Computer Chess Club Archives


Search

Terms

Messages

Subject: example of bad hashtable usage deep blue II

Author: Vincent Diepeveen

Date: 08:15:17 10/14/02

Go up one level in this thread


On October 12, 2002 at 22:42:28, Robert Hyatt wrote:

>On October 12, 2002 at 10:33:29, Omid David wrote:
>
>>On October 12, 2002 at 09:53:42, Robert Hyatt wrote:
>>
>>>On October 12, 2002 at 08:29:24, Omid David wrote:
>>>
>>>>On October 11, 2002 at 21:57:34, Robert Hyatt wrote:
>>>>
>>>>>On October 11, 2002 at 17:51:27, Omid David wrote:
>>>>>
>>>>>>On October 11, 2002 at 11:51:13, Robert Hyatt wrote:
>>>>>>
>>>>>>>On October 11, 2002 at 07:43:40, Uri Blass wrote:
>>>>>>>
>>>>>>>>On October 11, 2002 at 07:12:34, Vincent Diepeveen wrote:
>>>>>>>>
>>>>>>>>>On October 11, 2002 at 04:08:49, Uri Blass wrote:
>>>>>>>>>
>>>>>>>>>Isn't his article clear enough yet?
>>>>>>>>
>>>>>>>>Bob Hyatt still claims that it was 12 plies software and 6 plies hardware
>>>>>>>>so I prefer to hear an answer directly from Hsu.
>>>>>>>
>>>>>>>I agree.  But _I_ don't claiam _anything_ except that members of the DB team
>>>>>>>specifically told me that 12(6) means 12 plies in hardware, 6 in software.  I
>>>>>>>even posted the excerpt from the email that specifically said this...
>>>>>>>
>>>>>>>That is _all_ I have said about it...
>>>>>>>
>>>>>>
>>>>>>No matter what they say, even under extreme theoretical conditions it is
>>>>>>*impossible* to search 18 plies of brute force in chess, without any type of
>>>>>>forward pruning whatsoever, and no hash tables.
>>>>>
>>>>>
>>>>>However, they have _never_ said they didn't use forward pruning.  They have only
>>>>>said "we don't use null-move for forward pruning" and nothing else.  And they
>>>>>have
>>>>>slowly leaked details.  But they pretty much had to since I had gone over their
>>>>>log
>>>>>files and discovered that theyt had a _very_ good branching factor, too good for
>>>>>pure
>>>>>alpha/beta alone...
>>>>>
>>>>
>>>>Interesting... What was the average branching factor based on the logs you
>>>>reviewed?
>>>>
>>>>
>>>
>>>
>>>I don't remember the _exact_ number although I posted it here in CCC and several
>>>were involved
>>>in a long discussion about it.. but the number was something less than 4.0 I
>>>believe, which is
>>>_way_ below what a pure alpha/beta program can produce.
>>>
>>>
>>
>>I can't imagine a way for brute force alpha-beta to come up with a branching
>>factor of anything even close to that number (esp. without hash tables). With
>>regard to the branching factor, it seems that some kind of forward pruning was
>>indeed in place...
>>
>
>Remember, Deep Blue _did_ have hash tables.  Only the last few plies (done in
>hardware)
>didn't have hashing.  The first N plies hashed just like the rest of us...

No their thing was too inefficient to make use of hashtables. You
can see in the logfiles that in positions with loads of transpositions
where all software programs of today search up to 5 ply deeper than
in the other positions, that deep blue doesn't search a ply
deeper there.

