Jumbo 0.4.16

[Sven]

I have released a new version of Jumbo. You can download it here. It has some improvements in time management and evaluation (mostly king safety). My final tests are still running but my strength estimate is around +20 Elo compared to 0.4.0.

[Sven]

Copy&paste bug on my web page now fixed, instead of the link to the 0.4.16 Win64 binary there had been an old link by accident.

[MahmoudUthman]

shouldn't it's king safety see the danger at the end of this game ?
[pgn]
[Event "T6branch7th"]
[Site "DESKTOP-9IAP23G"]
[Date "2017.02.01"]
[Round "1"]
[White "Jumbo64-0.4.16"]
[Black "CPE"]
[Result "0-1"]
[BlackElo "2000"]
[ECO "B00"]
[Opening "Nimzowitsch Defence"]
[Time "00:27:05"]
[Variation "2.Nc3 e6"]
[WhiteElo "2000"]
[TimeControl "60"]
[Termination "normal"]
[PlyCount "50"]
[WhiteType "human"]
[BlackType "human"]

1. Nc3 {(b1c3 d7d5 d2d4 b8c6 g1f3 g8f6 c1f4 c8f5 c3b5 a8c8) +0.20/10 1} e6
{(e7-e6) -0.57/8 1} 2. e4 {(e2e4 b8c6 d2d4 g8f6 g1f3 d7d5 c1g5 h7h6 e4d5
e6d5) +0.38/10} Nc6 {(Nb8-c6) -0.11/7} 3. Nf3 {(g1f3 g8f6 e4e5 f6d5 c3d5
e6d5 d2d4 f8e7 f1d3 e8g8) +0.37/10 1} Bb4 {(Bf8-b4) -0.51/8 4} 4. a3 {(a2a3
b4a5 b2b4 a5b6 d2d4 g8f6 e4e5 f6d5 c3d5 e6d5 c1g5) +0.85/11 1} Bxc3
{(Bb4xc3) -0.77/8 3} 5. dxc3 {(d2c3 d7d6 f1c4 g8f6 d1e2 e8g8 c1g5 h7h6 g5f4
e6e5) +0.86/10 1} d6 {(d7-d6) -0.55/7} 6. Bb5 {(f1b5 e6e5 e1g1 g8f6 d1d3
e8g8 c1e3 c8e6 b5c4 d8e7) +0.84/10 1} e5 {(e6-e5) -0.52/7 1} 7. Be3 {(c1e3
g8f6 e1g1 e8g8 d1d3 c8e6 f3g5 d8d7 f2f3 a7a6 g5e6 d7e6) +0.69/10 1} Nf6
{(Ng8-f6) -0.28/7} 8. O-O {(e1g1 e8g8 d1d3 c8e6 f3g5 d8d7 g5e6 d7e6 f2f3)
+0.70/9 1} Nxe4 {(Nf6xe4) +0.02/7 1} 9. Qd5 {(d1d5 e4f6 b5c6 b7c6 d5c6 c8d7
c6a6 d8b8 a6a5 e8g8 a5b4 b8b5) +0.16/10 1} Nf6 {(Ne4-f6) +0.26/7} 10. Qc4
{(d5c4 e8g8 b5c6 c8e6 c4b5 b7c6 b5c6 d8d7 c6d7 e6d7) +0.09/10 1} Bd7
{(Bc8-d7) +0.61/7 1} 11. Rfd1 {(f1d1 e8g8 c4h4 d7g4 b5c6 b7c6 f3e5 g4d1
a1d1 d6e5 d1d8 f8d8) 0.00/8 1} a6 {(a7-a6) +0.83/7 1} 12. Bxc6 {(b5c6 b7c6
g1h1 f6g4 h2h3 g4e3 f2e3 e8g8 a3a4) -0.23/9 1} bxc6 {(b7xc6) +0.61/8 3} 13.
Bg5 {(e3g5 e8g8 a3a4 h7h6 g5e3 f6d5 g1h1 d7f5 c4c6 d5e3 f2e3 f5c2) -0.42/9
1} O-O {+1.02/7 1 (O-O)} 14. a4 {(a3a4 h7h6 g5e3 f6d5 a1a3 d5e3 f2e3 d7f5
c4c6 f5c2) -0.56/8 1} h6 {(h7-h6) +1.29/7} 15. Be3 {(g5e3 e5e4 f3d2 d7e6
c4e2 d8b8 b2b3 e6g4 f2f3) -0.73/9 1} Ng4 {(Nf6-g4) +1.49/7 1} 16. Re1
{(d1e1 d6d5 c4c5 a8b8 c3c4 b8b2 c4d5 c6d5 c5d5 g4e3 e1e3) -0.85/9 1} d5
{(d6-d5) +1.63/7 1} 17. Qc5 {(c4c5 a8b8 b2b4 e5e4 f3d4 d8h4 h2h3 g4e5 g1h1)
-1.01/9} e4 {(e5-e4) +1.64/7 1} 18. Nd4 {(f3d4 d8h4 h2h3 g4e5 c3c4 e5c4
d4c6 d7c6 c5c6) -0.79/9 1} Qh4 {(Qd8-h4) +1.75/7 1} 19. h3 {(h2h3 g4e5 c3c4
e5c4 d4c6 c4e3 e1e3 h4f6 c6d4 f6e5) -0.66/10} Ne5 {(Ng4-e5) +1.73/7 1} 20.
c4 {(c3c4 e5c4 d4c6 d7c6 c5c6 f8d8 a1d1 h4f6 c6f6 g7f6 e3h6 c4b2) -0.47/10}
Bxh3 {(Bd7xh3) +2.72/6} 21. gxh3 {(g2h3 h4h3 c4d5 h3g4 g1f1 g4h3 f1e2 h3g4
e2d2 c6d5 c5d5 f8e8) +0.95/9} Qxh3 {(Qh4xh3) +8.49/7} 22. Bf4 {(e3f4 h3g4
f4g3 a8b8 c4d5 b8b2 e1b1 b2b1 a1b1 c6d5 c5d5) +1.17/8} Nf3+ {(Ne5-f3+)
+M4/5} 23. Nxf3 {(d4f3 h3f3 f4c7) +1.65/1} exf3 {(e4xf3) +M3/4} 24. Qxf8+
{(c5f8 a8f8 c4d5 h3g2) -M2/3} Rxf8 {(Ra8xf8) +M2/3} 25. cxd5 {(c4d5 h3g2)
-M1/2} Qg2# {(Qh3-g2+) +M1/1} 0-1

