Fafis quo vadis?

[Guenther Simon]

There is still no source code available for version 2.6,
after it was obviously available for some hours at first.
It vanished after Dann looked into it and noticed it was not
the actual source, but an already precompiled kind of source,
which is hardly readable.
http://f27.parsimony.net/forum67828/messages/2318.htm

I noticed that the new version 2.6 searches now 6-7 plies! deeper
than 2.52...
What is your opinion - is it safe to assume Fafis is clean?
It would be interesting, if someone would run a tourney with both
Fafis versions including a few other programs, because I have
no CPU time left.
I would like to see the approx. strength increase...

This all is not intended as an offense against Rafael, but IMHO
these are reasonable questions(after what happened in the past).

Guenther

[Event "RWBC_Qualification_6th_Edition"]
[Site "ESPRESSO"]
[Date "2005.10.02"]
[Round "3"]
[Number "78"]
[White "ChessAlex_12r5"]
[Black "Fafis_252"]
[Result "1-0"]
[TimeControl "40/900"]

1. Nf3 d5 2. c4 dxc4 3. Qa4+ {+0.10/7} Nd7 4. e4 {+0.15/7} e5 {+0.12/7} 5.
Bxc4 {+0.21/7} Bd6 {+0.20/7} 6. d3 {+0.29/7} Nf6 {-0.01/6} 7. Ng5 {+0.46/6}
O-O {-0.05/7} 8. Qb3 {+0.63/6} Qe7 {-0.53/7} 9. Nxf7 {+0.55/6}
Rxf7 {-0.05/7} 10. Be3 {+0.21/6} Nc5 {-0.14/7} 11. Bxc5 {+0.42/6}
Bxc5 {-0.23/7} 12. O-O {+0.14/7} Ng4 {-0.09/7} 13. Nc3 {+0.15/7}
Nh6 {-0.08/7} 14. Kh1 {+0.17/7} Ng4 {-0.56/7} 15. Rac1 {+0.32/7}
Kf8 {-0.17/7} 16. Bxf7 {+0.70/6} Qh4 {-0.08/7} 17. h3 {+0.33/7}
Nxf2+ {-0.02/6} 18. Rxf2 {+0.33/7} Qxf2 {-0.02/6} 19. Bh5 {+0.33/7}
Be3 {+0.06/6} 20. Rd1 {+0.32/6} Bc5 {-0.41/6} 21. Be2 {+1.31/6}
Bxh3 {-2.89/7} 22. gxh3 {+4.14/7} Qg3 {-5.79/7} 23. Rf1+ {+5.15/7}
Bf2 {-4.55/7} 24. Qe6 {+5.26/7} a5 {-6.08/6} 25. Nd5 {+6.11/6}
Re8 {-9.73/7} 26. Qf5+ {+8.57/7} Kg8 {-9.80/8} 27. Rxf2 {+9.35/6}
Qxf2 {-10.30/7} 28. Qxf2 {+9.43/7} c6 {-10.26/7} 29. Nc7 {+9.59/6}
Rf8 {-10.85/8} 30. Qa7 {+10.50/7} g6 {-11.55/8} 31. Qxb7 {+11.66/7}
Rf7 {-12.74/8} 32. Qxc6 {+11.91/7} Re7 {-13.04/7} 33. Bg4 {+12.50/6}
Kg7 {-15.57/8} 34. Qd6 {+14.46/6} Rxc7 {-16.72/7} 35. Qxe5+ {+15.93/7}
Kf8 {-17.10/7} 36. Qxc7 {+99.93/5} Ke8 {-99.95/7} 37. Be6 {+99.95/4}
h5 {-99.97/8} 38. Qf7+ {+99.97/2} Kd8 {-99.99/8} 39. Qd7# {+99.99/1}
{White mates} 1-0

[Event "RWBC_Qualification_6th_Edition"]
[Site "ESPRESSO"]
[Date "2005.10.05"]
[Round "4"]
[Number "109"]
[White "Fafis_26"]
[Black "Damas_7b"]
[Result "1-0"]
[TimeControl "40/900"]