Let me give 2 examples with diep which also uses singular extensions,
but WITH hashtables and nullmove:

---------------------------------------
--> Bf8e7r <--
---------------------------------------
 123  T=7
qd3c3
 3(4) 95^ T=1
pc2c4 Pb5c4p qd3c4P Nd5b4 ra1e1 Kc8d8 re1e7B Kd8e7r
 3(5) 103  T=1
qd3f5P Kc8b7 qf5d7N
 4(5) 103  T=1
qd3f5P Kc8b7 qf5d7N
 5(5)[Qxf5](97)[c4](133) 133^ T=1
pc2c4 Pb5c4p qd3c4P Nd5b4 ra1e1 Kc8d8 re1e7B Kd8e7r
 5(5)[c4](159) 159  T=1
pc2c4 Nd5f6 pd4d5 Pb5c4p qd3c4P Nf6d5p
 6(5)[c4](138) 138  T=1
pc2c4 Pb5c4p qd3c4P Nd5b4 ra1e1 Kc8d8 re1e7B Kd8e7r
 7(5) #[c4](168) 168^ T=2
pc2c4 Pb5c4p qd3c4P Nd5b4 ra1e1 Kc8d8 re1e7B Kd8e7r
 7(6) #[c4](169)######################################### 169  T=2
pc2c4 Nd5b6 qd3f5P Nb6c4p qf5e6 Bc6f3n
 8(6) #[c4](182)######################################### 182  T=5
pc2c4 Pb5c4p qd3c4P Nd5b4 ra1e1 Rh8e8 re1e7B Re8e7r qc4b4N
 9(6) #[c4](189)######################################### 189  T=18
pc2c4 Pb5c4p qd3c4P Nd5b4 ra1e1 Rh8e8 nf3h4 Pf5f4 nh4f5 Nd7b6 qc4f7 Nb4d5
10(6) #[c4](219) 219^ T=21
pc2c4 Pb5c4p qd3c4P Nd5b4 ra1e1 Rh8e8 nf3h4 Pf5f4 nh4f5 Nd7b6 qc4f7 Pf4g3b
nf5e7B Re8e7n re1e7R
10(6) #[c4](220)######################################### 220  T=43
pc2c4 Pb5c4p qd3c4P Nd5b4 ra1e1 Rh8e8 nf3h4 Pf5f4 nh4f5 Nd7b6 qc4f7 Re8f8 qf7e6
Bc6d7 nf5e7B
11(6) #[c4](250) 250^ T=56
pc2c4 Pb5c4p qd3c4P Nd5b4 ra1e1 Rh8e8 nf3h4 Pf5f4 nh4f5 Nd7b6 qc4f7 Re8f8 qf7e6
Bc6d7 nf5e7B Kc8c7 bg3f4P
 #[c4](267)####################[TIMEOUT] 267  T=181
pc2c4 Pb5c4p qd3c4P Nd5b4 ra1e1 Rh8e8 nf3h4 Nd7b6 qc4f7 Nb4d5 nh4f5P Kc8d8
nf5g7P Re8h8
---------------------------------------
-->  19.   c4 <-- 21/84:24
---------------------------------------

Game 6 move 19.c4 with DIEP:

black timeleft=27:46.40.00
 r = k = - = - r   Bc1-f4   11    b7-b5
 o - = n b - o -   a2-a4    12    Bc8-b7
 - = b = - = - o   Rf1-e1   13    Nf6-d5
 = o = n = o = -   Bf4-g3   14    Kd8-c8
 - = - O - = - =   a4xb5    15    c6xb5
 = - = Q = N B -   Qd1-d3   16    Bb7-c6
 - O O = - O O O   Bg6-f5   17    e6xf5
 R - = - = - K -   Re1xe7   18    Bf8xe7
white timeleft=27:46.40.00
white to move               type 'help' for command overview

anal
Analysis mode is on
Engineflags=0 mask=1 denktime=10000000 maxtime=10000000
00:00     60   0k 0 0 3 (2) 1 (0,0) 1.372 Qd3xf5
00:00   4050   4k 0 0 324 (2) 2 (0,5) 0.123 Qd3xf5 Kc8-b7
++ f3-e5
00:00   4187   4k 0 0 670 (2) 2 (0,11) 0.397 Nf3-e5 Nd7xe5 d4xe5
00:00  10312  10k 0 0 2475 (2) 3 (0,49) 0.397 Nf3-e5 Nd7xe5 d4xe5
++ d3-f5
00:00   9596  10k 0 0 2687 (2) 3 (0,53) 0.798 Qd3xf5 Kc8-b7 Nf3-e5 Nd7xe5 Qf5xe5

