Not logged inRybka Chess Community Forum
Up Topic The Rybka Lounge / Computer Chess / Stockfish anti-shuffle patch of 31 March is still glitchy?
- - By MaLaoshi (*) Date 2019-04-03 11:16 Upvotes 1
Posting here since I can't log in to Fishcooking atm; feel free to cross-post.

Been looking forward to play with the anti-shuffle patch since I saw it pass FishTest. Let me first comment that something closely related has existed in Houdini years ago already: an option to lower internally the 50-move rule treshhold to some other value, so that variations ending in long shuffling sequences are evaluated as draws.

Let's start with something that should be easy:
3k4/1r2p3/r2pPp2/b1pP1Pp1/1pP3Pp/pP5P/P3K3/8 b - - 0 6


(This is a few moves into the solution of the pawn-fortress puzzle
3B4/1r2p3/r2p1p2/bkp1P1p1/1p1P1PPp/p1P4P/PPB1K3/8 w - - 0 1

) As long as White declines any rooks offered as bait on b5 etc, there is literally nothing for Black to do, i.e. the position is a draw. Indeed, Stockfish_19032417_x64_modern (i.e. pre-patch) quickly reaches a huge depth ~70, and returns PVs consisting of shuffling only. All this is as expected, and essentially correct except for the spurious ~ -15 eval.

So let's see what the shuffle-aware SF can do in this position:

Stockfish_19033111_x64_modern:
Found 639 tablebases
1/1  00:00   2k  1,563k  +17.48  Kc7
2/2  00:00   3k  2,709k  +17.60  Kc7 Kd2 Bb6
3/3  00:00   5k  4,693k  +17.67  Kc7 Kf2 Rb8
4/4  00:00   6k  6,069k  +17.67  Kc7 Kf2 Rb8 Kg2
5/5  00:00   14k  6,871k  +17.65  Kc7 Ke3 Bb6 Kf2 Rb8
6/6  00:00   16k  8,081k  +17.67  Kc7 Ke3 Bb6 Kf2 Rb8 Kg2
7/7  00:00   23k  11,477k  +17.72  Kc7 Kf2 Rb8 Kf3 Bb6 Kg2 Kb7
8/8  00:00   36k  17,964k  +17.76  Bb6 Kf2 Kc8 Kg2 Rc7 Kf2 Kb7 Kg2
9/9  00:00   60k  20,138k  +17.74  Bb6 Kf3 Kc8 Kg2 Ra5 Kf2 Rc7 Kg2 Kb7
10/10  00:00   68k  22,555k  +17.77  Bb6 Kf3 Kc8 Kf2 Rb8 Kf3 Kc7 Kg2 Kb7 Kf3
11/11  00:00   82k  27,485k  +17.77  Bb6 Kf3 Kc8 Kf2 Rb8 Kf3 Kc7 Kg2 Rc8 Kf2 Rd8
12/12  00:00   110k  36,786k  +17.76  Bb6 Kf3 Kc8 Kg2 Kc7 Kh2 Rb8 Kg2 Rg8 Kf3 Rc8 Kg2 Rd8
13/13  00:00   175k  35,010k  +17.83  Bb6 Kf3 Ke8 Kf2 Kf8 Kg2 Ra5 Kf2 Kg7 Kg2 Rc7 Kf2 Ra6
14/15  00:00   211k  42,239k  +17.78  Bb6 Kf3 Ke8 Kf2 Kf8 Kg2 Ra5 Kf3 Kg7 Kg2 Rc7 Kf2 Ra8 Kf3 Rd8
15/43  00:00   365k  45,600k  +17.69  Bb6 Kf2 Ke8 Kf3 Rb8 Kg2 Kd8 Kf2 Raa8 Kg1 Rb7 Kf2 Kc8 Kg2 Ra6
16/87  00:00   3,129k  53,949k  +17.16  Bb6 Kf2 Ra8 Kg1 Kc7 Kg2 Kb8 Kf2 Bc7 Kg2 Ra6 Kf1 Bb6 Kg2 Ra5
17/87  00:01   99,178k  56,836k  +13.11  Bb6 Kf2 Bc7 Kf1 Ba5 Kg2 Rab6 Kf1 Ra7 Kg1 Rb8 Kf2 Kc7 Kf1 Rba8 Kg1 Bb6 Kf2 Ra6 Kf1 R8a7 Kg2 Ba5 Kf1 Rb7 Kf2 Kc8 Kf1 Raa7 Kg1 Rb8 Kf1 Bb6 Kg1
18/49  00:01   100,201k  56,803k  +13.11  Bb6 Kf2 Kc7 Kf1 Rba7 Kg2 Ra8 Kf2 Rb8 Kg1 Kc8 Kf1 Ba5 Kg2 Raa8
19/42  00:01   101,131k  56,783k  +15.22  Bb6 Kf2 Kc7 Kf1 Ra5 Kg1 Kc8 Kf2 Rba7 Ke2 Rc7
20/59  00:02   136,277k  56,829k  +12.91  Bb6 Kf2 Ke8 Kg1 Ba5 Kg2 Kf8 Kg1 Ra8 Kf2 Bb6 Kf1 Rd7 exd7 Ra7 d8Q+ Bxd8 Kg2 Rb7 Kg1 Rd7 Kf2 Ba5 Kf1 Bb6 Kg2
21/51  00:02   145,116k  56,819k  +12.98  Bb6 Kf2 Rba7 Ke2 Ra5 Kf2 Rc7 Kg1 Ra6 Kg2 Ra8 Kf2 Kc8 Kg1 Rb8
22/40  00:02   146,499k  56,805k  +12.89  Bb6 Kf2 Rba7 Ke2 Ra5 Kf2 Kc7 Kg2 Rb7 Kg1 Raa7 Kf1 Kd8 Kg2 Ra5
23/60  00:02   168,219k  56,792k  +12.92  Bb6 Kf2 Rba7 Ke2 Kc7 Kf1 Rb7 Kf2 Raa7 Kf1 Ra8 Kg1 Ba7 Kg2 Kd8 Kf2 Rbb8 Kf1 Bb6 Kg1 Ra7 Kf1 Ke8 Kf2 Rba8 Kg2 Ra5 Kf2 Kd8 Kg1 Ba7 Kh1 Rb5 cxb5 Rb8 b6
24/49  00:03   179,229k  56,790k  +12.96  Bb6 Kf3 Rc7 Kg2 Rca7 Kf1 Ba5 Ke2 Bc7 Kf2 Ra8 Kg1 Rb8 Kf2 Ba5 Kf1 Rb7 Kg2 Ke8 Kf2 Rb5 Kg2 Ra8 cxb5 Rb8 Kf1
25/40+  00:03   179,880k  56,780k  +13.05  Bb6
25/40+  00:03   179,956k  56,786k  +13.15  Bb6
25/41+  00:03   181,233k  56,777k  +15.97  Bb6
25/59  00:04   236,594k  56,737k  +13.55  Bb6 Kf2 Ra8 Kf1 Kc7 Kf2 Ra4 Kg2 Rb8 Kg1 Rba8 bxa4 Kd8 a5 Rxa5 Kg2
26/38+  00:04   237,585k  56,743k  +13.65  Bb6
26/46+  00:04   237,877k  56,732k  +13.74  Bb6
26/46+  00:04   238,007k  56,736k  +15.26  Bb6
26/46+  00:04   240,353k  56,727k  +15.87  Ke8
26/46  00:04   246,855k  56,722k  +13.65  Ke8 Kf2 Bb6 Kg2 Rba7 Kf2 Ra8 Kg1 R6a7 Kf1 Kf8 Kg2 Rc8 Kg1 Ke8 Kf2 Kd8 Ke2 Rac7 Kf3 Rb7 Kf2 Kc7 Kg2 Ra7 Kg1 Ra6 Kf1 Rb8 Kf2 Ra7 Kf1 Ra4 bxa4 Ra8 Kg1
27/38+  00:04   247,568k  56,730k  +13.75  Ke8
27/38+  00:04   247,627k  56,717k  +13.84  Ke8
27/43+  00:04   257,648k  56,701k  +16.70  Ke8
27/48  00:04   267,926k  56,680k  +13.83  Ke8 Kf1 Kf8 Ke2 Bc7 Kf3 Bb6 Kg2 Rb8 Kf3 Ra4 Ke2 Rd8 Kf2 Rc8 bxa4 Ra8 a5 Rxa5 Ke2
28/38+  00:04   268,080k  56,677k  +13.92  Ke8
28/38+  00:04   268,095k  56,680k  +14.01  Ke8
28/50+  00:06   358,356k  56,577k  +16.07  Ke8
28/50  00:06   359,198k  56,576k  +15.91  Ke8 Kf2 Bb6 Kg2 Kf8 Kf2 Rb8 Kg2 Raa8 Kf2 Kg7 Kf1 Rd8 Kg2 Re8 Kf3 Reb8 Kf2 Kf8 Kf1 Rb7 Kg2 Rba7 Kh2 Kg7 Kg1 Rb7 Kf1 Ra6 Kg2 Ra4 bxa4 Ra7 Kf3
29/39+  00:06   360,612k  56,575k  +16.00  Ke8
29/40-  00:06   370,254k  56,588k  +15.81  Ke8 Kf2
29/40+  00:06   374,106k  56,588k  +15.95  Ke8
29/44  00:09   531,834k  56,765k  +15.78  Ke8 Kf2 Rba7 Kf1 Kf8 Kg2 Bb6 Kg1 Ra8 Kf2 R6a7 Kg2 Rb7 Kg1 Raa7 Kf2 Ra6 Kf1 Ra4 Kg2 Ke8 bxa4 Ra7 a5 Rxa5 Kh2 Kf8 Kg1
30/40+  00:09   543,797k  56,782k  +15.87  Ke8
30/41  00:09   551,124k  56,788k  +15.79  Ke8 Kf2 Rba7 Kg1 Bb6 Kf2 Rc7 Kg1 Rc8 Kg2 Rb8 Kf2 Ra7 Kg2 Ba5 Kh2 Rc8 Kg2 Rca8 Kf1 Bb6 Kf2 Kf8 Kg2 Rd8 Kf1 Rb8 Kg1 Rc8 Kf1 Ke8 Kf2 Ra4 bxa4 Ra8 Ke2
31/71+  00:58   3,340,013k  57,500k  +15.89  Ke8
31/71+  00:58   3,340,432k  57,500k  +15.98  Ke8
31/71+  00:58   3,346,099k  57,508k  +17.41  Ke8
31/102  00:58   3,364,581k  57,503k  +15.89  Ke8 Kf1 Bb6 Kf2 Rc7 Ke2 Kf8 Kf2 Kg7 Kg1 Ra8 Kg2 Rb7 Kf1 Rbb8 Kf2 Ra7 Kg1 Rc8 Kf1 Ba5 Kg2 Rca8 Kf1 Rh8 Kg2 Bc7 Kg1 Ra4 Kf2 Rb8 bxa4 Rb7 a5
32/87  03:14   11,156,856k  57,241k  +15.90  Ke8 Kf1 Kf8 Kf2 Rb8 Kg2 Rba8 Kg1 Bc7 Kg2 Bd8 Kf2 R8a7 Ke2 Ke8 Ke3 Bb6 Kf2 Kf8 Kf1 Ra8 Kg2 Ra4 Kf1 Kg8 Kf2 Rc8 Kg2 Kh7 Kf1 Kg7 bxa4 Ra8 Kg2 Rxa4
33/45  04:23   15,210,406k  57,668k  +15.90  Ke8 Kf1 Kf8 Kf2 Rb8 Kg2 Rba8 Kg1 Bc7 Kg2 Bd8 Kf2 R8a7 Ke2 Bb6 Kf1 Ra8 Kg2 R6a7 Kf2 Ke8 Kf1 Ra4 Kg2 R8a7 Kf2 Rb7 Kg2 Rc7 bxa4 Ra7 a5 Rxa5
34/38  04:23   15,210,546k  57,668k  +15.90  Ke8 Kf1 Kf8 Kf2 Rb8 Kg2 Rba8 Kg1 Bc7 Kg2 Bd8 Kf2 R8a7 Ke2 Bb6 Kf1 Ra8 Kg2 R6a7 Kf2 Ke8 Kf1 Ra4 Kg2 R8a7 Kf2 Rb7 Kg2 Rc7 bxa4 Ra7 a5 Rxa5 Kf1
35/38  04:23   15,211,193k  57,668k  +15.90  Ke8 Kf1 Kf8 Kf2 Rb8 Kg2 Rba8 Kg1 Bc7 Kg2 Bd8 Kf2 R8a7 Ke2 Bb6 Kf1 Ra8 Kg2 R6a7 Kf2 Ke8 Kf1 Ra4 Kg2 R8a7 Kf2 Rb7 Kg2 Rc7 bxa4 Ra7 a5 Rxa5 Kf1 b3
36/41  04:24   15,244,873k  57,673k  +15.90  Ke8 Kf1 Kf8 Kf2 Rb8 Kg2 Rba8 Kg1 Bc7 Kg2 Bd8 Kf2 R8a7 Ke2 Bb6 Kf1 Ra8 Kg2 R6a7 Kf2 Rc7 Kf1 Kg7 Kg2 Rb7 Kf1 Ra6 Kf2 Ba7 Kg2 Kf8 Kf1 Rc7 Kg2 Ra4 bxa4 b3 axb3
37/40  04:42   16,291,062k  57,760k  +15.90  Ke8 Kf1 Kf8 Kf2 Rb8 Kg2 Rba8 Kg1 Bc7 Kg2 Bd8 Kf2 R8a7 Ke2 Bb6 Kf1 Ra8 Kg2 R6a7 Kf2 Rc7 Kf1 Ra4 Kf2 Ke8 Ke2 Rb7 Kf1 Kd8 Kg2 Kc8 Kf2 Kb8 bxa4 Kc7 Kf3 Ra7
38/42  04:47   16,608,462k  57,777k  +15.90  Ke8 Kf1 Kf8 Kf2 Rb8 Kg2 Rba8 Kg1 Bc7 Kg2 Bd8 Kf2 R8a7 Ke2 Bb6 Kf1 Ra8 Kg2 R6a7 Kf2 Rc7 Kf1 Kg7 Kg2 Rb8 Kf2 Ra7 Ke2 Ra4 Kf2 Rc8 Kf1 Bd8 Ke2 Rc7 bxa4 Ra7 Ke3
39/45  25:35   89,622,339k  58,382k  +15.90  Ke8 Kf1 Kf8 Kf2 Rb8 Kg2 Rba8 Kg1 Bc7 Kg2 Bd8 Kf2 R8a7 Ke2 Bb6 Kf1 Ra8 Kg2 R6a7 Kf2 Rc8 Kf3 Ra5 Kg2 Raa8 Kf1 Kg7 Ke1 Ra6 Kf2 Ra4 Kf1 Rc7 Kg1 Kf8 bxa4 b3 Kf2 b2
40/46  36:20  127,015,756k  58,255k  +15.90  Ke8 Kf1 Kf8 Kf2 Rb8 Kg2 Rba8 Kg1 Bc7 Kg2 Bd8 Kf2 R8a7 Ke2 Bb6 Kf1 Ra8 Kg2 R6a7 Kf2 Rd8 Kg2 Kg8 Kf2 Ra4 Kg1 Rf8 Kf1 Ba7 bxa4 Rd8 Kg2 Rb8 Kh1 Bb6 Kg2 Rb7 a5 Bxa5 Kh2 Kf8
41/47  49:54  173,774,710k  58,029k  +15.90  Ke8 Kf1 Kf8 Kf2 Rb8 Kg2 Rba8 Kg1 Bc7 Kg2 Bd8 Kf2 R8a7 Ke2 Bb6 Kf1 Ra8 Kg2 R6a7 Kf2 Rd8 Kg2 Rb7 Kg1 Ba7 Kf2 Rb5 Ke2 Bb6 Kd2 Ke8 Kc1 Rb8 cxb5 Rc8 Kd2 Kd8 Kc2 Kc7 Kd3 Rb8
42/56  1:23:20  288,669,159k  57,730k  +15.90  Ke8 Kf1 Kf8 Kf2 Rb8 Kg2 Rba8 Kg1 Bc7 Kg2 Bd8 Kf2 R8a7 Ke2 Bb6 Kf1 Kg7 Kg2 Ra5 Kf1 Kg8 Kf2 Bc7 Kg2 Ra8 Kf2 Kg7 Ke2 Rb8 Kf2 Ra4 Kg2 Bb6 Kf2 Bd8 Kg2 Bc7 bxa4 Re8 Kh1 Rc8 Kg2 Kh8
43/47  1:29:37  310,168,869k  57,681k  +15.90  Ke8 Kf1 Kf8 Kf2 Rb8 Kg2 Rba8 Kg1 Bc7 Kg2 Bd8 Kf2 R8a7 Ke2 Bb6 Kf1 Kg7 Kg2 Ra5 Kf1 Kg8 Kf2 Bc7 Kg2 Ra4 Kg1 Kf8 Kh2 Ra8 Kg2 Ba5 Kf1 Bd8 Ke1 Kg8 Ke2 Bb6 bxa4 Kg7 Ke1 Rb8 Kf1 Kf8


