Gaviota, Wasp, The Baron, Pro Deo

[BrendanJNorman]

Mind you, I had a couple of beers just now with a business lunch, so my play is a bit sloppy, BUT....WHAT THE HELL is 36.Be5!?!?!?

[d]8/p3k1p1/bp1n1p1p/2pP1P2/2P2P2/2BK4/P6P/R7 w - -

Make you forfeit on time

Well in that case, it was a master stroke!

In all seriousness, this new release seems REALLY strong and interesting.

[Tony P.]

Actualy, Stockfish 8 slightly prefers 36. Be5 too (0.00; other moves are -1.21 at best, except 36. Re1+ Kd7 37. Be5 that's 0.00 too). White makes this sac in order to distract Black's knight from attacking the c-pawn, whose loss would be undesirable for White as it would create a protected passer for Black. The sac can't be declined - 36... Nxc4 37. Bb8 (+1.08) allows White to win the a-pawn and undermine the pawn chain.

[BrendanJNorman]

Actualy, Stockfish 8 slightly prefers 36. Be5 too (0.00, other moves are a few centipawns better for Black). White makes this sac in order to distract Black's knight from attacking the c-pawn, whose loss would be undesirable for White as it would create a protected passer for Black.

Of course - but the difference is that Prodeo time and time again, has a habit of playing such piece sacs for rolling passed pawns.

Finding the move and reaching the same position are two different things imo.

And yep, of course I was aware of the point of 36.Be5 - my "what the hell is this?" was more an expression of delighted amusement.

36.Be5 is essentially the only move in the position, but I still didn't see it until it was on the board, I admit.

This was a 3 1 game afterall.

[Tony P.]

Of course - but the difference is that Prodeo time and time again, has a habit of playing such piece sacs for rolling passed pawns.

That's a good point. Apparently, rolling passers have an unusually high weight in Prodeo's static eval

[PK]

I just started to test Rodent III. I hope I will soon have time to test the complete package: personnality, opening repertoire, on Rubinstein's own games, and to tweak the engine a bit, since I don't always agree with Pawel assessment of a player style.

@matejst, it would be interesting to see Your tweaks as well as reasoning behind them. I don't know as much about the players of the past as I woud like to, so any suggestion is appreciated.

[Rebel]

Of course - but the difference is that Prodeo time and time again, has a habit of playing such piece sacs for rolling passed pawns.

That's a good point. Apparently, rolling passers have an unusually high weight in Prodeo's static eval

You both have seen this well and it has been the case since the mid or late 80's. Consider the below position.

[d] 8/4kpbn/p1p3p1/Pp2p2p/1P2Pn2/N1P1BP2/5P1P/5BK1 w - - 0 1

And the analysis of Mephisto Gideon of 1993, Nxb5 is found almost instantly based on the unstoppable A5 passer that is created. No reductions or other forms of pruning back then that have the potential to undermine the evaluation. The current ProDeo needs 17 plies to find it while the formula has not changed during the years.