00:00  15953  16k 0 0 6541 (2) 4 (0,96) 0.473 Qd3xf5 Kc8-b7 Nf3-e5 Rh8-f8
++ c2-c4
00:00  24198  24k 0 0 14761 (2) 4 (6,122) 0.939 c2-c4 b5xc4 Qd3xc4 Nd5-b4
00:00  36171  36k 0 0 26767 (2) 5 (9,215) 1.329 c2-c4 Nd5-b4 Qd3xf5 b5xc4 Nf3-e5

00:00  46407  46k 0 0 42231 (2) 6 (10,278) 1.107 c2-c4 Nd5-b4 Qd3xf5 b5xc4 Qf5-e
6 Bc6xf3 Qe6xe7
00:01  74017  74k 0 0 120648 (2) 7 (13,381) 1.469 c2-c4 Nd5-b4 Qd3xf5 b5xc4 Nf3-
e5 Bc6-b5 Qf5-e6
00:04  86738  87k 0 0 358231 (2) 8 (33,553) 1.864 c2-c4 Nd5-b4 Qd3xf5 b5xc4 Qf5-
e6 Kc8-d8 Ra1-e1 Rh8-e8 Qe6xc4
00:11 111547 112k 0 0 1273877 (2) 9 (180,1570) 1.896 c2-c4 Nd5-b6 c4-c5 Nb6-d5 Q
d3xf5 Kc8-b7 Nf3-e5 Rh8-f8 Qf5-e6 Nd7xe5 Qe6xe5
00:32 121518 122k 0 0 3926277 (2) 10 (628,3960) 2.495 c2-c4 Nd5-b4 Qd3xf5 Rh8-f8
 Qf5-e6 Be7-d8 d4-d5 Rf8-f6 Qe6-g4 Bc6-b7 c4xb5 Nb4xd5 Qg4xg7
01:29 139712 140k 0 0 12477689 (2) 11 (669,4387) 3.180 c2-c4 Nd5-b4 Qd3xf5 Rh8-f
8 Qf5-e6 Be7-d8 c4xb5 Bc6xb5 Bg3-e5 g7-g5 Ra1-c1 Kc8-b7 h2-h3
03:25 151333 151k 0 0 31068755 (2) 12 (778,5153) 3.269 c2-c4 Nd5-b4 Qd3xf5 Rh8-f
8 Qf5-e6 Be7-d8 d4-d5 Rf8-f6 Qe6-g4 Bc6-b7 Nf3-e5 Rf6-d6 Ne5-c6 Rd6xc6

ok it is 200 seconds i need. Let's not cry for 20 seconds. But diep would
not show within 3 minutes the 12 ply mainline here. It's still busy with
mainline 12 ply after 3 minutes.

Now let's try some endgame where deep blue didn't get deep either:

---------------------------------------
--> nf3h4 <--
---------------------------------------
 -9  T=152
Nb6c8
 3(4) 12  T=0
Rd7f7 rg6e6 Rf7d7 re6e8 Rd7d1 kb1c2 Nb6c4p pb3c4N Rd1e1
 4(5) 12  T=0
Rd7f7 rg6e6 Rf7d7 re6e8 Rd7d1 kb1c2 Nb6c4p pb3c4N Rd1e1
 5(5)[Rf7](5)[Nc8](11) 11  T=0
Nb6c8 bg2d5 Nc8d6 kb1a1 Nf2e4 pg3g4 Ph5g4p
 6(5)[Nc8](-9) -9  T=0
Nb6c8 bg2f3 Nf2g4 bf3g4N Ph5g4b nh4g2 Nc8d6 ng2e3 Nd6e4 ne3g4P Rd7e7 kb1c2
 7(5) #[Nc8](3)######################## 3  T=1
Nb6c8 bg2d5 Nc8d6 kb1a1 Nf2e4 pg3g4 Ph5g4p
 8(6) #[Nc8](-5)######################## -5  T=2
Nb6c8 bg2d5 Nc8d6 bd5e6 Rd7e7 nh4g2 Nf2g4 ng2f4
 9(6) #[Nc8](-6)######################## -6  T=10
Nb6c8 bg2f3 Nf2g4 bf3g4N Ph5g4b nh4g2 Nc8d6 ng2e3 Nd6e4 ne3g4P Rd7e7 kb1c2
10(6) #[Nc8](-9)######################## -9  T=29
Nb6c8 bg2f3 Nf2g4 bf3g4N Ph5g4b nh4g2 Nc8d6 ng2e3 Nd6e4 ne3g4P Rd7e7 kb1c2
11(6) #[Nc8](-6)######################## -6  T=110
Nb6c8 bg2f3 Nf2g4 bf3g4N Ph5g4b nh4f5 Nc8e7 nf5e7N Rd7e7n rg6g7P Re7g7r
12(6) #[Nc8](-13)###[TIMEOUT] -13  T=192
Nb6c8 bg2f3 Nf2g4 bf3g4N Ph5g4b nh4f5 Nc8e7 nf5e7N Rd7e7n kb1c2 Pb7b6 pa2a3
Pa7a6 kc2d2
---------------------------------------
-->  34. ..   Nc8 <-- 6/46:3
---------------------------------------