One notices a couple things: *the depth is much lower now, in fact ridiculously low for such a simple position; the eval is not improved; the PVs have become worse, including moves close to their end where the bait has been taken, which in fact is a fatal mistake. Not visible in this log but visible in the Arena GUI, the TB and hash usage were both reported as zero. This is suspicious, since in fact TBs are useful for rejecting the bait-taking lines. To be sure, the pre-patch SF of 24 March reported normal hash&TB usage.

******************************

On to a second example:
q7/1p6/1Pp1k3/p1Pp4/P2Pp1p1/4PpPp/5P1P/5K1R b - - 6 5


(This is a few moves into the solution of the "claustrophilia" puzzle
8/1p6/1Pp2N1q/p1Ppk2p/P3p3/3PPpPp/3K1P1P/1R6 w - - 0 1

) I get the output

Stockfish_19033111_x64_modern:
Found 639 tablebases
1/2  00:00   77  19k  +8.63  Qa6+ Kg1
2/3  00:00   98  25k  +9.23  Qa6+ Kg1 Qc4
3/4  00:00   6k  568k  +9.23  Qa6+ Kg1 Qa8 Kf1
4/5  00:00   7k  637k  +8.26  Kf5 Kg1 Ke6 Kf1
5/6  00:00   8k  686k  +8.24  Kf6 Kg1 Qe8 Kf1 Qg6
6/6  00:00   8k  700k  +8.24  Kf6 Kg1 Qe8 Kf1 Qg6 Ke1
7/7  00:00   13k  1,006k  +8.82  Kf6 Kg1 Kg6 Kf1 Kg5 Kg1
8/8  00:00   26k  1,721k  +7.88  Kf6 Kg1 Kf7 Kf1 Ke6 Kg1 Qd8 Kf1
9/10  00:00   26k  1,763k  +7.88  Kf6 Kg1 Kf7 Kf1 Ke6 Kg1 Qd8 Kf1 Qh8 Ke1
10/14  00:00   42k  2,225k  +7.75  Kf6 Kg1 Ke6 Kf1 Kf5 Kg1 Qc8 Kf1 Qe6 Ke1 Qg8 Kf1
11/17  00:00   60k  2,835k  +7.65  Kf6 Kg1 Kg6 Kf1 Kg5 Kg1 Qe8 Kf1 Kf6 Ke1 Qa8 Kf1 Qb8 Kg1
12/41  00:00   138k  3,459k  +7.46  Kf6 Kg1 Qb8 Kf1 Ke7 Ke1 Qa8 Kf1 Ke6 Kg1 Qh8 Kf1 Kf5 Ke1
13/39  00:00   178k  4,151k  +7.44  Kf6 Kg1 Kf5 Kf1 Ke6 Kg1 Qb8 Kf1 Qf8 Rg1 Qa8 Rh1 Kf5 Kg1
14/54  00:00   563k  9,869k  +7.76  Kf6 Kg1 Qb8 Kf1 Qg8 Ke1 Qd8 Kf1 Ke6 Kg1 Kf5 Kf1 Qa8
15/23  00:00   651k  11,035k  +7.37  Kf6 Kg1 Ke7 Kf1 Ke6 Kg1 Qe8 Kf1 Kf6 Ke1 Kf7 Kf1 Ke6
16/52  00:00   1,033k  15,412k  +8.46  Kf6 Kg1 Ke7 Kf1 Ke6 Kg1 Kf6 Kf1 Kf5 Kg1 Qg8 Kf1 Qe8
17/33  00:00   1,288k  17,404k  +7.84  Kf6 Kg1 Qc8 Kf1 Qb8 Ke1 Kf5 Kf1 Qa8 Kg1 Qe8 Kf1 Ke6 Ke1 Qg8 Kd2 Qa8 Rc1 Qa6 Rc2 Qd3+ Kc1 Kf7 Kb2 Qf1 Ka2 Kg6 Kb2 Qg1 Ka2 Qxh2 Kb2 Qg2
18/75  00:00   43,002k  47,727k   0.00  Kf6 Kg1 Qf8 Kf1
19/17  00:00   43,021k  47,695k   0.00  Kf6 Kg1 Qf8 Kf1 Ke7 Ke1 Qh8 Kf1 Ke6 Ke1
20/25  00:00   43,078k  47,706k   0.00  Qg8 Ke1 Qe8 Kf1
21/17  00:00   43,095k  47,724k   0.00  Qg8 Ke1 Qe8 Kf1
22/13  00:00   43,125k  47,704k   0.00  Qg8 Ke1 Qe8 Kf1
23/50  00:00   43,276k  47,714k  +8.46  Kf7 Kg1 Ke7 Kf1 Ke6 Kg1 Qg8 Kf1 Qd8 Kg1 Qg5 Kf1
24/64  00:23   1,190,448k  51,118k  +7.68  Qb8 Rg1 Qg8 Rh1 Qg7 Rg1 Qg6 Rh1 Qh5 Ke1 Qh8 Kf1 Kf7 Ke1 Qf8 Kf1 Ke8 Ke1 Qg7 Kd1 Kd7 Ke1 Qg8 Kd2 Qa8 Rc1 Qa6 Rc2 Qd3+ Kc1 Qb3 Rd2 Qxa4 Rb2 Qc4+ Rc2
25/40-  00:23   1,190,579k  51,118k  +7.58  Qb8 Rg1
25/40+  00:23   1,190,623k  51,117k  +7.68  Qb8
25/40-  00:23   1,191,759k  51,120k  +7.49  Qb8 Ke1
25/40+  00:23   1,191,779k  51,121k  +7.65  Qb8
25/40-  00:23   1,193,269k  51,121k  +7.29  Qb8 Ke1
25/40+  00:23   1,193,287k  51,122k  +7.65  Qb8
25/40-  00:23   1,194,927k  51,122k  +6.93  Qb8 Ke1
25/40+  00:23   1,194,960k  51,123k  +7.57  Qb8
25/40-  00:23   1,196,353k  51,126k  +6.32  Qb8 Ke1
25/41+  00:23   1,196,741k  51,125k  +7.48  Qb8
25/51  00:54   2,862,420k  52,764k  +7.64  Qb8 Ke1 Kf6 Kf1 Qg8 Ke1 Ke7 Kf1 Ke6 Ke1 Qg5 Kf1 Kf7 Ke1 Qg7 Kf1 Qh8 Rg1 Qe8 Rh1 Qd8 Ke1 Kg7 Kd2 Qa8 Kc3 Kh8 Rc1 Qa6 Rc2 Qc4+ Kb2 Qxa4 Kb1 Qb4+ Rb2
26/31-  00:54   2,862,488k  52,765k  +7.44  Qb8 Ke1
26/31-  00:54   2,862,498k  52,765k  +7.29  Qb8 Ke1
26/31+  00:54   2,862,550k  52,765k  +7.64  Kf7
26/31-  00:54   2,862,562k  52,765k  +7.08  Kf7 Kg1
26/31-  00:54   2,862,614k  52,765k  +6.72  Kf7 Kg1
26/41+  00:54   2,862,712k  52,765k  +7.15  Kf7
26/42+  00:54   2,866,789k  52,763k  +7.62  Kf7
26/47-  01:07   3,569,532k  53,096k  +6.37  Kf7 Kg1
26/47+  01:07   3,569,816k  53,096k  +7.30  Kf7
26/57  01:16   4,070,911k  53,295k  +7.49  Kf7 Kg1 Qb8 Kf1 Qg8 Kg1 Qg6 Kf1 Ke8 Kg1 Kf8 Kf1 Qg8 Rg1 Ke8 Rh1 Qf7 Ke1 Qe6 Kd2 Qg8 Rc1 Kd7 Ke1 Qa8 Kf1 Qa6+ Kg1 Qd3 Re1 Qb3 Rc1 Qxa4 Rf1 Qa3 Re1 Ke6 Rb1
27/29-  01:16   4,071,026k  53,295k  +7.39  Kf7 Kg1
27/29-  01:16   4,071,050k  53,296k  +7.29  Kf7 Kg1
27/33-  01:16   4,071,462k  53,296k  +6.61  Kf7 Kg1
27/39+  01:16   4,071,700k  53,296k  +6.84  Kf7
27/39+  01:16   4,071,720k  53,295k  +7.31  Kf7
27/41-  01:16   4,077,239k  53,295k  +6.26  Kf7 Kg1
27/43+  01:16   4,077,957k  53,295k  +7.09  Kf7
27/48  02:10   7,023,616k  53,867k  +7.63  Kf7 Kg1 Qh8 Kf1 Ke8 Rg1 Qf6 Rh1 Qe6 Rg1 Ke7 Rh1 Qc8 Kg1 Qe8 Kf1 Kf6 Kg1 Qg8 Kf1 Qg6 Kg1 Qg7 Kf1 Kg5 Ke1 Qh8 Kd2 Qa8 Rc1 Qa6 Rc2 Qd3+ Kc1 Qa3+ Kb1 Qxa4 Rd2 Qc4
28/25-  02:10   7,023,709k  53,867k  +7.53  Kf7 Kg1
28/25-  02:10   7,023,823k  53,867k  +7.44  Kf7 Kg1
28/29-  02:10   7,024,325k  53,867k  +6.84  Kf7 Kg1
28/41+  02:10   7,025,633k  53,867k  +7.05  Kf7
28/41-  02:10   7,025,721k  53,867k  +6.48  Kf7 Kg1
28/41+  02:10   7,025,795k  53,866k  +7.00  Kf7
28/54+  02:46   8,991,016k  53,937k  +7.78  Kf7
28/54  02:48   9,088,349k  53,939k  +7.67  Kf7 Kg1 Qh8 Kf1 Qg7 Ke1 Ke8 Rf1 Kd7 Rh1 Qg8 Kf1 Qa8 Kg1 Qf8 Kf1 Ke6 Kg1 Qe8 Kf1 Ke7 Rg1 Qg6 Ke1 Qe6 Kd2 Qg8 Kc2 Qa8 Kb2 Kd7 Kc3 Qa6 Rc1 Qc4+ Kb2 Qxa4 Rg1 Qb5+ Kc1
29/33-  02:48   9,088,636k  53,939k  +7.58  Kf7 Kg1
29/33-  02:48   9,088,859k  53,939k  +7.48  Kf7 Kg1
29/43+  02:48   9,088,994k  53,939k  +7.58  Kf7
29/43-  02:48   9,090,646k  53,939k  +7.34  Kf7 Kg1
29/43+  02:48   9,090,708k  53,939k  +7.57  Kf7
29/43-  02:48   9,092,319k  53,938k  +7.07  Kf7 Kg1
29/45+  02:48   9,092,577k  53,938k  +7.52  Kf7
29/47-  02:49   9,121,074k  53,941k  +6.61  Kf7 Kg1
29/47+  02:49   9,121,499k  53,941k  +7.37  Kf7
29/66  04:29   14,573,470k  54,165k  +7.41  Kf7 Kg1 Qh8 Kf1 Qf8 Ke1 Qb8 Kf1 Qh8 Ke1 Qf6 Kf1 Kg7 Ke1 Qf7 Rg1 Qe6 Rh1 Kf8 Kd1 Ke8 Kd2 Kd7 Rc1 Qg8 Ke1 Qa8 Kf1 Qa6+ Kg1 Qd3 Ra1 Qc4 Kh1 Qc3 Rf1 Qc2 Kg1 Qxa4 Rb1 Qc4
30/43-  04:29   14,573,912k  54,165k  +7.32  Kf7 Kg1
30/43+  04:29   14,574,347k  54,165k  +7.41  Kf7
30/43-  04:29   14,574,752k  54,165k  +7.23  Kf7 Kg1
30/43+  04:29   14,574,875k  54,165k  +7.39  Kf7
30/47-  04:41   15,253,305k  54,202k  +7.03  Kf7 Kg1
30/47+  04:41   15,253,455k  54,202k  +7.34  Kf7
30/53-  04:45   15,451,378k  54,210k  +6.67  Kf7 Kg1
30/53+  04:45   15,451,515k  54,210k  +7.24  Kf7
30/65-  16:04   52,344,411k  54,260k  +6.06  Kf7 Kg1
30/65+  16:04   52,344,641k  54,260k  +7.04  Kf7


