Firefly 1.25 time management problem (ponder on)?

[Guenther Simon]

@Andrew especially

In some quick preparation games with ponder on,
Firefly lost always on time?
(tried some anticomp lines BTW
This was on my slower machine PIII 650Mhz with 384MB hash.
Of course there was no swapping because of sufficient Ram.
May be I had too big hashtables for 2/40 and 4/40?
(logs and debug available)

My ini file for Firefly:

Code: Select all

egtbpath C:\Sicherheit\Chess\Tbs
poshash 96MB
pawnhash 32MB
!indexer 1000
UseBitbase
log

FF used around 164MB hash with that settings.
I noticed that FF sees quite big time differences in its log files?

e.g.:

Code: Select all

time 1321
otim 7549
usermove Bf3
Search: i=9 a=231 b=277 s=259
   {Last iteration took 1.20s}
   {Stopped thinking}
   {Total think time = 4.66}
   {Not the predicted move}
   {Time Diff = -62.28s}
   {Time Left = 13.21s}
   {Moves to next TC = 7}
   {Think time = 1.47s}
 1   118   0   0   34...Rd6 35.Ra1 Rb8 36.Ra3 Rc8
 6   141   251   2154   34...Rd6 35.Ka3 Rc8 36.Rb3 a5
Search: i=6 a=95 b=141 s=141
 6   141   252   2154   34...Rd6 35.Ka3 Rc8 36.Rb3 a5
 6   233   254   7159   34...Rd6 35.Ka3 Rc8 36.Rb3 a5
Search: i=6 a=91 b=233 s=233
 6   233   254   7159   34...Rd6 35.Ka3 Rc8 36.Rb3 a5
 6   254   257   13472   34...Rd6 35.Ka3 a5 36.Rb3 Rb8 37.c4
Search: i=6 a=183 b=601 s=254
   {Last iteration took 1.34s}
 7   253   269   35792   34...Rd6 35.Ka3 a5 36.c4 Rc8 37.Be2
Search: i=7 a=231 b=277 s=253
   {Last iteration took 0.22s}
 8   241   325   171559   34...Rd6 35.c4 a5 36.Ka3
Search: i=8 a=230 b=276 s=241
   {Last iteration took 0.84s}
 9   229   451   490441   34...Rd6 35.c4 a5 36.Ka3 Rc8 37.Be2 Rdc6 38.Rg1
Search: i=9 a=218 b=264 s=229
   {Last iteration took 1.09s}
   {Stopped thinking}
   {Total think time = 4.78}
move Rd6
   {Predicting from PV}
   {Prediction : c4}
time 842
otim 7354
usermove Kc2
 1   229   0   0   35...Rc8 36.Be2 Rdc6 37.Rg1
 6   251   245   9206   35...Rc8 36.Kc3 a5 37.Rg1 Rf8
Search: i=6 a=206 b=252 s=251
   {Last iteration took 1.24s}
 7   245   290   132695   35...a5 36.Ka3 Rc8 37.Be2 Rdc6 38.d5 Rd6
Search: i=7 a=228 b=274 s=245
   {Last iteration took 0.44s}
 8   224   335   245230   35...a5 36.Kc3 Rfd8 37.Rg1 Kf7 38.c5
 8   242   418   468444   35...Rc8 36.Kb3 a5 37.Rg1 Rc7 38.d5 Ra7
Search: i=8 a=222 b=268 s=242
   {Last iteration took 1.38s}
 9   232   550   809153   35...Rc8 36.Kb3 a5 37.Kc3 Rcd8 38.Rg1 a4 39.Kb4
Search: i=9 a=219 b=265 s=232
   {Last iteration took 2.58s}
10   223   879   1663551   35...Rc8 36.Kc3 a5 37.Ra1 Kf7 38.Rg1 Rc7 39.Rc1
result 1-0 {White wins on time}
Search: i=10 a=209 b=255 s=223
   {Last iteration took 4.10s}
   {Stopped thinking}
   {Total think time = 10.99}
   {Not the predicted move}
   {Time Diff = -65.12s}
   {Time Left = 8.42s}
   {Moves to next TC = 6}
   {Think time = 1.05s}
force
 1   266   0   0   35...a5 36.Ra1 Kf7 37.Rg1
 6   289   250   12479   35...a5
Search: i=6 a=243 b=289 s=289
 6   289   250   12479   35...a5
 6   294   253   19615   35...a5 36.Rb3 a4 37.Ra3 b5
Search: i=6 a=239 b=381 s=294
   {Last iteration took 1.30s}
Search: i=7 a=271 b=317 s=271
 7   271   281   86164   35...a5
 7   254   285   96387   35...a5 36.c4 Rfd8 37.Rg1 a4 38.c5
Search: i=7 a=179 b=321 s=254
   {Last iteration took 0.44s}
 8   258   347   254667   35...a5 36.c4 Rc8 37.Kb3 Rc7 38.Rg1 Rd8
Search: i=8 a=231 b=277 s=258
   {Last iteration took 0.96s}
Search: i=9 a=235 b=281 s=239
   {Last iteration took 0.73s}
 9   258   472   575961   35...a5 36.c4 Rc8 37.Kb3 Rc7 38.Rg1 Rd8 39.d5 Kf7
   {Stopped thinking}
   {Total think time = 4.73}
ping 10
pong 10