++ d7-f7
00:00   4692   5k 0 0 2346 (2) 3 (2,54) 0.207 Rd7-f7 Bg2-d5 Rf7-e7
00:00   6170   6k 0 0 3394 (2) 4 (4,67) -0.005 Rd7-f7 Kb1-c2 Rf7-e7 Bg2-d5
++ d7-e7
00:00   7338   7k 0 0 4623 (2) 4 (6,78) 0.130 Rd7-e7 Nh4-f5 Re7-e1 Kb1-c2
00:00  14497  14k 0 0 10438 (2) 5 (7,125) 0.128 Rd7-e7 Bg2-f3 Nf2-g4 Nh4-f5 Re7-
e1 Kb1-c2
++ d7-f7
00:00  14891  15k 0 0 12360 (2) 5 (7,134) 0.184 Rd7-f7 Bg2-f3 Nf2-g4 Bf3xg4 h5xg
4
00:00  21627  22k 0 0 20546 (2) 6 (9,194) -0.152 Rd7-f7 Bg2-f3 Nb6-d7 Bf3xh5 Nf2
-e4 g3-g4
++ f2-d3
00:01  29224  29k 0 0 33024 (2) 6 (15,237) -0.064 Nf2-d3 Bg2-e4 Nd3-f2 Be4-f3 Nf
2-g4 Nh4-f5
00:01  56289  56k 0 0 94003 (2) 7 (34,411) -0.646 Nf2-d3 Nh4-f5 h5-h4 Nf5xh4 Nd3
-e5 Rg6-e6 Ne5-f7
++ d7-f7
00:02  65390  65k 0 0 136013 (2) 7 (58,525) 0.036 Rd7-f7 Kb1-c2 Nf2-g4 Kc2-c3 Ng
4-e5 Rg6-e6 Rf7-f8
++ b6-c8
00:02  74062  74k 0 0 191082 (2) 7 (67,595) 0.091 Nb6-c8 Kb1-c2 Nf2-g4 Bg2-f3 Ng
4-e5 Rg6-e6 Ne5xf3 Nh4xf3
00:03  79064  79k 0 0 270401 (2) 8 (108,818) -0.058 Nb6-c8 Bg2-f3 Nf2-g4 Bf3xg4
h5xg4 Kb1-c2 Rd7-e7 Nh4-f5 Re7-e2 Kc2-d3 Re2xa2 Rg6xg7 Kc7-c6 Nf5-e7 Kc6-b6
++ d7-f7
00:04  76313  76k 0 0 328149 (2) 8 (134,935) 0.089 Rd7-f7 Bg2-f3 Nf2-g4 Rg6-e6 N
b6-d7 g5-g6 Rf7-f6 Re6-e7
00:06 101434 101k 0 0 651207 (2) 9 (201,1405) -0.065 Rd7-f7 Rg6-e6 Nf2-g4 Re6-e8
 Rf7-d7 Kb1-c2 Rd7-f7 Bg2-f3 Kc7-d6
++ b6-c8
00:07 117169 117k 0 0 910404 (2) 9 (253,1696) -0.058 Nb6-c8 Bg2-f3 Nf2-g4 Bf3xg4
 h5xg4 Kb1-c2 Rd7-e7 Nh4-f5 Re7-e2 Kc2-d3 Re2xa2 Rg6xg7 Kc7-c6 Nf5-e7 Kc6-b6
00:11 148037 148k 0 0 1686144 (2) 10 (326,2282) -0.185 Nb6-c8 Bg2-d5 Nc8-d6 Nh4-
g2 Nd6-f5 Ng2-f4 Rd7-d6 Rg6xd6 Kc7xd6 Nf4xh5 g7-g6
++ d7-f7
00:23 184024 184k 0 0 4322739 (2) 11 (574,4170) 0.038 Nb6-c8 Bg2-f3 Nf2-g4 Rg6-e
6 Rd7-e7 Bf3xg4 h5xg4 Re6xe7 Nc8xe7 Kb1-c1 Kc7-d6 Kc1-d2 Kd6-e5 Kd2-e3
00:34 198553 199k 0 0 6850093 (2) 12 (798,5722) -0.039 Nb6-c8 Bg2-f3 Nf2-g4 Rg6-
e6 Rd7-e7 Re6xe7 Nc8xe7 Kb1-c2 b7-b6 Bf3xg4 h5xg4 Kc2-d3 Ne7-c6 a2-a3
01:00 213416 213k 0 0 12935199 (2) 13 (1200,8564) -0.069 Nb6-c8 Bg2-f3 Nf2-g4 Rg
6-e6 Rd7-e7 Re6xe7 Nc8xe7 Kb1-c2 b7-b6 Bf3xg4 h5xg4 Kc2-d3 Ne7-c6 Nh4-f5 Nc6-b4
Kd3-e2
02:20 223894 224k 0 0 31544563 (2) 14 (1945,14239) -0.124 Nb6-c8 Bg2-f3 Nf2-g4 R
g6-e6 Rd7-e7 Re6xe7 Nc8xe7 Kb1-c2 b7-b6 Bf3xg4 h5xg4 Kc2-d3 Ne7-c6 Nh4-f5 Nc6-b4
 Kd3-e2
06:06 225933 226k 0 0 82858962 (2) 15 (3821,26259) -0.114 Nb6-c8 Bg2-f3 Nf2-g4 R
g6-e6 Rd7-e7 Re6xe7 Nc8xe7 Kb1-c2 a7-a6 Kc2-c3 b7-b5 Bf3xg4 h5xg4 Kc3-d3 Ne7-c6
Nh4-f5 Nc6-b4 Kd3-c3

After a lot of doubting diep finally gets to Nc8.
Note i miss the doubting
in deep blue completely always. It never doubts. Let's not
discuss how simple it's eval is otherwise we get distracted
from the measurement facts.

After 3 minutes deep blue is busy at 12 ply. DIEP is busy after 3 minutes
at 15 ply. So +3 ply there for diep.

There is nearly 3 ply difference in search depth for DIEP. There is
hardly 1 ply difference in deep blue.

Even worse is the game 2 positions when todays software does not
doubt:

---------------------------------------
--> Qc7b6 <--
---------------------------------------
 123  T=168
ra8d8 Qd7d8r qc6b7R Qd8c7 qb7b5P Kf7e7 pg2g3 Ke7d8 kg1g2 Qc7c8 qb5b6 Qc8c7 qb6a6
Kd8d7 qa6a8 Qc7c8
 3(4)[Kf1](30) 30^ T=2
kg1f1 Rb7b8 ra8a6 Qb6c6q pd5c6Q Kf7g8 be4d5 Kg8h7 ra6a7 Rb8c8 kf1e2 Bd6b4p
 3(5)[Kf1](61) 61^ T=3
kg1f1 Rb7b8 ra8a6 Qb6c6q pd5c6Q Kf7g8 be4d5 Kg8h7 ra6a7 Rb8c8 kf1e2 Bd6b4p
 3(5) 75  T=4
kg1h1 Rb7b8 qc6d7 Kf7g8 qd7e7 Bd6e7q
 4(5)[Kh1](80)[Kh2](84) 84  T=4
kg1h2 Rb7b8 ra8a6 Qb6c6q pd5c6Q Kf7f8 ra6a7
 5(5)[Kh2](81)[Kf1](86) 86  T=7
kg1f1 Rb7b8 ra8a6 Qb6c6q pd5c6Q Kf7g8 be4d5 Kg8h7 ra6a7 Rb8c8 kf1e2 Bd6b4p
 6(5) #[Kf1](113)#### 113  T=9
kg1f1 Rb7b8 ra8a6 Qb6c6q pd5c6Q Kf7g8 be4d5 Kg8h7 ra6a7 Rb8c8 kf1e2 Bd6b4p
 7(5) #[Kf1](96)#### 96  T=10
kg1f1 Rb7b8 ra8a6 Qb6c6q pd5c6Q Kf7g8 be4d5 Kg8h7 ra6a7 Rb8c8 kf1e2 Bd6b4p
 8(6) #[Kf1](126)[Kf1](126) 126^ T=14
kg1f1 Rb7b8 ra8a6 Qb6c6q pd5c6Q Kf7g8 be4d5 Kg8h7 ra6a7 Rb8d8 bd5e6 Bd6b8 pc6c7
Bb8a7r pc7d8R/q
 #[Kf1](116)#### 116  T=25
kg1f1 Rb7b8 ra8a6 Qb6c6q pd5c6Q Kf7g8 be4d5 Kg8h7 ra6a7 Rb8c8 kf1e2 Pe5e4 bd5e4P
 9(6) #[Kf1](116)#### 116  T=47
kg1f1 Rb7b8 ra8a6 Qb6c6q pd5c6Q Kf7g8 be4d5 Kg8h7 ra6a7 Rb8c8 kf1e2 Pe5e4 bd5e4P
10(6) #[Kf1](146)[Kf1](146) 146^ T=59
kg1f1 Rb7b8 ra8a6 Qb6c6q pd5c6Q Kf7g8 be4d5 Kg8h7 ra6a7 Rb8c8 ra7b7 Bd6b4p
pc3b4B Rc8c6p bd5c6R
 #[Kf1](156)#### 156  T=143
kg1f1 Rb7b8 ra8a6 Qb6c6q pd5c6Q Kf7g8 be4d5 Kg8h7 ra6a7 Rb8c8 ra7b7 Ph6h5 rb7b5P
Kh7h6 rb5b7 Rc8c7 rb7b8
11(6)[TIMEOUT] 156  T=192
kg1f1
---------------------------------------
-->  44.   Kf1 <-- 16/78:52
---------------------------------------

Deep Blue plays Kf1? here busy at 11 ply mainline it gets timeout.
Score for Kf1?, which is a move that blows the position to a draw.