[/pgn]

[Sven]

shouldn't it's king safety see the danger at the end of this game ?

Thank you for the test game. Yes, of course it should ... but it doesn't I'm in the beginning of adding king safety to Jumbo, and so far my results are not convincing, I know that. Of course this game was already lost for White (after a catastrophic opening) when Black played Bxh3 in this game. But at least Bxh3 was not in the PV of Jumbo ... It needs at least depth 11 (which was too much for the blitz TC you were obviously using) to see the loss after gxh3.

[Graham Banks]

I have released a new version of Jumbo. You can download it here. It has some improvements in time management and evaluation (mostly king safety). My final tests are still running but my strength estimate is around +20 Elo compared to 0.4.0.

Thanks Sven.

[Daniel Anulliero]

shouldn't it's king safety see the danger at the end of this game ?

Thank you for the test game. Yes, of course it should ... but it doesn't I'm in the beginning of adding king safety to Jumbo, and so far my results are not convincing, I know that. Of course this game was already lost for White (after a catastrophic opening) when Black played Bxh3 in this game. But at least Bxh3 was not in the PV of Jumbo ... It needs at least depth 11 (which was too much for the blitz TC you were obviously using) to see the loss after gxh3.

Ooh yes , good king safety is a pain to implement correctly.!.. I noticed even if I borrow the KS scheme from a very strong engine it won't work correctly in my weak engine lol
See you and good luck for Graham's D7 Sven

PS: also in this game , strange Jumbo has castled instead of protecting his 'e' pawn ...
Bests
Dany

[Sven]

shouldn't it's king safety see the danger at the end of this game ?

Thank you for the test game. Yes, of course it should ... but it doesn't I'm in the beginning of adding king safety to Jumbo, and so far my results are not convincing, I know that. Of course this game was already lost for White (after a catastrophic opening) when Black played Bxh3 in this game. But at least Bxh3 was not in the PV of Jumbo ... It needs at least depth 11 (which was too much for the blitz TC you were obviously using) to see the loss after gxh3.

PS: also in this game , strange Jumbo has castled instead of protecting his 'e' pawn ...

At that point "strange Jumbo" might have seen that O-O Nxe4 Qd5 Nf6 Bxc6+ bxc6 Qxc6+ Bd7 Qc4 would win back the pawn. But maybe depth was not sufficient (here as well) to see what happens after the line above. In the game Jumbo switched from Bxc6+ to Qc4, playing with a pawn down. Don't know which is better, though ... Certainly Jumbo is not a blitz crack

[Daniel Anulliero]

shouldn't it's king safety see the danger at the end of this game ?

Thank you for the test game. Yes, of course it should ... but it doesn't I'm in the beginning of adding king safety to Jumbo, and so far my results are not convincing, I know that. Of course this game was already lost for White (after a catastrophic opening) when Black played Bxh3 in this game. But at least Bxh3 was not in the PV of Jumbo ... It needs at least depth 11 (which was too much for the blitz TC you were obviously using) to see the loss after gxh3.

PS: also in this game , strange Jumbo has castled instead of protecting his 'e' pawn ...

At that point "strange Jumbo" might have seen that O-O Nxe4 Qd5 Nf6 Bxc6+ bxc6 Qxc6+ Bd7 Qc4 would win back the pawn. But maybe depth was not sufficient (here as well) to see what happens after the line above. In the game Jumbo switched from Bxc6+ to Qc4, playing with a pawn down. Don't know which is better, though ... Certainly Jumbo is not a blitz crack

I didn't say "strange Jumbo" lol I would say "It's strange that Jumbo has castled instead of protecting his pawn"
I missed some ponctuation
So for the game , that could be also some horizon effect problems ... and don't worry , Isa is not a blitz crack too !