[Event "Computer chess game"]
[Site "KABA"]
[Date "2005.03.29"]
[Round "-"]
[White "Guenther Simon"]
[Black "FireFly v1.2.5"]
[Result "1-0"]
[TimeControl "40/120"]

1. c3 e5 2. g3 d5 3. Bg2 Nf6 4. d3 Nc6 5. Qc2 Be7 {+1.19/6} 6. e4
dxe4 {+1.38/7} 7. dxe4 O-O {+1.52/6} 8. Nf3 Bg4 {+1.48/6} 9. O-O
Bxf3 {+1.06/6} 10. Bxf3 Qd7 {+0.86/6} 11. Be3 Rfd8 {+0.87/6} 12. Rd1
Qh3 {+0.44/7} 13. Bg2 Qe6 {+0.66/7} 14. Nd2 Ng4 {+0.69/7} 15. Nf1
Rxd1 {+0.78/7} 16. Qxd1 Rd8 17. Qc2 Nxe3 {+0.77/7} 18. Nxe3 Bc5 {+0.32/7}
19. Nd5 Bb6 20. Rd1 Bc5 {+0.24/7} 21. b3
{White wins on time} 1-0

[Event "Computer chess game"]
[Site "KABA"]
[Date "2005.03.29"]
[Round "-"]
[White "FireFly v1.2.5"]
[Black "Guenther Simon"]
[Result "0-1"]
[TimeControl "40/120"]

1. d4 d6 2. e4 b6 3. Nc3 {+0.41/7} e6 4. Qf3 {+0.48/6} Bb7 5. Be3 {+0.52/6}
Nf6 6. Bf4 {+0.70/6} Be7 7. Rd1 {+0.59/7} a6 8. Bc4 {+0.83/6} O-O 9.
Nge2 {+0.75/6} Qc8 10. O-O {+0.95/6} c5 11. d5 {+0.35/6} e5 12.
Bg5 {+0.88/7} b5 13. Bxb5 {-0.36/6} axb5 14. Nxb5 {-1.30/6} Qd7 15.
Bxf6 {-0.97/7} Bxf6 16. Nbc3 {-1.19/8} Na6
{Black wins on time} 0-1

[Event "Computer chess game"]
[Site "KABA"]
[Date "2005.03.29"]
[Round "-"]
[White "Guenther Simon"]
[Black "FireFly v1.2.5"]
[Result "1-0"]
[TimeControl "40/240"]