00:03  88864  89k 0 0 271926 (2) 9 (28,312) 1.003 Kg1-f1 Rb7-b8 Qc6-d7 Kf7-g8 Ra
8-a7 Bd6-f8 Qd7-f7 Kg8-h7 d5-d6 Qb6-e3 Qf7-c7 Qe3xe4 Qc7xb8 Qe4-d3 Kf1-f2 Bf8xd6
 Qb8xb5 Qd3xc3
++ g1-h2
00:04  85000  85k 0 0 387604 (2) 9 (50,518) 1.056 Kg1-h2 Rb7-b8 Qc6-d7 Kf7-g8 Ra
8-a7 Bd6-f8 Be4-f3 Qb6-e3 d5-d6 Qe3-d3 Qd7-e6 Kg8-h8
00:04  91082  91k 0 0 448125 (2) 10 (55,546) 1.056 Kg1-h2 Rb7-b8 Qc6-d7 Kf7-g8 R
a8-a7 Bd6-f8 Be4-f3 Qb6-e3 d5-d6 Qe3-d3 Qd7-e6 Kg8-h8
++ g1-h1
00:08 119563 120k 0 0 974442 (2) 10 (193,1279) 1.686 Kg1-h1 Rb7-b8 Qc6-d7 Kf7-g8
 Ra8-a7 Bd6-f8 Qd7-f7 Kg8-h7 d5-d6 Qb6-e3 Qf7-c7 Qe3-b6 Qc7xb6 Rb8xb6
00:08 124217 124k 0 0 1090626 (2) 11 (203,1354) 1.686 Kg1-h1 Rb7-b8 Qc6-d7 Kf7-g
8 Ra8-a7 Bd6-f8 Qd7-f7 Kg8-h7 d5-d6 Qb6-e3 Qf7-c7 Qe3-b6 Qc7xb6 Rb8xb6
00:16 148508 149k 0 0 2392471 (2) 12 (317,2238) 1.077 Kg1-h1 Rb7-b8 Qc6-d7 Kf7-g
8 Ra8-a7 Bd6-f8 d5-d6 Qb6-e3 Be4-d5 Kg8-h7 Qd7-f7 Qe3-c1 Kh1-h2 Qc1-f4 Kh2-h1 Qf
4xf5 Qf7-g8 Kh7-g6 Ra7-f7 Qf5-b1 Kh1-h2 Qb1-d3 Bd5xc4 Qd3xc4 Rf7xf8
++ g1-h2
00:25 164463 164k 0 0 4230011 (2) 12 (606,3498) 1.885 Kg1-h2 Rb7-b8 Ra8-a6 Qb6xc
6 d5xc6 Rb8-c8 Ra6-a7 Rc8-c7 Ra7-a5 Bd6-f8 Be4-d5 Kf7-e7 Ra5xb5 Ke7-e8
00:29 167093 167k 0 0 4977715 (2) 13 (730,4253) 1.885 Kg1-h2 Rb7-b8 Ra8-a6 Qb6xc
6 d5xc6 Rb8-c8 Ra6-a7 Rc8-c7 Ra7-a5
00:39 174785 175k 0 0 6895275 (2) 14 (990,5960) 1.885 Kg1-h2 Rb7-b8 Ra8-a6 Qb6xc
6 d5xc6 Rb8-c8 Ra6-a7 Rc8-c7 Ra7-a5
01:07 187554 188k 0 0 12749953 (2) 15 (1896,10809) 1.879 Kg1-h2 Rb7-b8 Ra8-a6 Qb
6xc6 d5xc6 Rb8-c8 Ra6-a7 Kf7-f8 Ra7-b7 Rc8-c7 Rb7xb5 Kf8-e7 g2-g3 Rc7-a7 Kh2-g2
Ra7-a2 Kg2-f3
++ g1-h1
01:31 194665 195k 0 0 17757373 (2) 15 (1999,11475) 1.887 Kg1-h1 Rb7-b8 Ra8-a6 Qb
6xc6 d5xc6 Kf7-f8 Ra6-a7 Rb8-c8 Ra7-b7 Rc8-c7 Rb7xb5 Rc7-a7 Kh1-h2 Ra7-a3 Rb5-b7
 Bd6-e7 Be4-d5 g7-g5
++ g1-f1
02:15 201192 201k 0 0 27342059 (2) 15 (2468,13754) 2.615 Kg1-f1 Rb7-b8 Ra8-a6 Qb
6xc6 d5xc6 Rb8-c8 Ra6-a5 Kf7-f8 Ra5xb5 Kf8-e7 Be4-d5 Rc8-b8 Rb5xb8 Bd6xb8 c6-c7
Bb8xc7 Bd5-g8 g7-g5 f5xg6 Bc7-b6 Bg8xc4
03:42 196695 197k 0 0 43693878 (2) 16 (3835,19815) 2.615 Kg1-f1 Rb7-b8 Ra8-a6 Qb
6xc6 d5xc6 Rb8-c8 Ra6-a5 Kf7-f8 Ra5xb5 Kf8-e7 Be4-d5 Rc8-b8 Rb5xb8 Bd6xb8 c6-c7
Bb8xc7 Bd5-g8 g7-g5 f5xg6 Bc7-b6 Bg8xc4
09:55 191170 191k 0 0 113746222 (2) 17 (8722,41288) 1.965 Kg1-f1 Rb7-b8 Ra8-a6 Q
b6-e3 Qc6-d7 Kf7-g8 Qd7xd6 Rb8-f8 Qd6-e6 Kg8-h8 Be4-f3 Qe3xc3 Qe6-e7 Rf8-g8 Qe7-
c5 Qc3-d3 Kf1-g1 Qd3xf5 d5-d6 Qf5-g5
++ g1-h1
16:12 198065 198k 0 0 192543255 (2) 17 (10061,48543) 2.399 Kg1-h1 Rb7-b8 Ra8-a6
Qb6xc6 d5xc6 Kf7-f8 Ra6-a7 Rb8-c8 Ra7-b7 h6-h5 Rb7xb5 Kf8-e7 Be4-d5 h5-h4 Rb5-b7
 Rc8-c7 b4-b5 Rc7xb7 c6xb7 Bd6-b8 g2-g4 h4xg3 Bd5-g8 Bb8-a7