Looks to me as if the problems of SF 31 Mar are the same as in the previous test position: low depth, nonsensical PVs, unimproved evals, and also the hash&TB usage was again reported as zero. Perhaps these are corner cases, where the attacking side literally has no non-shuffle moves left. But still, in such a case, SF should return a draw eval (or increase the 36-move limit if we are at root) instead of forcing the defender to make some nonsensical move.

To be sure, my point is not to argue for reverting this patch. It's a long time since I've played with Houdini, but as far as I remember its modified 50-move rule worked pretty much as advertised. In Stockfish, with its much deeper search than old Houdinis, something similar should be a good deal more powerful. But it doesn't look to me as if all the kinks have been worked out already.
Parent - - By Peter Grayson (****) Date 2019-04-03 15:07 Upvotes 1
The first position suggests an engine issue.
Running on a Q9550 +512 Mb hash it finds 6..Ke8 60/61 ply depth at 03:48.

Running on a dual Xeon E5-2687w allocated 16 cores + 4Gb hash on several runs, it gets stuck for a while at 25 to 28 ply depending on move selection with typical example  3:49 minutes to reach 26/62 ply depth with 6..Rb8 despite recording 47000 kN/s - see attached txt files.

The engine identifies a draw (stalemeate) very quickly in the second position ...

Q9550 CPU at 2.83 GHz + 512Mb hash
New game
q7/1p6/1Pp1k3/p1Pp4/P2Pp1p1/4PpPp/5P1P/5K1R b - - 6 5


Analysis by Stockfish 310319 64:

5...Kf6 6.Kg1 Kf5 7.Kf1 Kg5 8.Kg1 Kf6 9.Kf1 Ke7 10.Kg1 Ke8 11.Kf1 Kd7
  -+ (-8.40)  Depth: 13/32   00:00:00  22kN
5...Kf6 6.Kg1 Qe8 7.Kf1 Kf7 8.Kg1 Ke7 9.Kf1 Qg8 10.Kg1 Qd8 11.Kf1 Ke6 12.Ke1 Kd7
  -+ (-6.36)  Depth: 14/57   00:00:00  698kN
5...Kf6 6.Kg1 Qe8 7.Kf1 Kf7 8.Kg1 Ke6 9.Kf1 Qg8 10.Kg1 Kd7 11.Kf1 Qa8
  -+ (-6.27)  Depth: 15/57   00:00:00  757kN
5...Kf6 6.Kg1 Qe8 7.Kf1 Kf7 8.Kg1 Qd8 9.Kf1 Ke6 10.Kg1 Ke7 11.Kf1 Qf8 12.Kg1 Qb8
  -+ (-5.98)  Depth: 16/57   00:00:00  846kN
5...Kf6 6.Kg1 Qb8 7.Kf1 Kg6 8.Kg1 Kh7 9.Kf1 Qa8 10.Kg1 Qd8 11.Kf1 Qf6 12.Ke1
  -+ (-5.84)  Depth: 17/57   00:00:00  990kN
5...Kf6 6.Kg1 Qd8 7.Kf1 Qe7 8.Kg1
  = (0.00)  Depth: 18/102   00:00:01  4745kN
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Kd7 8.Kg1 Qd8 9.Kf1 Kc8 10.Ke1 Qe8 11.Kd2
  -+ (-7.14)  Depth: 19/102   00:00:01  4753kN
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Kd7 8.Kg1 Qd8 9.Kf1 Kc8 10.Ke1 Qe8 11.Kd2
  = (0.00)  Depth: 20/102   00:00:01  4792kN
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Kd7 8.Kg1 Qd8 9.Kf1 Kc8 10.Ke1 Qe8 11.Kd2
  = (0.00)  Depth: 21/5   00:00:01  4795kN
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Kd7 8.Kg1 Qd8 9.Kf1 Kc8 10.Ke1 Qe8 11.Kd2
  = (0.00)  Depth: 22/5   00:00:01  4801kN
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Kd7 8.Kg1 Qd8 9.Kf1 Kc8 10.Ke1 Qe8 11.Kd2
  = (0.00)  Depth: 23/5   00:00:01  4813kN
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Kd7 8.Kg1 Qd8 9.Kf1 Kc8 10.Ke1 Qe8 11.Kd2
  = (0.00)  Depth: 24/5   00:00:01  4825kN
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qh8 8.Ke1 Qe8 9.Kf1
  = (0.00)  Depth: 25/13   00:00:01  4852kN
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qh8 8.Ke1 Qe8 9.Kf1 Qa8 10.Kg1 Kf6 11.Kf1 Qa6+ 12.Kg1
  = (0.00)  Depth: 26/16   00:00:01  4872kN
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qh8 8.Ke1 Qe8 9.Kf1 Qa8 10.Kg1 Kf6 11.Kf1 Qa6+ 12.Kg1
  = (0.00)  Depth: 27/5   00:00:01  4892kN
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qh8 8.Ke1 Qe8 9.Kf1 Qa8 10.Kg1 Kf6 11.Kf1 Qa6+ 12.Kg1
  = (0.00)  Depth: 28/5   00:00:01  4903kN
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Kd7 8.Kg1 Qd8 9.Kf1 Kc8 10.Ke1 Qe8 11.Kf1
  = (0.00)  Depth: 29/21   00:00:01  4921kN
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qe8 8.Ke1 Kf6 9.Kf1
  = (0.00)  Depth: 30/17   00:00:01  4946kN
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qe8 8.Ke1 Kf6 9.Kf1
  = (0.00)  Depth: 31/5   00:00:01  4960kN
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qe8 8.Ke1 Kf6 9.Kf1
  = (0.00)  Depth: 32/5   00:00:01  4988kN
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qg8 8.Kg1 Kd7 9.Kf1 Qg5 10.Ke1 Kc8 11.Kd2 Kb8 12.Ke1
  = (0.00)  Depth: 33/19   00:00:01  5044kN
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qg8 8.Kg1 Kd7 9.Kf1 Qg5 10.Ke1 Kc8 11.Kd2 Kb8 12.Ke1
  = (0.00)  Depth: 34/5   00:00:01  5055kN
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qh8 8.Ke1 Qe8 9.Kf1 Qa8 10.Kg1 Kf6 11.Kf1 Kg5 12.Kg1
  = (0.00)  Depth: 35/17   00:00:01  5061kN
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qh8 8.Ke1 Qe8 9.Kf1 Qa8 10.Kg1 Kf6 11.Kf1 Kg5 12.Kg1
  = (0.00)  Depth: 36/17   00:00:01  5113kN
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qh8 8.Ke1 Qe8 9.Kf1 Qa8 10.Kg1 Kf6 11.Kf1 Kg5 12.Kg1
  = (0.00)  Depth: 37/5   00:00:01  5147kN
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qg8 8.Kg1
  = (0.00)  Depth: 38/29   00:00:01  5185kN
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Kd7 8.Kg1
  = (0.00)  Depth: 39/21   00:00:01  5223kN
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qd8 8.Ke1 Qh8 9.Kf1 Qa8 10.Kg1 Kf6 11.Kf1
  = (0.00)  Depth: 40/45   00:00:01  5245kN
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qd8 8.Ke1 Qh8 9.Kf1 Qa8 10.Kg1 Kf6 11.Kf1
  = (0.00)  Depth: 41/45   00:00:01  5278kN
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qd8 8.Ke1 Qh8 9.Kf1 Qa8 10.Kg1 Kf6 11.Kf1
  = (0.00)  Depth: 42/5   00:00:01  5315kN
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qf8 8.Kg1
  = (0.00)  Depth: 43/45   00:00:01  5393kN
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qb8 8.Kg1 Kf6 9.Kf1 Qh8 10.Rg1 Qh6 11.Rh1
  = (0.00)  Depth: 44/29   00:00:01  5454kN
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qb8 8.Kg1 Kf6 9.Kf1 Qh8 10.Rg1 Qh6 11.Rh1
  = (0.00)  Depth: 45/5   00:00:01  5492kN
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qb8 8.Kg1 Kf6 9.Kf1 Qh8 10.Rg1 Qh6 11.Rh1
  = (0.00)  Depth: 46/5   00:00:01  5531kN
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qb8 8.Kg1 Kf6 9.Kf1 Qh8 10.Rg1 Qh6 11.Rh1
  = (0.00)  Depth: 47/5   00:00:01  5579kN
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qb8 8.Kg1 Kf6 9.Kf1 Qh8 10.Rg1 Qh6 11.Rh1
  = (0.00)  Depth: 48/31   00:00:01  5661kN
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qc8 8.Ke1 Qe8 9.Kf1
  = (0.00)  Depth: 49/25   00:00:01  5793kN
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qc8 8.Ke1 Qe8 9.Kf1
  = (0.00)  Depth: 50/5   00:00:01  5901kN
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Kd7 8.Kg1 Qd8 9.Kf1 Kc8 10.Ke1 Qe8 11.Kf1
  = (0.00)  Depth: 51/21   00:00:01  6040kN
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qd8 8.Ke1 Qh8 9.Kf1 Qa8 10.Kg1 Kf6 11.Kf1 Kg5 12.Kg1 Qf8 13.Kf1 Qc8 14.Kg1
  = (0.00)  Depth: 52/25   00:00:01  6159kN
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qd8 8.Ke1 Qh8 9.Kf1 Qa8 10.Kg1 Kf6 11.Kf1 Kg5 12.Kg1 Qf8 13.Kf1 Qc8 14.Kg1
  = (0.00)  Depth: 53/5   00:00:01  6271kN
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qc8 8.Ke1 Qf8 9.Kf1 Qa8 10.Kg1 Kf6 11.Kf1 Kg5 12.Kg1
  = (0.00)  Depth: 54/25   00:00:01  6499kN
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qc8 8.Ke1 Qf8 9.Kf1 Qa8 10.Kg1 Kf6 11.Kf1 Kg5 12.Kg1
  = (0.00)  Depth: 55/5   00:00:01  6618kN
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qe8 8.Ke1 Kf6 9.Kf1 Kf7 10.Kg1 Kg8 11.Kf1
  = (0.00)  Depth: 56/27   00:00:01  6784kN
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qe8 8.Ke1 Kf6 9.Kf1 Kf7 10.Kg1 Kg8 11.Kf1
  = (0.00)  Depth: 57/5   00:00:01  6871kN
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qd8 8.Ke1 Qh8 9.Kf1 Qa8 10.Kg1 Kf6 11.Kf1 Qa6+ 12.Kg1
  = (0.00)  Depth: 58/15   00:00:01  7181kN
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qd8 8.Ke1 Qh8 9.Kf1 Qa8 10.Kg1 Kf6 11.Kf1 Qa6+ 12.Kg1
  = (0.00)  Depth: 59/5   00:00:01  7363kN
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qd8 8.Ke1 Qh8 9.Kf1 Qa8 10.Kg1 Kf6 11.Kf1 Qa6+ 12.Kg1
  = (0.00)  Depth: 60/5   00:00:01  7396kN
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qd8 8.Ke1 Qh8 9.Kf1 Qa8 10.Kg1 Kf6 11.Kf1 Qa6+ 12.Kg1
  = (0.00)  Depth: 61/5   00:00:01  7438kN
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qd8 8.Ke1 Qh8 9.Kf1 Qa8 10.Kg1 Kf6 11.Kf1 Qa6+ 12.Kg1
  = (0.00)  Depth: 62/29   00:00:01  7761kN
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Kd7 8.Kg1
  = (0.00)  Depth: 63/49   00:00:02  8025kN
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qd8 8.Ke1 Qh8 9.Kf1
  = (0.00)  Depth: 64/15   00:00:02  8125kN
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qd8 8.Ke1 Qh8 9.Kf1
  = (0.00)  Depth: 65/5   00:00:02  8327kN
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qd8 8.Ke1 Qh8 9.Kf1
  = (0.00)  Depth: 66/5   00:00:02  8598kN
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qb8 8.Kg1 Kf6 9.Kf1 Qh8 10.Rg1 Qh6 11.Rh1
  = (0.00)  Depth: 67/19   00:00:02  9788kN
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qb8 8.Kg1 Kf6 9.Kf1 Qh8 10.Rg1 Qh6 11.Rh1
  = (0.00)  Depth: 68/5   00:00:02  10459kN
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qb8 8.Kg1 Kf6 9.Kf1 Qh8 10.Rg1 Qh6 11.Rh1
  = (0.00)  Depth: 69/13   00:00:02  10664kN
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qb8 8.Kg1 Kf6 9.Kf1 Qh8 10.Rg1 Qh6 11.Rh1
  = (0.00)  Depth: 70/17   00:00:02  11860kN
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qb8 8.Kg1 Kf6 9.Kf1 Qh8 10.Rg1 Qh6 11.Rh1
  = (0.00)  Depth: 71/5   00:00:02  12643kN, tb=4
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qc8 8.Ke1 Qe8 9.Kf1 Qa8 10.Kg1 Qh8 11.Kf1
  = (0.00)  Depth: 72/23   00:00:03  18398kN, tb=257
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qc8 8.Ke1 Qe8 9.Kf1 Qa8 10.Kg1 Kf5 11.Kf1
  = (0.00)  Depth: 73/21   00:00:03  19147kN, tb=406
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qc8 8.Ke1 Qe8 9.Kf1 Qa8 10.Kg1 Kf5 11.Kf1
  = (0.00)  Depth: 74/5   00:00:03  19350kN, tb=428
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qh8 8.Ke1 Qe8 9.Kf1
  = (0.00)  Depth: 75/17   00:00:04  21288kN, tb=530
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qd8 8.Ke1 Qa8
  = (0.00)  Depth: 76/8   00:00:16  99708kN, tb=39029
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qd8 8.Ke1 Qa8
  = (0.00)  Depth: 77/5   00:00:16  99709kN, tb=39029
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qd8 8.Ke1 Qe8 9.Kf1
  = (0.00)  Depth: 78/17   00:00:16  99719kN, tb=39029
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qd8 8.Ke1 Qe8 9.Kf1
  = (0.00)  Depth: 79/5   00:00:16  99720kN, tb=39029
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qg8 8.Ke1 Qe8 9.Kf1
  = (0.00)  Depth: 80/17   00:00:16  99744kN, tb=39029
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qg8 8.Ke1 Qe8 9.Kf1 Kf7 10.Kg1 Qa8 11.Kf1 Qd8 12.Kg1
  = (0.00)  Depth: 81/17   00:00:16  99754kN, tb=39029
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qg8 8.Ke1 Qe8 9.Kf1 Kf7 10.Kg1 Qa8 11.Kf1 Qd8 12.Kg1 Qe8 13.Kf1 Qg8
  = (0.00)  Depth: 82/20   00:00:16  100MN, tb=39029
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qg8 8.Ke1 Qe8 9.Kf1 Kf7 10.Kg1 Qa8 11.Kf1 Qd8 12.Kg1 Qe8 13.Kf1 Qg8
  = (0.00)  Depth: 83/5   00:00:16  101MN, tb=39029
5...Qa6+ 6.Kg1 Qa8 7.Kf1 Qd8 8.Ke1 Qe8 9.Kf1 Kf7 10.Kg1 Qa8 11.Kf1 Qd8 12.Kg1
  = (0.00)  Depth: 84/17   00:00:16  101MN, tb=39029

(Grayson, Newport, South Wales 03.04.2019)
Attachment: Q9550blockadeanalysis.txt (16k)
Parent - - By MaLaoshi (*) Date 2019-04-04 08:21
Interesting; so your observation is that throwing more hardware at these test positions made SF 31 Mar behave worse? My own runs were with 28 threads, 96GB hash, so if there is a scaling problem I'd have probably been on the bad side of it. Normally SF scales so well, I didn't think to test with less cores and memory. How is it for you with, say, SF 24 march? Similar issues or all the scaling trouble seems to come from this anti-shuffle patch?
Parent - By Peter Grayson (****) Date 2019-04-04 11:18
It certainly seems to indicate that way.

In the second position, again running with 16 cores and 4 Gb hash, the Stockfish 31031964 engne seems to get stuck around the 28 ply depth mark when the noticeable difference is the secondary search reflects just 53 ply depth (28/53) whereas on the Quad it finds the draw at just 17 primary ply depth but 103 secondary search depth (17/103).

On the dual Xeon machine I gave up on the 16 core search on position two when it had not moved from its output at 5..Kf5 28/53 00:39 1734MN because it had not changed after 12 minutes! Reducing hash to 512 Mb caused it to get stuck at 25 ply. Resetting the engine allocated number of cores to 4 and retaining 512 Mb hash the engine found the draw at 18/103 ply in 00:00 and was displaying 99/13 ply at just 27 seconds therefore it tends to confirm issues may arise when more than one CPU.

Peter
Parent - By Peter Grayson (****) Date 2019-04-04 11:25 Upvotes 1
I forgot to add that in the first position, remove the rooks and  with just the bishop a draw is shown straight away. Add in one rook and the draw is no longer seen.

Peter
Parent - - By MrKris (**) Date 2019-04-03 22:20
Latest Stockfish, including the "shuffle detection" patch.

Stockfish_19033111 modern x64, AMD PhII X6 6-cpu, Sy6, 256 MB hash:
{position as in the games}
51/53 plies, 18:48 (mins:secs), +14.67 (-+ eval.) 6...Ra8 7.Ke3 (etc.)

Both: Game/14 mins. + 1 sec./move
Both: A huge 3x deeper plies for CF, it is not that fast.

As black the new Stockfish offered a rook many times beginning with 24...Rb5, as white the old CFish never accepted.

