
>>> a b c d e f g h
>>>
>>>White(1): st 600
>>>search time set to 600.00.
>>>White(1): go
>>> clearing hash tables
>>> time surplus 0.00 time limit 10:00 (10:00)
>>> nss depth time score variation (1)
>>> 7 1.03 ++ 1. Qe1!!
>>> 7-> 1.47 2.63 1. Qe1 Qf4 2. Rb7 e5 3. Rxb4 e4
>>> 8 2.30 2.73 1. Qe1 Qf4 2. h5 g5 3. Rb7 e5 4. h6
>>> Re6
>>> 8-> 2.76 2.73 1. Qe1 Qf4 2. h5 g5 3. Rb7 e5 4. h6
>>> Re6
>>> 9 5.66 ++ 1. Qe1!!
>>> 9-> 8.80 3.12 1. Qe1 Qf4 2. h5 g5 3. Rb7 e5 4. h6
>>> Re6
>>> 10 12.88 3.32 1. Qe1 Qf4 2. h5 Kg5 3. Rg7 Kh4 4.
>>> Rxg6 b3 5. h6 e5
>>> 10-> 15.48 3.32 1. Qe1 Qf4 2. h5 Kg5 3. Rg7 Kh4 4.
>>> Rxg6 b3 5. h6 e5
>>> 11 26.03 ++ 1. Qe1!!
>>> 11 1:30 4.16 1. Qe1 Qf4 2. h5 Kg5 3. Kg2 gxh5 4.
>>> Rg7+ Kh6 5. Nxh5 Re7 6. Nxf4 Rxg7+
>>> 7. Kf3
>>> 11-> 1:37 4.16 1. Qe1 Qf4 2. h5 Kg5 3. Kg2 gxh5 4.
>>> Rg7+ Kh6 5. Nxh5 Re7 6. Nxf4 Rxg7+
>>> 7. Kf3
>>> 12 1:59 4.30 1. Qe1 Qf4 2. h5 Kg5 3. Kg2 Rf8 4.
>>> Qxe6 Qf3+ 5. Kh2 Qd3 6. Qe7+ Kg4 7.
>>> Qxf8 Qxg3+ 8. Kh1 Qe1+ 9. Kh2 gxh5
>>> 12-> 2:20 4.30 1. Qe1 Qf4 2. h5 Kg5 3. Kg2 Rf8 4.
>>> Qxe6 Qf3+ 5. Kh2 Qd3 6. Qe7+ Kg4 7.
>>> Qxf8 Qxg3+ 8. Kh1 Qe1+ 9. Kh2 gxh5
>>> 13 3:49 ++ 1. Qe1!!
>>> 13 6:07 4.70 1. Qe1 Qf4 2. h5 Kg5 3. Kg2 Rf8 4.
>>> Qxe6 Qf3+ 5. Kh2 b3 6. hxg6 Rf6 7.
>>> Ne4+ Kxg6 8. Nxf6 Qxf6
>>> 13-> 7:05 4.70 1. Qe1 Qf4 2. h5 Kg5 3. Kg2 Rf8 4.
>>> Qxe6 Qf3+ 5. Kh2 b3 6. hxg6 Rf6 7.
>>> Ne4+ Kxg6 8. Nxf6 Qxf6
>>> time=10:00 cpu=98% mat=1 n=183310432 fh=93% nps=305k
>>> ext-> chk=20376026 cap=407860 pp=527915 1rep=1291452 mate=29595
>>> predicted=0 nodes=183310432 evals=38692257
>>> endgame tablebase-> probes done=0 successful=0
>>>
>>>Obviously, crafty also likes Qe1 and thinks white is winning. But its PV
>>>deviates alredy in the second white move from your analysis.
>>>So I gave it more time and let it run over night:
>>>
>>>Black(1): st 43200
>>>search time set to 43200.00.
>>>Black(1): setboard 4r3/7R/4p1p1/8/1p4kP/5qN1/4Q2K/8 w - -
>>>
>>> +---+---+---+---+---+---+---+---+
>>> 8 | | | | | *R| | | |
>>> +---+---+---+---+---+---+---+---+
>>> 7 | | | | | | | | R |
>>> +---+---+---+---+---+---+---+---+
>>> 6 | | | | | *P| | *P| |
>>> +---+---+---+---+---+---+---+---+
>>> 5 | | | | | | | | |
>>> +---+---+---+---+---+---+---+---+
>>> 4 | | *P| | | | | *K| P |
>>> +---+---+---+---+---+---+---+---+
>>> 3 | | | | | | *Q| N | |
>>> +---+---+---+---+---+---+---+---+
>>> 2 | | | | | Q | | | K |
>>> +---+---+---+---+---+---+---+---+
>>> 1 | | | | | | | | |
>>> +---+---+---+---+---+---+---+---+
>>> a b c d e f g h
>>>
>>>White(1): go
>>>crafty is behind 1.5 on time, reducing by 1/3.
>>> time surplus 0.00 time limit 720:00 (720:00)
>>> nss depth time score variation (1)
>>> 10-> 0.40 4.70 1. Qe1 Qf4 2. h5 <HT>
>>> 11 0.40 4.70 1. Qe1 Qf4 2. h5 <HT>
>>> 11-> 1.29 4.70 1. Qe1 Qf4 2. h5 <HT>
>>> 12 1.29 4.70 1. Qe1 Qf4 2. h5 <HT>
>>> 12-> 6.15 4.70 1. Qe1 Qf4 2. h5 <HT>
>>> 13 6.16 4.70 1. Qe1 Qf4 2. h5 <HT>
>>> 13-> 32.96 4.70 1. Qe1 Qf4 2. h5 <HT>
>>> 14 28:28 ++ 1. Qe1!!