I'm outsearching deep blue everywhere here. Note that it is interesting
to see how important a few extra moves sometimes are. In this case
my selective search. It used to find Kh1 way sooner because it found it
at depth 14 or 15. So within the 3 minutes. Now it needs 17 ply. That's
bad. But it is because i restricted conditions for DIEP's
7 ply selective search (between main search and quiescencesearch).

>And you are right, of course.  There are details they have not completely
>revealed about
>whatever forward pruning they did to reach that BF...

their overall b.f. as we can see is very poor.

126 million . 10% effective . 180 seconds = 2.3 billion nodes

==> 12th root from 2.3 B = 6.0



>>
>>
>>
>>>
>>>
>>>>
>>>>>
>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>>
>>>>>>>>>
>>>>>>>>>reporting a 12.2 average search depth fullwidth.
>>>>>>>>>I guess you never searched with a decent program fullwidth
>>>>>>>>>with extensions. If you did, you would understand that
>>>>>>>>>getting 12.2 ply fullwidth with loads of extensions is already nearly
>>>>>>>>>impossible. Every extended line is searched to the full depth,
>>>>>>>>>no pruning happens!
>>>>>>>>
>>>>>>>>I agree that 12.2 plies with a lot of extensions and no pruning is impossible
>>>>>>>>for normal programs and also is impossible for deep blue in case that
>>>>>>>>there were real 6 more plies in the hardware.
>>>>>>>>
>>>>>>>>The only case when it may be possible is if the 6 more plies in the hardware are
>>>>>>>>real selective search and it means more pruning than null move with R=3 and in
>>>>>>>>this case the 6 plies in the hardware may be eqvivalent to only 2 plies in
>>>>>>>>software because of big probability to miss things.
>>>>>>>>
>>>>>>>>>
>>>>>>>>>The interesting 2 questions are
>>>>>>>>>  a) did DB use 'no-progress pruning' in SOFTWARE (we know
>>>>>>>>>     already it used it in hardware).
>>>>>>>>
>>>>>>>>They explained in the article that they did not want to take risks of missing
>>>>>>>>something in the first plies so it is clear that they did no pruning in the
>>>>>>>>first 12 plies.
>>>>>>>>
>>>>>>>>If they did some pruning in the software it is clearly after it.
>>>>>>>>
>>>>>>>>I do not know what is exactly no progress pruning.
>>>>>>>>Is there a difference between it and null move pruning?
>>>>>>>>
>>>>>>>>Is no progress pruning more aggresive than null move pruning?
>>>>>>>>
>>>>>>>>Uri



This page took 0.01 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.