[Event "3.5.1"]
[Site "Arena"]
[Date "2019.04.03"]
[Round "?"]
[White "CFish_10"]
[Black "Stockfish_19033111"]
[Result "1/2-1/2"]
[BlackElo "2000"]
[Time "11:00:45"]
[WhiteElo "2200"]
[TimeControl "840+1"]
[SetUp "1"]
[FEN "3k4/1r2p3/r2pPp2/b1pP1Pp1/1pP3Pp/pP5P/P3K3/8 b - - 0 6"]
[Termination "normal"]
[PlyCount "100"]
[WhiteType "program"]
[BlackType "program"]

{FEN
3k4/1r2p3/r2pPp2/b1pP1Pp1/1pP3Pp/pP5P/P3K3/8 b - - 0 6
} 6. ... Kc8 {+15.98/24 19} 7. Kf2 {-17.95/57 26} Ra8 {+15.84/32 44} 8. Ke1
{-17.95/60 11} Kc7 {+16.73/24 30} 9. Kd1 {-17.95/56 7} Rbb8 {+16.55/16 41}
10. Kc1 {-17.95/60 7} Rb6 {+15.92/18 46} 11. Kc2 {-17.95/59 8} Rbb8
{+14.69/24 37} 12. Kb1 {-17.95/64 30} Rg8 {+15.64/18 68} 13. Ka1 {-17.95/61
9} Bb6 {+15.35/16 59} 14. Kb1 {-17.92/1 0} Rgb8 {+15.19/20 46} 15. Ka1
{-17.95/64 23} Rf8 {+14.69/16 51} 16. Kb1 {-17.90/1} Rg8 {+14.53/18 34} 17.
Kc1 {-17.95/64 19} Rh8 {+14.34/21 10} 18. Kb1 {-17.95/62 19} Rhb8
{+14.64/21 31} 19. Ka1 {-17.95/64 10} Ra6 {+14.31/18 39} 20. Kb1 {-17.92/1}
Ra7 {+14.51/19 25} 21. Kc1 {-17.95/65 8} Raa8 {+14.92/19 58} 22. Kb1
{-17.95/66 19} Kb7 {+14.51/17 47} 23. Kc1 {-17.95/63 10} Ra5 {+15.29/18 6}
24. Kb1 {-17.95/63 42} Rb5 {+15.64/23 8} 25. Kc2 {-17.95/61 31} Ba5
{+15.29/26 11} 26. Kd3 {-17.95/58 25} Ka6 {+15.23/23 10} 27. Kd2 {-17.95/56
12} Kb7 {+15.25/23 3} 28. Kd1 {-17.95/58 12} Rc8 {+15.21/18 38} 29. Kc2
{-17.95/55 24} Rh8 {+15.38/18 10} 30. Kb1 {-17.95/53 45} Ra8 {+15.11/19 29}
31. Kc1 {-17.95/50 25} Re8 {+15.15/22 4} 32. Kb1 {-17.95/48 33} Bb6
{+14.89/19 21} 33. Kc1 {-17.95/44 8} Kc7 {+15.15/17 2} 34. Kc2 {-17.95/45
31} Ba5 {+14.98/20 15} 35. Kc1 {-17.95/43 7} Rf8 {+14.99/17 3} 36. Kb1
{-17.95/42 20} Rb6 {+15.18/19 7} 37. Ka1 {-17.95/36 40} Kb7 {+15.08/21 4}
38. Kb1 {-17.83/1} Ra6 {+15.01/16 2} 39. Kc1 {-17.95/34 22} Rh8 {+14.49/15
1} 40. Kb1 {-17.95/31 10} Ra7 {+14.86/18 0} 41. Kc2 {-17.95/30 81} Bb6
{+15.01/17 0} 42. Kb1 {-17.95/28 27} Rc8 {+15.23/17 0} 43. Ka1 {-17.95/27
54} Ra5 {+15.07/16 0} 44. Kb1 {-17.93/1} Kc7 {+15.08/19 1} 45. Ka1
{-17.95/21 22} Rb5 {+15.00/18 0} 46. Kb1 {-17.95/20 34} Ba5 {+14.84/18 1}
47. Kc1 {-17.86/16 5} Rb6 {+14.97/15 0} 48. Kb1 {-17.92/14 4} Rd8
{+15.12/15 0} 49. Kc1 {-17.95/15 1} Rb5 {+13.24/16 0} 50. Kc2 {-17.92/17 4}
Rdb8 {+9.91/20 1} 51. Kb1 {0.00/21 1} R8b6 {0.00/44 0} 52. Kc1 {0.00/53 3}
Ra6 {0.00/51 0} 53. Kd1 {0.00/69 2} Bb6 {0.00/69 1} 54. Kc2 {0.00/86 1} Ba5
{0.00/85 0} 55. Kd1 {0.00/97 1} Bb6 {0.00/100 0} 56. Ke1 {0.00/105 4 50
moves rule} 1/2-1/2


The new Stockfish LOST AS WHITE!
As black the old CFish had offered a rook several times, then the new Stockfish as white could not resist any longer and finally accepted 46.exd7??.

[Event "3.5.1"]
[Site "Arena"]
[Date "2019.04.03"]
[Round "?"]
[White "Stockfish_19033111"]
[Black "CFish_10"]
[Result "0-1"]
[BlackElo "2200"]
[Time "14:26:02"]
[WhiteElo "2000"]
[TimeControl "840+1"]
[SetUp "1"]
[FEN "3k4/1r2p3/r2pPp2/b1pP1Pp1/1pP3Pp/pP5P/P3K3/8 b - - 0 6"]
[Termination "normal"]
[PlyCount "113"]
[WhiteType "program"]
[BlackType "program"]

{FEN
3k4/1r2p3/r2pPp2/b1pP1Pp1/1pP3Pp/pP5P/P3K3/8 b - - 0 6
} 6. ... Ke8 {+18.24/58 28} 7. Kf1 {-13.86/23 53} Kf8 {+18.24/58 14} 8. Kg2
{-13.76/23 16} Kg7 {+18.24/59 12} 9. Kf2 {-13.56/21 13} Bb6 {+18.24/60 8}
10. Kg1 {-13.43/22 9} Ba5 {+18.24/63 24} 11. Kf1 {-14.18/25 65} Bb6
{+18.24/63 28} 12. Kg2 {-14.21/18 55} Rb8 {+18.24/64 10} 13. Kg1 {-14.19/17
104} Ra7 {+18.24/63 8} 14. Kg2 {-13.89/19 26} Raa8 {+18.24/61 80} 15. Kh2
{-14.10/20 21} Rd8 {+18.24/61 20} 16. Kg2 {-14.03/20 19} Ba5 {+18.24/64 85}
17. Kh2 {-14.37/22 67} Bb6 {+18.24/62 10} 18. Kg2 {-14.21/18 15} Rab8
{+18.24/65 58} 19. Kf1 {-14.16/19 7} Re8 {+18.24/67 15} 20. Kg2 {-14.39/22
48} Rbd8 {+18.24/65 5} 21. Kg1 {-13.92/18 6} Rb8 {+18.24/61 5} 22. Kf2
{-13.37/21 20} Red8 {+18.24/65 25} 23. Kg1 {-14.15/20 41} Rbc8 {+18.24/64
10} 24. Kg2 {-13.61/17 9} Ba5 {+18.24/64 76} 25. Kf2 {-14.18/20 20} Bb6
{+18.24/62 8} 26. Kf1 {-13.61/17 12} Rc6 {+18.24/60 11} 27. Kf2 {-13.33/20
9} Ba5 {+18.24/58 14} 28. Ke2 {-14.21/22 19} Bb6 {+18.24/55 6} 29. Ke3
{-13.82/18 8} Rcc8 {+18.24/54 10} 30. Kf3 {-13.64/18 6} Rc6 {+18.24/53 6}
31. Kg2 {-14.49/27 10} Rf8 {+18.24/50 6} 32. Kf3 {-14.39/17 53} Rfc8
{+18.24/49 11} 33. Ke2 {-14.47/20 10} Ra8 {+18.24/46 5} 34. Kf3 {-14.70/22
28} Rd8 {+18.24/43 59} 35. Kg2 {-14.53/16 6} Rc7 {+18.24/41 5} 36. Kf2
{-14.59/18 9} Ba5 {+18.24/38 44} 37. Kg2 {-14.28/18 2} Kg8 {+18.24/37 17}
38. Kg1 {-14.59/20 8} Bb6 {+18.24/35 45} 39. Kg2 {-14.46/18 2} Ba5
{+18.24/33 15} 40. Kf2 {-14.54/22 7} Bb6 {+18.24/30 7} 41. Ke2 {-14.50/24
25} Ba5 {+18.24/29 28} 42. Kf2 {-14.50/20 12} Kg7 {+18.24/26 20} 43. Kg2
{-14.51/17 10} Rb7 {+18.24/27 12} 44. Kf3 {-14.36/15 2} Bb6 {+18.24/23 7}
45. Kg2 {-14.35/16 4} Rdd7 {+18.24/20 4} 46. exd7 {-11.16/20 2} Rxd7
{+18.25/30 1} 47. Kf1 {-11.19/20 1} Rd8 {+18.93/26 1} 48. Ke2 {-74.47/34 2}
e5 {+25.49/25 0} 49. fxe6 {-M16/42 2} Rf8 {+M16/32 0} 50. Kf2 {-M14/43 0}
Bd8 {+M14/37 0} 51. Kg2 {-M13/41 0} f5 {+M13/39 0} 52. gxf5 {-M11/44 0}
Rxf5 {+M12/38 0} 53. Kg1 {-M10/47 0} Be7 {+M10/43 1} 54. Kg2 {-M10/45 0}
Kh6 {+M9/46 0} 55. Kg1 {-M8/53 0} Kh5 {+M8/52 0} 56. Kg2 {-M7/75 0} g4
{+M7/65 0} 57. Kg1 {-M6/60 0} gxh3 {+M6/105 0} 58. Kh1 {-M5/127 0} Rf2
{+M5/127 0} 59. Kg1 {-12.23/1} Rxa2 {+M4/127 0} 60. Kh1 {-M3/127 0} Rb2
{+M3/127 0} 61. Kg1 {-12.96/1} a2 {+M2/127 0} 62. Kf1 {-M1/127 0} a1=Q#
{+M1/127 0} 0-1
Parent - By MaLaoshi (*) Date 2019-04-04 09:35 Upvotes 3
Just throw the game?! That settles it then, the latest Fish is buggy. Still useful to check that in these and other cases, SF 24 Mar behaves as it should, i.e., that the problems are due to this one specific patch. I see that on Fishtest, Marco Costalba is testing some modifications to shuffle detection; can somebody point him to this discussion? Not sure if he's seen it, and as I said before I can't log into Fishcooking atm.
- - By gsgs (***) Date 2019-04-04 09:14 Upvotes 1
would StockFish be stronger, if we just replace the 50-move-rule by a 15(20?)-move-rule ?
(except tablebases)
In practice positions where you need more than 15 moves to progress should be rare.

well, tell StockFish that only the opponent, not sf,  can claim for draw after 15 moves
Parent - - By MaLaoshi (*) Date 2019-04-04 09:46 Upvotes 1
Yes that's what Houdini did years ago (just like Houdini was ahead of its time with "smart fail-high"). Or rather, Houdini had it as a user-controllable option which would be my favorite. The SF 31 Mar patch has this idea but without user control, but then it seems to have been combined with some extension logic that may not be essential, and doesn't work flawlessly yet. I think it's worth to pursue this, since fortresses etc are such a noticeable remaining weakness of SF.
Parent - By gsgs (***) Date 2019-04-04 09:53 Upvotes 1
also this shuffling was sooo tedious to watch in live online games like TCEC
Parent - By MrKris (**) Date 2019-04-05 03:20
In this game both reached 0.00 at move 157 only after there was just 1 pawn remaining, and it was blocked by a black bishop.
AMD PhII X6 6-cpu, Syzygy5&6, 256 MB hash

[Event ""]
[Site ""]
[Date "2019.04.04"]
[Round "?"]
[White "Stockfish_19032417"]
[Black "Stockfish_19033111 w/ shuffle detect"]
[Result "1/2-1/2"]
[BlackElo "2000"]
[Time "17:59:26"]
[WhiteElo "2200"]
[TimeControl "840+1"]
[SetUp "1"]
[FEN "k4Bb1/6P1/5P2/4P3/3P4/2P1K3/bP6/2B5 w - - 0 5"]
[Termination "adjudication"]
[PlyCount "326"]
[WhiteType "program"]
[BlackType "program"]