1. c4 b6 2. Nc3 e6 3. Nf3 {+0.04/16} Bb7 4. e3 {+0.25/15} Bxf3 {+0.19/7} 5.
Qxf3 {+0.67/15} c6 {+0.00/8} 6. d4 {+0.86/14} d5 {+0.09/7} 7.
cxd5 {+1.34/14} exd5 {+0.16/7} 8. Bd2 {+1.64/13} Nf6 {+0.37/7} 9.
Bd3 {+1.14/13} Bb4 {+0.13/7} 10. Rc1 {+1.19/12} a5 {+0.13/7} 11.
O-O {+2.25/14} O-O {+0.16/7} 12. a3 {+0.83/13} Bd6 {+0.28/7} 13.
e4 {+1.23/13} dxe4 {+0.30/7} 14. Nxe4 {+1.18/13} Be7 {+0.29/7} 15.
Be3 {+0.95/12} Ra7 {+0.20/7} 16. Qh3 {+0.90/12} Nd5 {+0.20/6} 17.
Bd2 {+0.78/12} a4 {+0.72/6} 18. Ng5 {+0.83/11} h6 {+0.76/7} 19.
Ne4 {+0.77/11} Rd7 {+0.77/6} 20. Qf5 {+0.93/12} Re8 {+0.77/6} 21.
Nd6 {+3.51/14} Nf6 {-1.39/7} 22. Nxe8 {+3.54/13} Qxe8 {-1.67/7} 23.
Be3 {+3.43/12} Rd6 {-1.90/6} 24. Rfe1 {+3.44/12} c5 {-1.64/6} 25.
Bc4 {+3.83/12} cxd4 {-2.67/7} 26. Bf4 {+3.95/13} Rd8 {-3.27/6} 27.
Bc7 {+4.89/13} d3 {-2.86/6} 28. Bb5 {+4.86/12} Qf8 {-2.99/7} 29.
Bxd8 {+10.00/13} d2 {-4.83/7} 30. Bxe7 {+14.58/15} dxe1=Q+ {-9.55/7} 31.
Rxe1 {+15.97/17} g6 {-10.30/7} 32. Qxf6 {+15.97/15} Qg7 {-12.83/7} 33.
Qxb6 {+20.42/16} Na6 {-12.83/6} 34. Qd8+ {+99.84/16} Kh7 {-18.85/6} 35.
Bf6 {+99.84/12} Nc7 {-21.72/7} 36. Bxg7 {+99.84/11} Kxg7 {-22.41/7} 37.
Qxc7 {+99.84/8} h5 {-22.41/6} 38. Be8 {+99.84/2} Kh8 {-99.94/6} 39.
Qxf7 {+99.84/2} g5 {-99.96/4} 40. Qf8+ {+99.86/2} Kh7 {-99.98/2} 41.
Re7# {+99.98/2}
{White Mates} 1-0

Analysis of both versions on a random position from AC5:
(It is necessary to move/rename the books otherwise it would start
the analysis with a move from book)
[diag]r4rk1/1Q3ppp/2p1b3/p1Pp4/3P4/2P2N2/4BPPP/q1B2RK1 w Qq - 0 1[/diag]
r4rk1/1Q3ppp/2p1b3/p1Pp4/3P4/2P2N2/4BPPP/q1B2RK1 w Qq - 0 1

Fafis 2.6:

Code: Select all

13790 <first : setboard r4rk1/1Q3ppp/2p1b3/p1Pp4/3P4/2P2N2/4BPPP/q1B2RK1 w Qq - 0 1
13981 <first : Fafis >Error {Unknown command}
17255 >first : black
17255 >first : time 2160000
otim 2160000
17255 >first : white
go
17315 <first : black
17325 <first : time 2160000
17325 <first : otim 2160000
17325 <first : white
17325 <first : go
17335 <first : Fafis> Opening book missing!!
17345 <first : 2   248   1   340   Bc1f4 Qa1a4
17355 <first : 3   304   2   1262   Qb7xc6 Qa1xc3 Bc1f4
17375 <first : 4   284   3   2319   Qb7xc6 Qa1xc3 Bc1d2 Rf8c8 Bd2xc3 Rc8xc6
17405 <first : 5   295   7   7269   Qb7xc6 Qa1xc3 Bc1d2 Qc3c2 Qc6d6
17456 <first : 6   284   12   16300   Qb7xc6 Qa1xc3 Bc1d2 Qc3c2 Rf1c1 Qc2e4
17596 <first : 7   291   26   39368   Qb7xc6 Qa1xc3 Bc1d2 Qc3c2 Rf1c1 Qc2e4 Be2c4
17966 <first : 8   279   62   122051   Qb7xc6 Qa1xc3 Bc1d2 Qc3c2 Rf1c1 Qc2e4 Be2b5 f7f6
18447 <first : 9   286   112   227154   Qb7xc6 Qa1xc3 Bc1d2 Qc3c2 Rf1c1 Qc2e4 Be2b5 Rf8b8 Rc1e1
19458 <first : 10   290   212   451735   Qb7xc6 Qa1xc3 Bc1d2 Qc3c2 Be2a6 Qc2a2 Ba6d3 Ra8b8 Bd2f4 Rb8b2
21241 <first : 11   296   391   897421   Qb7xc6 Qa1xc3 Bc1d2 Qc3c2 Be2a6 Qc2a2 Ba6d3 Ra8b8 Bd2f4 Rb8b2 Nf3g5
28752 <first : 12   285   1142   2620930   Nf3g5 Qa1xc3 Ng5xe6 f7xe6 Qb7xc6 Qc3xd4 Qc6xe6 Rf8f7 Bc1e3 Qd4e4 Qe6d6 Rf7f8
41520 <first : 13   301   2419   5472266   Nf3g5 Ra8e8 Qb7xc6 Qa1xc3 Bc1e3 Qc3b3 Be2f3 Re8c8 Qc6d6 Rf8d8 Qd6e5 Rd8e8 Rf1e1
65194 <first : 14   339   4786   10475542   Nf3g5 Qa1xc3 Ng5xe6 f7xe6 Qb7xc6 Qc3xd4 Qc6xe6 Kg8h8 Bc1e3 Rf8f6 Qe6d7 Qd4e4 Rf1d1 Rf6g6
102087 <first : 15   330   8475   18538966   Nf3g5 Qa1xc3 Ng5xe6 f7xe6 Qb7xc6 Qc3xd4 Qc6xe6 Kg8h8 Bc1e3 Rf8f6 Qe6d7 Qd4e4 c5c6 d5d4 Qd7xd4 Qe4xc6

Fafis 2.52:
(does some weird things BTW, starts before the 'go' command and
seems to freeze in an illegal PV)

Code: Select all

3435 >first : setboard r4rk1/1Q3ppp/2p1b3/p1Pp4/3P4/2P2N2/4BPPP/q1B2RK1 w Qq - 0 1
4226 <first : 1   -322   44   7   Qb7xa8 Rf8xa8
4236 <first : 1   180   45   38   Qb7xc6 Qa1xc3
4236 <first : 2   180   45   155   Qb7xc6 Qa1xc3
4306 <first : 3   199   52   5344   Qb7xc6 Rf8c8 Qc6b7 Qa1xc3
4486 <first : 4   187   70   24266   Qb7xc6 Qa1xc3 Qc6b7 Rf8e8
5568 >first : black
5568 >first : time 2160000
otim 2160000
5568 >first : white
go
5648 <first : Fafis >Error {Unknown command}
6149 <first : 5   232   236   160916   Qb7xc6 Ra8c8 Qc6b5 Qa1xc3 Rf1e1
7230 <first : 5   8647   344   274751   Qb7xa8 Kg8h8 Qa8xe8 Kf8xe8 Bc1f4 Qa1b2

-no more output after ply 5 illegal PV -

[Pradu]

Code: Select all

13790 <first : setboard r4rk1/1Q3ppp/2p1b3/p1Pp4/3P4/2P2N2/4BPPP/q1B2RK1 w Qq - 0 1
13981 <first : Fafis >Error {Unknown command}

The "new" one dosen't seem to understand the setboard command while the old one does, how odd...

[Jim Ablett]

This all is not intended as an offense against Rafael, but IMHO
these are reasonable questions(after what happened in the past).

The clone issue was bad enough, but the intentionally embedded virus
in the code was unforgivable!

Jim.

[Jim Ablett]

The clone issue was bad enough, but the intentionally embedded virus
in the code was unforgivable! Mad

The infected engine was submitted to AVG virus labs, who stated
it was no accidental infection, and that they had never come across
a program like this before. (embedded payload virus)

Jim.

[Pradu]

Guenther your computer ok?

[Guenther Simon]

Jim said:

Quote:
This all is not intended as an offense against Rafael, but IMHO
these are reasonable questions(after what happened in the past).


The clone issue was bad enough, but the intentionally embedded virus
in the code was unforgivable! Mad

Jim.