1. d4 Nf6 2. c3 d5 3. Nf3 e6 4. g3 Bd6 {+0.49/8} 5. Bg2 O-O {+0.48/8} 6.
Bf4 Bxf4 {+0.86/10} 7. gxf4 Qd7 {+0.87/8} 8. e3 Qb5 {+0.95/9} 9. Qc2
Qa6 {+0.95/8} 10. Na3 Bd7 {+0.82/8} 11. Qd2 Ne4 {+1.04/9} 12. Qe2
Qxe2+ {+0.93/8} 13. Kxe2 Nc6 {+0.87/8} 14. h4 Ne7 {+1.04/8} 15. h5
Nf5 {+0.97/8} 16. Ng5 Bc6 {+1.01/7} 17. Nxe4 dxe4 {+0.93/9} 18. Nc4
Bb5 {+1.30/8} 19. b3 Nd6 {+1.30/8} 20. Kd2 Bxc4 {+1.72/10} 21. bxc4
Nxc4+ {+1.73/8} 22. Kc2 Nd6 {+1.79/9} 23. h6 g6 {+1.70/10} 24. a4
Rad8 {+1.81/8} 25. Rab1 b6 {+1.70/8} 26. Bf1 Rd7 {+1.88/8} 27. Bb5
Rdd8 {+1.76/10} 28. Bc6 f6 {+1.60/8} 29. a5 Nc4 {+1.67/8} 30. axb6
Na3+ {+1.53/8} 31. Kb2 Nxb1 {+2.26/9} 32. Rxb1 cxb6 {+2.42/8} 33. Bxe4
f5 {+2.44/8} 34. Bf3 Rd6 {+2.29/9} 35. Kc2
{White wins on time} 1-0

[Guenther Simon]

Hi Andrew,

I have the same problem with ponder off it seems?
This time I tried less hash (48 + 16 ~ tot. 99MB used acc. Taskmanager).
BTW my time control was moves/time as usual not Fischer,
so your test did not help much.
Please try something like 2 - 4 min/40 as a Human, may be you
can reproduce it? Hopefully there is no bug regarding Humans
and Alessandros WB?

Cheers,
Guenther

Hi Guenther,

FF runs fine here (excuse) even with 5min + 5s. Tried with 1min + 1s also and no lost on time.

Your hash tables sizes is much larger than I normally use.
I usualy run with 32/16MB for positional/pawn hash. Have you tried smaller hash settings? Large hash tables does not necessarily mean stronger play Wink.

BTW, I have a 1.3.0 beta 1 on my home page if you care to try it. Same time management though. It is stronger during tests.

I'll test some large hash settings and see if that's the root of the problem.

[Guenther Simon]

Hi again Andrew,

You should try from the other end of time controls.
I have rebooted and changed my WB to standard 4.27,
but the problem remains for FF at fast tc 40/X.
I added an engine-engine game too, no change.
Well I am curious now what's up
Please try some games 2min/40 vs. programs and vs. yourself.
BTW this is WIN2KSP4 on this machine.

Best,
Guenther

[Event "Computer chess game"]
[Site "KABA"]
[Date "2005.03.29"]
[Round "-"]
[White "Guenther Simon"]
[Black "FireFly v1.2.5"]
[Result "1-0"]
[TimeControl "40/60"]

1. d3 g6 2. e3 Nf6 3. f4 Bg7 4. Nf3 Ng4 5. Nc3 Bxc3+ 6. bxc3 O-O 7. Bd2 d5
8. h3 Nf6 9. g3 e6 10. Bg2 Nc6 11. O-O Qd6 12. Nd4 Ne7 13. Nb5 Qd7 14. a4
c5 15. d4 a6 16. Na3 Qxa4 17. dxc5 Nd7 18. e4 dxe4 19. c4
{White wins on time} 1-0

[Event "Computer chess game"]
[Site "KABA"]
[Date "2005.03.29"]
[Round "-"]
[White "Gerbil, Release 02"]
[Black "FireFly v1.2.5"]
[Result "1-0"]
[TimeControl "40/60"]

1. Nf3 c5 2. c4 Nc6 3. d4 cxd4 4. Nxd4 e6 5. Nb5 Nf6 6. Bf4 e5 7. Bd2 Be7
8. Bg5 O-O 9. N1c3 Bb4 10. Nd6 Qa5 11. Qd3 e4 12. Qd2 Bxd6 13. Bxf6 Bb4 14.
Bd4 Nxd4 15. Qxd4 Rd8 16. f3 d5 17. O-O-O Bxc3 18. Qxc3
{White wins on time} 1-0

[Guenther Simon]

Hi Andrew

No time loss in 40/2min under standard WB now!
I will go back to Alessandros WB now and test this too.
Thanks for the quick fix! I guess I will test 40/1 too.

Cheers,
Guenther

Hi Guenther,

Please check your email for FF130b2. Slightly modified time management code.

Hope this is better for your.

Andrew.