5. d5 {+10.02/49 16} Baxd5 {-9.80/48 15} 6. Kd4 {+10.02/52 9} Kb8 {-9.80/51
9} 7. b4 {+10.02/48 7} Kc7 {-9.80/52 7} 8. Kc5 {+10.02/48 9} Bc4 {-9.80/53
8} 9. b5 {+10.02/45 7} Bge6 {-9.80/53 9} 10. Be7 {+10.02/47 7} Bd3
{-9.80/48 8} 11. b6+ {+10.02/47 7} Kd7 {-9.79/44 7} 12. b7 {+10.02/54 9}
Kc7 {-8.52/45 7} 13. Bd8+ {+10.02/55 8} Kxb7 {-8.50/44 9} 14. Kd6
{+10.01/54 8} Bdc4 {-8.50/46 9} 15. Bc7 {+10.01/58 8} Kc8 {-8.49/46 14} 16.
Bb6 {+10.01/62 8} Kb7 {-8.49/49 8} 17. Bce3 {+10.01/61 9} Kc8 {-8.47/45 11}
18. Bg5 {+10.01/61 8} Bf7 {-8.39/29 8} 19. Bh4 {+10.01/62 9} Bfe6 {-8.39/39
9} 20. Bc7 {+10.01/64 12} Bf7 {-8.39/30 9} 21. Bg3 {+10.01/66 12} Bfe6
{-8.36/33 8} 22. Bh2 {+10.01/65 9} Bf7 {-8.23/32 8} 23. Bb6 {+10.01/66 11}
Bfe6 {-8.22/28 8} 24. Ba5 {+10.01/66 9} Bcd5 {-8.15/29 13} 25. Bc7
{+10.01/57 15} Bc4 {-8.08/35 10} 26. Bf4 {+10.01/58 10} Bf7 {-8.07/30 9}
27. Bb6 {+10.01/60 9} Bfe6 {-7.98/39 9} 28. Bfe3 {+10.01/62 12} Bed5
{-7.98/49 11} 29. Bc7 {+10.01/66 12} Bf7 {-7.97/53 12} 30. Ba5 {+10.01/62
9} Bfd5 {-7.97/51 10} 31. Bab6 {+10.01/63 36} Bg8 {-7.96/51 10} 32. Ke7
{+10.01/56 10} Bgf7 {-7.96/52 10} 33. Bf2 {+10.01/59 9} Bfd5 {-7.95/53 11}
34. Bg1 {+10.01/59 144} Bf7 {-7.95/50 10} 35. Bge3 {+10.01/56 18} Kb7
{-7.95/53 10} 36. Kd7 {+9.96/52 108} Bfe6+ {-7.94/53 12} 37. Kd6 {+9.82/53
45} Kc8 {-7.94/54 9} 38. Ba5 {+10.00/48 5} Bf7 {-7.93/50 9} 39. Bf2
{+9.96/48 60} Bfe6 {-7.93/54 15} 40. Kc6 {+10.01/43 5} Bcd5+ {-7.92/51 17}
41. Kd6 {+8.53/46 38} Bc4 {-7.92/52 23} 42. Bg3 {+8.49/35 6} Bg8 {-7.92/49
9} 43. Bh2 {+9.83/32 9} Bgf7 {-7.92/48 31} 44. Bf4 {+9.82/36 21} Bfe6
{-7.91/48 24} 45. Bh2 {+9.83/30 4} Bg8 {-7.91/44 11} 46. Bg3 {+8.15/50 10}
Bgd5 {-7.82/43 30} 47. Be1 {+9.97/40 4} Bg8 {-7.82/41 12} 48. Bc7 {+8.15/49
7} Bge6 {-7.82/40 18} 49. Bg3 {+8.15/44 2} Bf7 {-7.48/36 15} 50. Be1
{+8.15/47 2} Bfe6 {-7.11/29 11} 51. Bb6 {+8.25/50 2} Bf7 {-6.78/34 19} 52.
Bg3 {+8.15/50 3} Bfe6 {-6.64/28 5} 53. Bh2 {+8.15/48 2} Bcd5 {-6.48/33 10}
54. Bf4 {+9.13/45 2} Bc4 {-6.35/28 6} 55. Bfe3 {+8.15/53 5} Kb7 {-6.33/31
19} 56. Bf2 {+8.15/48 2} Kc8 {-6.24/31 24} 57. Kc6 {+8.15/49 2} Bcd5+
{-5.70/36 13} 58. Kd6 {+8.15/52 1} Bc4 {-5.70/41 3} 59. Ba5 {+8.15/52 2}
Bf7 {-5.70/46 5} 60. Bc7 {+8.15/50 2} Bfe6 {-5.70/40 7} 61. Ba7 {+8.15/51
2} Kb7 {-3.79/37 5} 62. Bg1 {+4.66/43 5} Kc8 {-3.78/46 4} 63. f7 {+4.66/44
1} Bxf7 {-3.78/40 4} 64. Ba5 {+4.66/43 2} Kb7 {-3.78/48 4} 65. Bgb6
{+4.66/43 1} Bfe6 {-3.77/43 4} 66. Bc7 {+4.66/45 1} Kc8 {-3.77/49 5} 67.
Bb4 {+4.66/43 1} Ba2 {-3.77/42 4} 68. Bca5 {+4.66/46 1} Kb7 {-3.77/46 11}
69. Bd8 {+4.66/47 2} Bac4 {-3.77/45 4} 70. Bba5 {+4.66/45 1} Bcd5 {-3.77/48
3} 71. Bab6 {+4.66/46 1} Bc4 {-3.77/50 4} 72. Bd4 {+4.66/46 2} Ka6
{-3.77/40 3} 73. Kc5 {+4.66/46 1} Kb7 {-3.77/41 3} 74. Kd6 {+4.66/48 2} Ka6
{-3.76/47 3} 75. Kc6 {+4.66/50 2} Bed5+ {-3.76/48 4} 76. Kc5 {+4.66/49 1}
Kb7 {-3.77/45 6} 77. Ba5 {+4.66/47 1} Be6 {-3.76/47 3} 78. Bb6 {+4.66/47 1}
Ka6 {-3.76/43 3} 79. Be3 {+4.66/49 2} Kb7 {-3.76/41 3} 80. Bd4 {+4.66/49 2}
Ka6 {-3.76/44 3} 81. Bc7 {+4.66/49 1} Kb7 {-3.76/47 3} 82. Kd6 {+4.66/49 1}
Kc8 {-3.76/36 3} 83. Ba5 {+4.66/47 2} Kb7 {-3.76/48 3} 84. Bab6 {+4.66/45
1} Kc8 {-3.76/47 4} 85. Ke7 {+4.66/47 1} Kb7 {-3.76/45 3} 86. Kd6 {+4.66/50
1} Kc8 {-3.75/45 3} 87. Bc7 {+4.66/47 1} Bf5 {-3.75/45 3} 88. Bdb6
{+4.66/46 1} Bfe6 {-3.75/47 3} 89. Bd8 {+4.66/44 1} Bg4 {-3.75/42 3} 90.
Bf6 {+4.66/44 1} Bge6 {-3.75/42 4} 91. Bf2 {+4.66/46 1} Bf5 {-3.75/44 5}
92. Bb6 {+4.66/46 1} Kb7 {-3.75/44 2} 93. Bc7 {+4.66/42 5} Kc8 {-3.75/44 3}
94. Bcd8 {+4.66/41 10} Bfe6 {-3.75/40 3} 95. Ba5 {+4.66/39 3} Bf5 {-3.74/37
3} 96. Bb4 {+4.66/35 0} Bfe6 {-3.74/33 3} 97. Be7 {+4.66/33 1} Bf5
{-3.73/28 2} 98. Ba5 {+4.66/33 4} Bfe6 {-3.71/26 2} 99. Bb6 {+4.66/32 4}
Kb7 {-3.69/29 2} 100. Kc5 {+4.66/29 2} Kc8 {-3.65/29 2} 101. Kd6 {+4.66/25
0} Kb7 {-3.56/34 2} 102. Bc7 {+4.66/27 1} Kc8 {-3.46/31 2} 103. Bh4
{+4.56/22 0} Kb7 {-3.06/29 6} 104. Bf2 {+4.46/22 1} Kc8 {-3.06/27 1} 105.
Be3 {+4.57/19 0} Kb7 {-2.96/26 2} 106. Bd8 {+3.94/25 1} Kc8 {-2.79/26 2}
107. Beg5 {+3.92/28 0} Bf5 {-2.36/27 6} 108. Bb6 {+3.92/30 0} Bfe6
{-2.23/26 5} 109. Bh6 {+3.09/33 2} Bh3 {-2.56/23 2} 110. e6 {+3.09/29 0}
Bhxe6 {-1.91/23 1} 111. Ba5 {+3.09/32 0} Bg4 {-1.91/24 8} 112. Ke7
{+3.09/32 0} Kb7 {-1.90/18 1} 113. Kd6 {+3.09/37 0} Bge6 {-1.75/18 3} 114.
Bd8 {+3.09/36 0} Kc8 {-1.72/18 1} 115. Bhg5 {+3.09/36 0} Bh3 {-1.71/18 1}
116. Bge7 {+3.09/37 0} Bg4 {-1.65/21 1} 117. Bh4 {+3.09/37 0} Bh3 {-1.72/21
6} 118. Bb6 {+3.09/37 0} Bf5 {-1.53/18 2} 119. Bd4 {+3.09/36 0} Bfe6
{-1.54/20 6} 120. Bdf2 {+3.09/37 0} Bd7 {-1.43/19 1} 121. Bc5 {+3.09/39 0}
Bf5 {-1.56/22 5} 122. Ba3 {+3.09/34 0} Bfe6 {-1.46/20 0} 123. Bf6 {+3.09/38
0} Bg4 {-1.53/22 2} 124. Be7 {+3.09/39 0} Bf5 {-1.58/20 4} 125. Bc5
{+3.09/39 0} Bd7 {-1.42/19 4} 126. Bb6 {+3.09/39 0} Bde6 {-1.33/21 2} 127.
Bc5 {+3.09/41 0} Bd7 {-1.33/17 2} 128. Bg1 {+3.09/39 0} Bh3 {-1.31/18 1}
129. Bf6 {+3.09/39 0} Bf5 {-1.15/21 2} 130. Kc5 {+3.09/39 0} Bfe6 {-1.08/17
0} 131. Kc6 {+3.09/41 0} Bh3 {-1.07/20 2} 132. Be7 {+3.09/40 0} Bg4
{-1.13/24 5} 133. Kc5 {+3.09/42 0} Bge2 {-1.14/18 0} 134. Kc6 {+3.09/41 0}
Bg4 {-1.09/21 2} 135. Kd6 {+3.09/41 0} Bf5 {-1.07/20 3} 136. Kc5 {+3.09/42
0} Bfe6 {-1.02/18 1} 137. Kd6 {+3.09/44 1} Bh3 {-1.17/19 4} 138. Bb6
{+3.09/40 0} Bhe6 {-1.02/20 1} 139. Bf6 {+3.09/40 1} Bh3 {-1.04/20 1} 140.
Kc5 {+3.09/40 1} Bhe6 {-1.02/18 0} 141. Kc6 {+3.09/40 6} Bh3 {-1.06/20 1}
142. Be7 {+3.09/38 3} Bf5 {-1.02/20 1} 143. Kc5 {+3.09/36 1} Bfe6 {-0.98/18
0} 144. Kd6 {+3.09/33 1} Bh3 {-1.10/22 4} 145. Kc5 {+3.09/30 0} Bhe6
{-1.16/19 0} 146. Kd6 {+3.09/29 0} Bg4 {-1.05/20 1} 147. Be3 {+3.09/28 0}
Bh3 {-1.00/19 0} 148. Kc5 {+3.09/25 1} Bhe6 {-0.75/27 0} 149. Kc6 {+3.09/24
0} Bd7+ {-0.72/24 0} 150. Kc5 {+1.06/21 1} Bde6 {-0.71/23 0} 151. Bf6
{+0.75/20 0} Kd7 {-0.69/23 0} 152. Be5 {+0.32/22 0} Ke7 {-0.13/23 2} 153.
Kd4 {0.00/29 0} Kf7 {0.00/30 0} 154. g8=Q+ {0.00/32 0} Kxg8 {0.00/36 0}
155. Bc1 {0.00/35 0} Kf7 {0.00/40 0} 156. Bb2 {0.00/41 0} Kg6 {0.00/37 0}
157. Ba1 {0.00/42 0} Kf7 {0.00/40 0} 158. Bb2 {0.00/46 0} Ke7 {0.00/43 0}
159. Ba1 {0.00/41 0} Bd3 {0.00/50 0} 160. Bh2 {0.00/44 0} Kf7 {0.00/48 0}
161. Ke3 {0.00/45 0} Bd5 {0.00/48 0} 162. Kd4 {0.00/43 0} Be6 {0.00/54 0}
163. Bg3 {0.00/45 0} Bdc4 {0.00/47 0} 164. Bb2 {0.00/45 0} Kf6 {0.00/37 0}
165. Ba1 {0.00/46 0} Kf7 {0.00/48 0} 166. Bb2 {0.00/48 0} Ke7 {0.00/38 0}
167. Bf4 {0.00/42 0} Be2 {0.00/38 0  User Adjudication} 1/2-1/2