Hi Jim,

I didn't make this thread up for warming this up again and I really
don't know if there still was a chance that the virus was introduced
by another person or by accident(e.g. when the obfuscating tool was
infected,or some one else compiled that version of Fafis)?

My interest is more with the present Fafis and if it is possible
to increase search depths +6/7 in 4 weeks with 'normal' means?

From the readme for version 2.6:

Code: Select all

---> Fafis v 2.6         -02/10/2005-

   -Some experiment between c#/c++/asm
   -Added some eval funtions.
   -Asm code.
   -Some Bitboard added for speed.

[Roman Hartmann]

Jim said:

Quote:
This all is not intended as an offense against Rafael, but IMHO
these are reasonable questions(after what happened in the past).


The clone issue was bad enough, but the intentionally embedded virus
in the code was unforgivable! Mad

Jim.

[ ... ]

My interest is more with the present Fafis and if it is possible
to increase search depths +6/7 in 4 weeks with 'normal' means?

From the readme for version 2.6:

Code: Select all

---> Fafis v 2.6         -02/10/2005-

   -Some experiment between c#/c++/asm
   -Added some eval funtions.
   -Asm code.
   -Some Bitboard added for speed.

Hi Guenther,
if you add nullmove and/or move ordering you can increase your search depths by several plies. I added nullmove and fixed my move ordering some weeks ago and got instantly about 4 additional plies. So if he had some bugs in the move ordering part (like I had) or added some pruning (like nullmove) it's definitely possible that the engine searches a good deal deeper now.

best regards
Roman

[Guenther Simon]

Roman said:

Hi Guenther,
if you add nullmove and/or move ordering you can increase your search depths by several plies. I added nullmove and fixed my move ordering some weeks ago and got instantly about 4 additional plies. So if he had some bugs in the move ordering part (like I had) or added some pruning (like nullmove) it's definitely possible that the engine searches a good deal deeper now.

best regards
Roman

But I would say this depends on the given depth of the previous version?
IMHO it is much easier to get from depth 4 to 10/11( in 20-30 sec. on a 2.6Ghz)
machine than from depth 7 to 13/14. (Note: all depths of course in opening positions.)
I also believe nullmove/ and or move ordering already was there,
but I have no source of 2.52 here to verify.

Guenther

P.S. Roce does much better now in the Qualification, than in my test
before (may be it is better at slower tcs?)

[Event "RWBC_Qualification_6th_Edition"]
[Site "ESPRESSO"]
[Date "2005.10.06"]
[Round "5"]
[Number "134"]
[White "Roce_00344"]
[Black "Elf_130"]
[Result "1-0"]
[TimeControl "40/900"]