>>> 14 178:14 7.14 1. Qe1 Qf4 2. Qd1+ Qf3 3. Qg1 Qf4 4.
>>> Qg2 e5 5. Qh3+ Kf3 6. Qf1+ Ke3 7. Qe2+
>>> Kd4 8. Rd7+ Kc5 9. Qc2+ Kb6 10. Qxg6+
>>> Ka5 11. Rd5+ Ka4 12. Qxe8+ Ka3 13.
>>> Qa8+ Kb2 14. h5
>>> 14-> 183:31 7.14 1. Qe1 Qf4 2. Qd1+ Qf3 3. Qg1 Qf4 4.
>>> Qg2 e5 5. Qh3+ Kf3 6. Qf1+ Ke3 7. Qe2+
>>> Kd4 8. Rd7+ Kc5 9. Qc2+ Kb6 10. Qxg6+
>>> Ka5 11. Rd5+ Ka4 12. Qxe8+ Ka3 13.
>>> Qa8+ Kb2 14. h5
>>> 15 203:23 ++ 1. Qe1!!
>>> 15 370:43 9.25 1. Qe1 Qf4 2. Qd1+ Qf3 3. Qg1 Qf4 4.
>>> Qg2 b3 5. Rg7 Kxh4 6. Rxg6 Qg4 7. Rh6+
>>> Kg5 8. Qd2+ Qf4 9. Rh5+ Kg6 10. Qxf4
>>> Re7 11. Qe5
>>> 15-> 381:24 9.25 1. Qe1 Qf4 2. Qd1+ Qf3 3. Qg1 Qf4 4.
>>> Qg2 b3 5. Rg7 Kxh4 6. Rxg6 Qg4 7. Rh6+
>>> Kg5 8. Qd2+ Qf4 9. Rh5+ Kg6 10. Qxf4
>>> Re7 11. Qe5
>>> 16 443:28 ++ 1. Qe1!!
>>> time=720:00 cpu=98% mat=1 n=1909505130 fh=0% nps=44k
>>> ext-> chk=-2060764322 cap=13731413 pp=33182471 1rep=154509366
>>>mate=1272643
>>> predicted=0 nodes=1909505130 evals=837289426
>>> endgame tablebase-> probes done=0 successful=0
>>>
>>>(some counters did wrap multiply)
>>>
>>>Crafty replaces your 41...Qf3 with 4...b3.
>>>After crafty's move we have [D]4r3/7R/4p1p1/8/5qkP/1p4N1/6QK/8 w - -
>>>and Chest finds a mate in 10:
>>>
>>>PV: Rg7 Qf3 Rxg6+ Kf4 Qb2 Qxg3+ Rxg3 Rg8 Qd2+ Ke4 Rxg8 e5 Rg3 Kf5 Qd6 b2 Qg6+
>>>Kf4 Qg4#
>>>(37 minutes on a K7/600 with 350 MB hash)
>>>
>>>If black plays Kxh4 instead of Qf3 after the initial Rg7, the mate is one
>>>move shorter:
>>>[D]4r3/6R1/4p1p1/8/5q1k/1p4N1/6QK/8 w - -
>>>PV: Rxg6 Qg4 Qf2 Rf8 Rh6+ Kg5 Qxf8 Qf3 Ne4+ Kg4 Rg6+ Kh4 Qe7+ Qf6 Qh7#
>>>(32 seconds)
>>>
>>>I haven't analysed 41...Qf3, sorry.
>>>
>>>Regards,
>>>Heiner
>>
>>Chessmaster6000 says only mate in 12 after 41...Qf3.
>>
>>The main line begins with 42.Qh3+ Kf4 43.Rf7+ Ke5 44.Rxf3 Rd8 45.Re3+
>>
>>Uri
>
>I believe that after 41...Qf3+
>42.Qh3+ is mate in 11 because Chessmaster can see mate in 9 after
>42.Qh3+ Kf4 43.Rf7+ Ke5 if I give it some minutes.
>
>Uri
Yes, Chest agrees to this: after 41...Qf3+
[D]4r3/7R/4p1p1/8/1p4kP/5qN1/6QK/8 w - -
Chest finds a mate in 11:
PV: Qh3+ Kf4 Rf7+ Ke3 Nf1+ Kd4 Qxf3 Rh8 Rd7+ Kc5 Rc7+ Kd6 Qc6+ Ke5 Qc5+ Ke4 Ng3+
Kf4 Rf7+ Kg4 Qd4#
(85.4 minutes)
CM's main line from above also is part of the solution tree. After
three more full moves (after 44...Rd8) Chest finds the following mate in 8:
PV: Re3+ Kd5 Rd3+ Kc5 Ne4+ Kb5 Rxd8 g5 Qd3+ Kb6 Rb8+ Ka7 Rxb4 g4 Qa3#
(5.6 minutes)
After Crafty's 41...b3 : mate in 10
After Fritz's 41...Qf3 : mate in 11
In this sense, Fritz's defense is a bit better. The depth is correct,
although the main line is not completely correct: on 44...Kd6 the mate
with 45.Ne4+ is one shorter, but 44...Kd4 45.Rf4+ restores the old length
(as do 44...Ra8/Rc8/Rd8/Rh8 45.Re3+).
This is still no complete proof, but convincing support for Ian's original
statement, that white has a forced win.
Regards,
Heiner
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.