From above game, 63. f7 {+4.66/44 1} Bxf7 {-3.78/40 4}, analysis
FEN:
2k5/2B2bP1/3K4/4P3/2b5/2P5/8/6B1 w - - 0 64

Houdini 6.03 standard ,  Contempt -2 , FiftyMoveDistance 10 :
('hashfull' if in a game might help also, but I did not try it)
 12/23  00:00   1,334k  11,596k  +1.69  64.Ba5 Bg8 65.e6 Bcxe6 66.c4 Bxc4 67.Bh2 Kb7 68.Bb4 Bce6 69.Bg3 Bd5 70.Ke7
...
 20/33  00:01   19,120k  12,504k  +0.82  64.Be3 Bfe6 65.Bc1 Kb7 66.Bb2 Kc8 67.Bb6 Kb7 68.Bg1 Kc8 69.Bd4 Kd8 70.Bc1 Kc8 71.Ba3 Kb7 72.Bb4 Kb8 73.g8R+ Bxg8 74.Ba5 Kb7 75.Be3 Bge6 76.Bd8
 21/36  00:02   30,021k  12,704k  +0.72  64.Be3 Bfe6 65.Ba5 Kb7 66.Bd8 Kc8 67.Bf6 Kb8 68.Bg1 Kb7 69.Bc5 Kc8 70.Bh4 Kb8 71.Bg5 Kc8 72.Ba7 Kb7 73.g8R Bxg8 74.Bf2 Kc8 75.Bf6
...
 31/43  00:57   825,818k  14,432k  +0.11  64.Be3 Bfe6 65.Bf2 Bg8 66.Bh4 Bgf7 67.Bcd8 Bfe6 68.Bb6 Kb8 69.Ba5 Kb7 70.Be1 Kb8 71.Bf2 Kb7 72.Bd8 Kc8 73.g8N Bxg8 74.Ba5 Bge6 75.Ba7 Kb7 76.Bg1 Bg8 77.Be3 Ka6 78.Bc7 Bge6 79.Bf2 Kb5 80.Bcb6 Ka4
 31/43  00:58   839,477k  14,428k  +0.11  64.Be3 Bfe6 65.Bf2 Bg8 66.Bh4 Bgf7 67.Bcd8 Bfe6 68.Bb6 Kb8 69.Ba5 Kb7 70.Be1 Kb8 71.Bf2 Kb7 72.Bd8 Kc8 73.g8N Bxg8 74.Ba5 Bge6 75.Ba7 Kb7 76.Bg1 Bg8 77.Be3 Ka6 78.Bc7 Bge6 79.Bf2 Kb5 80.Bcb6 Ka4
 32/44-  01:20   1,158,508k  14,426k  +0.04  64.Be3 Bfe6
 32/45  03:41   3,222,580k  14,539k  +0.01  64.Be3 Bfe6 65.Bf2 Bg8 66.Bh4 Bgf7 67.Bcd8 Bfe6 68.Kc6 Bcd5+ 69.Kc5 Bc4 70.Kb6 Kd7 71.Bdf6 Kc8 72.Bg3 Kd7 73.g8R Bxg8 74.Kc5 Bge6 75.Kd4 Kc6 76.Ke3 Kd7 77.Kf4 Kc6 78.Kg5 Kd7 79.Kh6 Kc6 80.Bf4 Kd5 81.Kg5
 32/45  03:51   3,355,146k  14,484k  +0.01  64.Be3 Bfe6 65.Bf2 Bg8 66.Bh4 Bgf7 67.Bcd8 Bfe6 68.Kc6 Bcd5+ 69.Kc5 Bc4 70.Kb6 Kd7 71.Bdf6 Kc8 72.Bg3 Kd7 73.g8R Bxg8 74.Kc5 Bge6 75.Kd4 Kc6 76.Ke3 Kd7 77.Kf4 Kc6 78.Kg5 Kd7 79.Kh6 Kc6 80.Bf4 Kd5 81.Kg5
 33/45  03:58   3,442,781k  14,437k   0.00  64.Be3 Bfe6 65.Bf2 Bg8 66.Bh4 Bgf7 67.Bcd8 Bfe6 68.Kc6 Bcd5+ 69.Kb5 Bc4+ 70.Kb4 Kd7 71.Bde7 Kc6 72.Bf2 Kd7 73.g8B Bxg8 74.Bd6 Kc6 75.Be7 Kd7
...
 36/47  04:25   3,787,216k  14,244k   0.00  64.Be3 Bfe6 65.Bf2 Bg8 66.Bh4 Bgf7 67.Bcd8 Bfe6 68.Kc6 Bcd5+ 69.Kb5 Bc4+ 70.Kb4 Kd7 71.Bde7 Kc6 72.Bf2 Kd7 73.g8B Bxg8 74.Bd6 Kc6 75.Be7 Kd7
- - By Peter Grayson (****) Date 2019-04-05 12:26
The Example 2 problem when the engine is run with more than 1 CPU appears to have been fixed. Just tested the last Stockfish 040419 64 POPCNT release and running with the full 2 CPUs and 16 cores it solves straight away.

Peter
Parent - - By MaLaoshi (*) Date 2019-04-09 07:46
Every SF version will behave a little differently (even the same Fish from run to run can give different results). But if I look on http://www.abrok.eu/stockfish, none of the commits after 31 March have touched this shuffle-detection logic. Indeed, in Example 1 above, I still get the same nonsensical PVs; as MrKris has shown, this may cause SF to outright lose this dead-drawn position. So in my opinion, SF is still broken. Looking on Fishtest, people are still working on this; so hopefully a fix is on the way.
Parent - By Peter Grayson (****) Date 2019-04-09 10:54
Thanks for update and detail.
Peter
Parent - By MrKris (**) Date 2019-04-11 04:07
Chinese Wall
4k3/8/8/B2p3p/p1pPp1pP/PpP1PpP1/1P3P2/K2N3R w - - 0 1


Houdini-6--50 Move Distance = 15 :
15/31  00:00   501k  4,812k  +4.67  1.Kb1 Kd7 2.Bb4 Ke6 3.Kc1 Kd7 4.Kd2 Kc6 5.Ke1 Kc7 6.Kf1 Kd7 7.Rh3 Kc6 8.Kg1 Kd7 9.Bf8 gxh3 10.Kh2 Ke6 11.Kxh3 Kf7
...
33/49-  00:04   25,913k  5,628k  +1.60  1.Kb1 Kd7
33/49-  00:07   45,260k  5,830k  +0.14  1.Kb1 Kd7
33/49  00:08   48,421k  5,863k   0.00  1.Kb1 Kd7 2.Re1 Kc6 3.Bb4 Kd7 4.Kc1 Kd8 5.Bd6 Kd7 6.Rh1 ...
...
67/115  06:04   3,071,806k  8,417k   0.00  1.Kb1 Kd7 2.Re1 Kc6 3.Bb4 Kd7 4.Kc1 Kd8 5.Bd6 Kd7 6.Rh1 ...

Stockfish_19041019:
1/1  00:00   10  5k  +14.79  1.Kb1
...
40/39+  00:18   80,764k  4,392k  +10.50  1.Kb1
40/48+  00:18   81,197k  4,394k  +10.59  1.Kb1
{stuck @ 40 ply: 40/51  06:04   1,546,686k  4,240k  +13.42  1.Kb1}

Stockfish_19032417:
Found 510 tablebases
1/1  00:00   10  5k  +14.61  1.Kb1
...
79/81  06:04   2,010,447k  5,511k  +14.83  1.Bb4 Kd7 2.Kb1 Kc6 3.Rf1 Kc7 4.Bc5 Kb7 5.Bd6 Kb6 6.Be5 Kb7 7.Re1 ...
Parent - - By MrKris (**) Date 2019-04-11 06:28
From the first post this thread, 2nd example - a black bishop added on h7 inadvertantly gives it a win:
(Similar to P.G.'s idea above for the 1st position, but here if there is a R instead on h7 then 5...Qa6+ 6.Kg1 Rh4!!) 

PV's: 2-cores each, AMD PhII X6, Sy.5+6 50-move off
- Houdini 6.03 with 50-move rule distance = 15: quickly finds the winning method for black.
- Stockfish_19041019: shuffles with black until it finds a way for white to lose by moving out of the stalemate.
- Stockfish_19032417: shuffles with black, white keeps the stalemate resource but drifts away, by then it is a 50-move rule draw in the PV but high black evaluation.


[Event "added ...Bh7"]
[Site "simultaneously 2-cores each"]
[Date "2019.04.10"]
[Round "?"]
[White ""]
[Black "Sf_190324,190410,H6-50d=15"]
[Result "*"]
[BlackElo "2000"]
[Time "21:36:14"]
[WhiteElo "2000"]
[TimeControl "360+1"]
[SetUp "1"]
[FEN "q7/1p5b/1Pp1k3/p1Pp4/P2Pp1p1/4PpPp/5P1P/5K1R b - - 0 5"]
[Termination "unterminated"]
[PlyCount "0"]
[WhiteType "program"]
[BlackType "human"]

{'q7/1p5b/1Pp1k3/p1Pp4/P2Pp1p1/4PpPp/5P1P/5K1R b - - 0 5'
Houdini 6 50moveDistance=15:  48/96, 10:19, +4.59 black winning method 14...Qxb6} 5. ... Qh8

( {Stockfish_19041019: 31/103, 03:59, +6.54 -- but black just shuffling until white destroys its stalemate, depth stalled at 32 ply below:} 5. ... Qd8 6.Kg1 Kd7 7.Kf1 Bg6 8.Rg1 Ke6 9.Rh1 Qg8 10.Ke1 Qc8 11.Kf1 Kd7 12.Kg1 Bf7 13.Kf1 Ke7 14.Ke1 Qe6 15.Kd1 Bg8 16.Ke1 Kf6 17.Kd2 Kf5 18.Kc2 Qf6 19.Kd2 Ke6 20.Rc1 Qh8 21.Ke1 Qh5 22.Kf1 Qe8 23.Kg1 Qa8 24.Ra1 Kd7 25.Re1 Qa6 26.Rc1 Be6 27.Kh1 Bf7 28.Kg1 Kc8 29.Rc3 Qd3 30.Rxd3 exd3 31.Kf1 )

( {Stockfish_19032417: 78/101, 12:46, +11.52 -- but only shuffling by black, but white mostly keeps the stalemate--the PV is a 50 move draw} 5. ... Qa6+ 6.Kg1 Qa8 7.Kf1 Kf7 8.Kg1 Qg8 9.Kf1 Bf5 10.Ke1 Kg7 11.Kf1 Bd7 12.Ke1 Qe8 13.Kf1 Qa8 14.Kg1 Qd8 15.Kf1 Qe8 16.Ke1 Kf7 17.Kf1 Qe6 18.Rg1 Kg8 19.Rh1 Qf5 20.Kg1 Be6 21.Kf1 Kf8 22.Kg1 Qf6 23.Kf1 Bf5 24.Kg1 Qg7 25.Kf1 Qf7 26.Kg1 Kg8 27.Kf1 Qd7 28.Kg1 Qe6 29.Kf1 Bg6 30.Rg1 Qe7 31.Rh1 Qd8 32.Kg1 Qc8 33.Kf1 Kg7 34.Kg1 Qd7 35.Kf1 Bf5 36.Ke1 Kh6 37.Kf1 Kg6 38.Kg1 Qf7 39.Kf1 Qf6 40.Ke1 Qg5 41.Kf1 Kf6 42.Kg1 Kf7 43.Kf1 Qh6 44.Ke1 Qh5 45.Kf1 Bg6 46.Kg1 Qh8 47.Kf1 Qd8 48.Kg1 Kg8 49.Kf1 Qe7 50.Ke1 Bh7 51.Kd2 Qd7 52.Kc2 Qd8 53.Rg1 Qa8 54.Kb1 Qa6 {50-move rule!?} )

6.Kg1 Kf5 7.Kf1 Qa8 8.Kg1 Kf6 9.Kf1 Ke7 10.Kg1 Kd7 11.Kf1 Bg6 12.Kg1 Be8 13.Kf1 Qa6+ 14.Kg1 Qxb6 15.cxb6 c5 16.dxc5 Kc6 17.Kf1 Kxc5 18.Ke1 Kxb6 19.Kd2 Bxa4 20.Rc1 Bc6 21.Ra1 Bb5 22.Kc3 Bc4 23.Rc1 Kc5 24.Ra1 b6 25.Kc2 Bd3+ 26.Kc3 Kb5 27.Ra3 Be2 28.Rb3+ Ka6 29.Rb2 Bd3 30.Kd4 Bc4 31.Rb1 a4 32.Kc3 b5 33.Kb4 d4 34.exd4 Kb6 {all 3 engines agree black is winning, ...e4-e3 is very strong}
*


Houdini 6 50d=15 first had the ...Qxb6 idea in only 7 seconds! The last long PV is shown as the main moves above.
20/53  00:07   39,217k  5,083k  +1.42  5. ...Bf5 6.Kg1 Ke7 7.Kf1 Kf7 8.Kg1 Bc8 9.Kf1 Bd7 10.Kg1 Be8 11.Kf1 Ke7 12.Kg1 Kd7 13.Kf1 Qa6+ 14.Kg1 Qxb6 ...