1. e4 c6 2. d4 d5 3. Nc3 dxe4 4. Nxe4 Bf5 5. Ng3 Bg6 6. h4 h6 7. Nf3 Nd7 8.
Bd3 {-0.41/10} Bxd3 {-0.64/8} 9. cxd3 {-0.45/11} Qb6 {+0.20/7} 10.
O-O {-0.37/10} e6 {+0.14/7} 11. Ne4 {-0.09/9} Be7 {+0.29/7} 12.
Rb1 {-0.41/10} O-O-O {+0.34/7} 13. Ne5 {+0.06/9} Nxe5 {+0.59/8} 14.
dxe5 {-1.02/11} Qa5 {+0.41/7} 15. Bf4 {-1.18/10} Qxa2 {+0.31/7} 16.
d4 {-0.73/9} Qd5 {+0.75/8} 17. Nd6+ {-0.94/10} Bxd6 {+0.80/8} 18.
exd6 {-1.18/11} Nf6 {+0.92/8} 19. Qa4 {-0.93/10} Ne4 {+0.97/7} 20.
Qxa7 {+0.22/9} Nxd6 {+0.80/7} 21. Rbc1 {-0.03/9} Rhe8 {+0.77/7}?? 22.
Rc5 {+0.90/9} Qe4 {-0.82/8} 23. Bxd6 {+4.11/9} Rxd6 {-0.73/7} 24.
Qa8+ {+3.91/9} Kc7 {-3.30/7} 25. Qxe8 {+3.05/11} Rd7 {-3.17/8} 26.
Rc3 {+2.94/10} Qxh4 {-2.73/8} 27. Rf3 {+2.90/10} Qd8 {-2.91/9} 28.
Qxd8+ {+3.07/11} Kxd8 {-2.80/8} 29. Ra1 {+2.15/8} Kc7 {-2.90/9} 30.
Ra8 {+2.40/8} f5 {-2.94/9} 31. Re3 {+2.58/8} Rxd4 {-2.98/9} 32.
Rxe6 {+2.33/8} Rd1+ {-3.14/8} 33. Kh2 {+3.61/9} Rd2 {-3.49/9} 34.
Re7+ {+2.76/8} Kd6 {-3.51/8} 35. Rxb7 {+3.83/9} Rxf2 {-3.55/8} 36.
Rxg7 {+3.44/8} Rxb2 {-3.66/8} 37. Rh8 {+3.52/8} Kd5 {-3.59/9} 38.
Rxh6 {+4.60/9} c5 {-3.79/8} 39. Rf7 {+3.74/8} Ke5 {-3.91/9} 40.
Rh5 {+3.79/8} Rf2 {-3.74/8} 41. Kg1 {+4.73/10} Rf4 {-4.93/9} 42.
g3 {+5.54/9} Rf3 {-5.08/9} 43. Kg2 {+5.77/9} Rc3 {-6.07/9} 44.
Rfxf5+ {+5.84/9} Kd6 {-6.07/7} 45. Rf6+ {+5.76/9} Ke7 {-6.07/8} 46.
Ra6 {+5.74/10} Rc2+ {-6.37/8} 47. Kh3 {+9.88/9} Kf8 {-6.62/8} 48.
Ra7 {+6.10/9} Kg8 {-6.82/9} 49. Rg5+ {+9.87/9} Kf8 {-6.52/8} 50.
Rd5 {+6.10/9} Ke8 {-6.52/9} 51. Rh5 {+10.88/9} Kf8 {+0.00/2} 52.
Rh8# {+99.99/9}
{White mates} 1-0

Big blunder by Elf though with 21...Rhe8??

[Uri Blass]

I think that it is certainly possible to get a lot of plies in few weeks if you
add a lot of pruning.

Adding pruning and improving the order of moves certainly can cause increasing depth from depth 7 to depth 13/14 and it can be done in a short time.

Uri

[Dann Corbit]

His line about hand coded assembly is pure crap.
I'm not saying that he's lying...
No, wait. Yes I am.

[Roman Hartmann]

[ ... ]

P.S. Roce does much better now in the Qualification, than in my test
before (may be it is better at slower tcs?)

[ ... ]

I guess Roce plays a bit better under long timecontrol. There is still no QS and if Roce gets more time and is able to search a bit deeper it might get away with that. But because of the missing QS Roce can't play any 'silent' moves. If there is nothing going on on the board Roce will usually outplay itself by producing some unsound moves.

best regards
Roman

[Jim Ablett]

I didn't make this thread up for warming this up again and I really
don't know if there still was a chance that the virus was introduced
by another person or by accident(e.g. when the obfuscating tool was
infected,or some one else compiled that version of Fafis)?

Sorry Guenther, you're right. When the virus issue occurred
I kept it to myself what I really thought. Just needed to get it
off my chest I guess. What's passed is passed.
Shouldn't have brought it up again.

regards,
Jim.

[Roger Brown]

[ ... ]

P.S. Roce does much better now in the Qualification, than in my test
before (may be it is better at slower tcs?)

[ ... ]

I guess Roce plays a bit better under long timecontrol. There is still no QS and if Roce gets more time and is able to search a bit deeper it might get away with that. But because of the missing QS Roce can't play any 'silent' moves. If there is nothing going on on the board Roce will usually outplay itself by producing some unsound moves.

best regards
Roman

Hello Roman,

Wonderful.

So now I can take Roce off my list of sparring partners I can actually compete with. I really was enjoying playing at my full (patzer) strength and actually winning a game or two.

Sigh.

What next? Try and weaken Shredder or something?



Thanks for your (formerly) beatable engine.



Later.

[Roman Hartmann]

Hello Roman,

Wonderful.

So now I can take Roce off my list of sparring partners I can actually compete with. I really was enjoying playing at my full (patzer) strength and actually winning a game or two.

Sigh.

What next? Try and weaken Shredder or something?



Thanks for your (formerly) beatable engine.



Later.

Hi Roger,
the older version is, of course, still available on my site. But even the newer version is probably beatable by you as it doesn't know too much about chess. It's just searching a bit deeper now

best regards
Roman