Code: Select all

Mephisto Gideon Professional (1993):
   1	00:00	           7	7.000	-0,22	Bxf4 exf4
   1	00:00	          38	38.000	+0,03	Nc2
   1	00:00	          63	63.000	+0,03	c4
   2	00:00	         596	596.000	-0,06	c4 Ng5
   2	00:00	         795	795.000	-0,05	Nc2 Bh6
   3	00:00	       1.517	1.517.000	-0,03	Nc2 Ng5 Bc5+ Kd7
   3	00:00	       2.401	2.401.000	+0,03	Bc5+ Ke6 Nc2
   4	00:00	       3.644	3.644.000	-0,15	Bc5+ Ke6 Nc2 Bh6
   4	00:00	       5.454	5.454.000	-0,06	Nc2 Bh6 c4 f5
   5	00:00	      11.803	11.803.000	-0,06	Nc2 Ng5 Bc5+ Kd7 Ne1
   6	00:00	      24.974	24.974.000	-0,17	Nc2 Ng5 Bc5+ Kd7 Ne1 Bf6
   6	00:00	      36.522	36.522.000	-0,16	Nb1 Ng5 Nd2 f5 c4
   7	00:00	      74.788	74.788.000	-0,16	Nb1 Ng5 Nd2 Bf6 c4 Kd7 Bc5
   7	00:00	      89.440	89.440.000	-0,08	Nc2 Ng5 Ne1 Bf6 c4 Kd7 Bc5
   7	00:00	     111.511	111.511.000	-0,08	c4
   7	00:00	     122.265	122.265.000	+0,08	c4 Ng5 cxb5 Nxf3+ Kh1 cxb5 Nxb5 Bh6
   8	00:00	     151.786	151.786.000	+0,06	c4 Ng5 cxb5 Nxf3+ Kh1 cxb5 Nxb5 Kd7 Nc3
   9	00:00	     233.550	233.550.000	+0,12	c4 Ng5 cxb5 Nxf3+ Kh1 cxb5 Nxb5 axb5 Bxb5 Kd6
   9	00:00	     314.415	314.415.000	+0,12	Nxb5
   9	00:00	     351.727	351.727.000	+0,34	Nxb5 cxb5 c4 bxc4 Bxc4 Ng5 Bxa6 Nxf3+ Kh1 Ng5 Bb7
  10	00:00	     427.975	427.975.000	+0,30	Nxb5 cxb5 c4 bxc4 Bxc4 Ng5 Bxa6 Nxf3+ Kh1 Ng5 Bc5+ Kd7 Bc4 Nxe4 Bxf7
  11	00:00	     655.108	655.108.000	+0,30	Nxb5 cxb5 c4 bxc4 Bxc4 Ng5 Bxa6 Nxf3+ Kh1 Ng5 Bc5+ Kd7 Bc4 Nxe4 Bxf7
  11	00:00	     842.183	842.183.000	+0,30	c4
  11	00:00	     655.108	655.108.000	+0,30	Nxb5 cxb5 c4 bxc4 Bxc4 Ng5 Bxa6 Nxf3+ Kh1 Ng5 Bc5+ Kd7 Bc4 Nxe4 Bxf7
  12	00:01	   3.920.499	3.920.499	+1,74	Nxb5 cxb5 Bxb5 Ng5 Bxa6 Kd6 Kh1 Kc7 Bc4 f5 a6 Nxf3 Bd5 Bf8

[matejst]

Dear Pawel,

I will try to tweak the Rubinstein personnality. I analyzed his games extensively when I was young and I appreciate this player very much. I think that he was an all-round player for his time, very objective at his best, balanced, not really defensive.

Since you gave all the tools one needs (the opening book is important), I plan to analyze his games again and experiment with the values in the personnality file. They are self-explanatory, but I still don't know well how Rodent plays overall, nor do I know the full effect of the parameters, so it will be a (rather pleasant) process. I'll pm you the personnality file when I'm finished, so you can test it and publish, or do whatever you want with it.

[Rebel]

Dear Brendan,

About Pro Deo: its search is faster now, but its game makes less sense. And I have already seen Ed's tweaking tool. Recently, I also found a huge collection of excellent personnalities for Pro Deo on another site.

I am very disappointed that chess engines aren't opponents any more, hadn't become teachers, but are only used for hyperbullet matches and tournaments.

I really agree that it's a shame that only a few programmers are trying to create OPPONENTS for humans now (which was the original intention of computer chess) and are more focussed on never-ending Elo gain.

If styling is the real concern have a look (again perhaps) at the MP version of ProDeo 2.2. It's about styling, that it also is a bit stronger is irrelevant. Create a style with the online personality editor and store it as Q3.ENG in the Q3 folder.

[jmartus]

Whenever I try to load Prodeo 2.2 SMP version q3.exe or Rebeluci engine in the q3 folder in arena it just says error starting engine

[Rebel]

Whenever I try to load Prodeo 2.2 SMP version q3.exe or Rebeluci engine in the q3 folder in arena it just says error starting engine

Not sure if I understand correctly, but q3.exe isn't a loadable engine, it's a slave engine under the control of ProDeo.