Stockfish_19041019 reached ply 32 in 4 minutes but could not get beyond that:
32/49+  03:59   930,902k  3,889k  +6.73  5. ... Qd8
32/49  12:46   3,111,398k  4,057k  +11.05  5. ... Qd8 << output when stopped

-
-
-
{P.S. Remove the ...Bh7 back to the original draw, done separately:
Okay- Stockfish_19041019:  22/28  00:00   3,715k  5,953k   0.00  5. ...Qf8 ...
Fail- Stockfish_19032417: 77/97  00:03   18,827k  5,924k  +7.49  5. ... Ke7 ...
Okay- Houdini-6--50moveD=15: 18/31  00:01   17,006k  12,863k   0.00  5. ... Kf5 -}
Parent - - By Akbarfan (***) Date 2019-04-12 12:33 Upvotes 1
Author: Marco Costalba
Date: Fri Apr 12 13:48:04 2019 +0200
Timestamp: 1555069684

Revert "Shuffle detection #2064"

It causes a serious regression hanging a simple fixed
depth search. Reproducible with:

position fen
q1B5/1P1q4/8/8/8/6R1/8/1K1k4 w - - 0 1

go depth 13

The reason is a search tree explosion due to:

if (... && depth < 3 * ONE_PLY)
extension = ONE_PLY;
Parent - - By MrKris (**) Date 2019-04-14 17:10
SP-CC found the problem in 1/40th the number of games as the pass on Fishnet:

https://www.sp-cc.de/index.htm comment is gone but it still shows on the graph.
" ... Stockfish 190301 caused a serious regression -9 Elo [Total 5,000 games]
Each Stockfish-version plays 1000 games (3'+1'') versus Komodo 12.3, Houdini 6, Fire 7.1, Ethereal 11.25, Komodo 12.3 MCTS. ... "

http://abrok.eu/stockfish/ passed at almost 1/4 million games vs. its previous self.
" Author: Moez Jellouli
Date: Sun Mar 31 10:51:08 2019 +0200
Timestamp: 1554022268
Shuffle detection #2064 ...
STC
LLR: -2.94 (-2.94,2.94) [0.50,4.50]
Total: 32595 W: 7273 L: 7275 D: 18047 Elo -0.02
http://tests.stockfishchess.org/tests/view/5c90aa330ebc5925cfff1768
LTC
LLR: 2.95 (-2.94,2.94) [0.00,3.50]
Total: 51249 W: 8807 L: 8486 D: 33956 Elo +2.18
http://tests.stockfishchess.org/tests/view/5c90b2450ebc5925cfff1800
VLTC
LLR: 2.96 (-2.94,2.94) [0.00,3.50]
Total: 137974 W: 20503 L: 19983 D: 97488 Elo +1.31
http://tests.stockfishchess.org/tests/view/5c9243a90ebc5925cfff2a93 "
Total Games: 221,818   Ave. Elo: +1.16
Parent - By MaLaoshi (*) Date 2019-04-19 09:28
Hmm, sp-cc is worth following but please note that their data points all have error bars of 7~8 Elo. So a -9 Elo regression is hardly significant, particularly if it follows immediately after a +10 Elo progression which was not expected from Fishtest's own regression tests over many more games.
- - By MaLaoshi (*) Date 2019-04-19 09:22
Update: this notorious patch was reverted on 12 Apr, and I just verified that SF 18 Apr 2019 works pretty much the same as the one of 24 Mar. Back to square one... So at least the bug is gone, but shuffling positions are no closer to being recognized. Still waiting for the very simple option of having a user-selectible "X move rule" (X <= 50), which Houdini had years and years ago, and which is effective in all these examples.
Parent - - By Peter Grayson (****) Date 2019-04-19 09:35

> So at least the bug is gone, but shuffling positions are no closer to being recognized.


With the shuffle patch some positions were scored correctly = 0.00 or very close to it, without the shuffle patch the positions score completely erroneously. The question is, how does this affect move selection and how early in the move tree? With the patch a correctly scored 0.00 position may be disregarded in favour of one with more promise but without the patch the erroneously scored position may be selected when the engine mistakenly identifies it as a target to achieve when all it can achieve is a draw.

If the positions are scored accurately it makes the engine more valuable as an analysis tool. I wonder how difficult it would be to find a compromise? Are the MCTS or LC0 engines better able to address this problem?

Peter
Parent - By MrKris (**) Date 2019-04-21 05:06

[Event "Premier, game 98"]
[Site "TCEC 15"]
[Date "2019.04.20"]
[Round "?"]
[White "Ethereal 11.38"]
[Black "LCZero v0.21.1-n41800"]
[Result "1/2-1/2"]
[BlackElo "2000"]
[Time "19:33:21"]
[WhiteElo "2000"]
[TimeControl "480+4"]
[SetUp "1"]
[FEN "8/6k1/6p1/8/7P/1B6/P2b3P/7K w - - 0 43"]
[Termination "normal"]
[PlyCount "0"]
[WhiteType "human"]
[BlackType "human"]

{'8/6k1/6p1/8/7P/1B6/P2b3P/7K w - - 0 43' - after 42...Bx(R)d2} 43. Kg2 {TCEC 1.52} Kh6 {TCEC 0.13} 44. Bd1 Be1 45. a4 Bxh4 46. a5 Bd8 47. a6 Bb6 48. Bg4 g5 49. Kf3 Kg6 50. Ke4 Bg1
51. Bf5+ Kf6 52. h3 Ba7 53. Bg4 Bg1 54. Kd5 Kf7 55. Bf5 Kf6 56. Bd7 Ke7 57. Kc6 Be3
58. Bg4 Kf7 59. Kb7 Kg6 60. Kc6 Bf2 61. Kd7 Bg1 62. Bd1 Bf2 63. Be2 {TCEC 1.52 still} Bg1
{TCEC 0.05 -- Lc0's first TCEC drawscore; draw on move 93} 1/2-1/2


https://en.wikipedia.org/wiki/Computer_chess
" ... 1956 – Los Alamos chess {6x6 board no bishops https://en.wikipedia.org/wiki/Los_Alamos_chess } is the first program to play a chess-like game, developed by Paul Stein and Mark Wells for the MANIAC I { http://www.ed-thelen.org/comp-hist/BRL61-m.html#MANIAC-I } computer.
1956 – John McCarthy invents the alpha-beta search algorithm.
1957 – The first programs that can play a full game of chess are developed, one by Alex Bernstein[3] and one by Russian programmers using a BESM {K memory, Kflops https://en.wikipedia.org/wiki/BESM }.
1958 – NSS becomes the first chess program to use the alpha-beta search algorithm. ... "

Here, AMD PhII X6, Sy.6, same position as above:
Hannibal1.4bx64 (1-cpu, no etb): FEN: 8/6k1/6p1/8/7P/1B6/P2b3P/7K w - - 0 43 
   8  00:00         1,391  92,733   0.00  43.a2-a3 g6-g5 44.h4xg5 Bd2xg5 45.h2-h3 Bg5-c1 46.a3-a4 Bc1-b2
..
  38  00:05    34,095,581  6,939,869   0.00  43.a2-a3 g6-g5 44.h4xg5 Bd2xg5 45.h2-h3 Bg5-c1 46.a3-a4 Bc1-b2 47.h3-h4 Bb2-c1 48.h4-h5 Bc1-d2 49.a4-a5 Bd2xa5 50.Kh1-g2 Ba5-d2 51.Bb3-c2 Bd2-c1 52.Bc2-d1 Bc1-d2 53.Bd1-c2 Bd2-c1


Stockfish_19041816: FEN: 8/6k1/6p1/8/7P/1B6/P2b3P/7K w - - 0 43
Found 510 tablebases, 65.6 million TBhits
 1/1  00:00   79  40k  +0.33  43.Kg2
 2/2  00:00   292  146k  +0.69  43.Kg2 Bc1
 3/3  00:00   4k  1,434k  +0.48  43.a4 Kh6 44.Bd1
...
 96/101  11:57   9,469,747k  13,202k  +0.87  43.Kg2 Ba5 44.a4 Bd8 45.Kh3 Kf6 46.Kg4 Bc7 47.Bc2 Kg7 48.Bd1 Kh6 49.Be2 Bd8 50.Kh3 Bb6 51.Bg4 Kg7 52.Kg2 Bc7 53.Bd1 Kf6 54.h3 Bb6 55.Be2 Kf5 56.Bd3+ Kf6 57.Kf3 Bc7 58.Be4 Ba5 59.Bc2 Bd8 60.Bb1 Ba5 61.Be4 Bd2 62.Kg4 Bc3 63.Bd3 Bd2 64.Bf1 Kg7 65.Ba6 Kf6 66.Bd3 Kg7 67.Bc2 Kf6 68.Kg3 Be1+ 69.Kf3 Ke5 70.Kg4 Kf6 71.Bb3 Ba5 72.Bd5 Be1 73.Bb7 Ba5 74.Kg3 Be1+ 75.Kf3 Kf5 76.Bc8+ Kf6 77.Kg4 Bc3 78.Bd7 Bd2 79.Kf3 Ba5 80.Ke2 Ke7 81.Bb5 Kf6 82.Bd3 Bc7 83.Kf3 Ba5 84.Kf4 Bd2+ 85.Ke4 Kg7 86.Bc2 Be1 87.Kf4 Bd2+ 88.Ke5 Kh6 89.Bd1 Ba5 90.Kf4 Bd8 91.Ke4 Kg7 92.Bg4 Kf7 93.Kf4
 97/101  12:45   10,128,227k  13,233k  +0.87  43.Kg2 Ba5 44.a4 Bd8 45.Kh3 Kf6 46.Kg4 Bc7 47.Bc2 Kg7 48.Bd1 Kh6 49.Be2 Bd8 50.Kh3 Bb6 51.Bg4 Kg7 52.Kg2 Bc7 53.Bd1 Kf6 54.h3 Bb6 55.Be2 Kf5 56.Bd3+ Kf6 57.Kf3 Bc7 58.Be4 Ba5 59.Bc2 Bd8 60.Bb1 Ba5 61.Be4 Bd2 62.Kg4 Bc3 63.Bd3 Bd2 64.Bf1 Kg7 65.Ba6 Kf6 66.Bd3 Kg7 67.Bc2 Kf6 68.Kg3 Be1+ 69.Kf3 Ke5 70.Kg4 Kf6 71.Bb3 Ba5 72.Bd5 Be1 73.Bb7 Ba5 74.Kg3 Be1+ 75.Kf3 Kf5 76.Bc8+ Kf6 77.Kg4 Bc3 78.Bd7 Bd2 79.Kf3 Ba5 80.Ke2 Ke7 81.Bb5 Kf6 82.Bd3 Bc7 83.Kf3 Ba5 84.Kf4 Bd2+ 85.Ke4 Kg7 86.Bc2 Be1 87.Kf4 Bd2+ 88.Ke5 Kh6 89.Bd1 Ba5 90.Kf4 Bd8 91.Ke4 Kg7 92.Bg4 Kf7 93.Kf4
Up Topic The Rybka Lounge / Computer Chess / Stockfish anti-shuffle patch of 31 March is still glitchy?

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill