Not logged inRybka Chess Community Forum
Up Topic The Rybka Lounge / Test Positions / Rybka 4 vs Rybka 3
- - By Dadi Jonsson (Silver) [is] Date 2010-05-25 22:14
Here is a position I ran into just now while looking for something else. It was posted as an example of a Rybka 3 weakness. Majd tested this position and Rybka 3 plays b6 even on 8 cores (b6 draws due f4+ and the white king can't escape from the rook checks)

k7/1R6/P7/1P3p2/7P/5PK1/8/1r6 w - - 0 70


I analyzed this position with the single core version of Rybka 4 32-bit. Here are the results:

[+14.05]  d=24  70.Kf4 Rh1 71.Kg5 f4 72.Rc7 Rb1 73.Rc5 Ka7 74.h5 Rg1+ 75.Kxf4 Rg7 76.h6 Rh7 77.Rc6 Rf7+ 78.Ke3 Rd7 79.f4 Rf7 80.Ke4 Re7+ 81.Kd5 Rh7 82.Ke5 Rd7 83.Rg6 Re7+ 84.Kd6 Re8 85.h7 (0:10:08) 36450kN
[+12.87]  d=23  70.Kf4 Rh1 71.Kg5 f4 72.Rc7 Rb1 73.Rc5 Ka7 74.h5 Rg1+ 75.Kxf4 Rg7 76.h6 Rh7 77.Rc6 Rf7+ 78.Ke3 Rd7 79.f4 Rf7 80.Ke4 Re7+ 81.Kd5 Rh7 82.Ke5 Rd7 83.Rg6 Rd8 84.h7 Rh8 85.Rg7+ (0:04:13) 15216kN
[+12.87]  d=22  70.Kf4 Rh1 71.Kg5 f4 72.Rc7 Rb1 73.Rc5 Ka7 74.h5 Rg1+ 75.Kxf4 Rg7 76.h6 Rh7 77.Rc6 Rf7+ 78.Ke3 Rd7 79.f4 Rf7 80.Ke4 Re7+ 81.Kd5 Rh7 82.Ke5 Rd7 83.Rg6 Rd8 84.h7 Rh8 85.Rg7+ (0:04:07) 14925kN
[+10.61]  d=21  70.Kf4 Rh1 71.Kg5 f4 72.Rc7 Rb1 73.Rc5 Ka7 74.h5 Rg1+ 75.Kxf4 Rg7 76.h6 Rh7 77.Rc6 Rf7+ 78.Ke3 Rd7 79.f4 Rf7 80.Ke4 Re7+ 81.Kd5 Rh7 82.Ke5 Rd7 83.Rg6 Re7+ 84.Kd6 Re8 85.h7 (0:01:35) 5477kN
[+10.61]  d=20  70.Kf4 Rh1 71.Kg5 f4 72.Rc7 Rb1 73.Rc5 Ka7 74.h5 Rg1+ 75.Kxf4 Rg7 76.h6 Rh7 77.Rc6 Rf7+ 78.Ke3 Rd7 79.f4 Rf7 80.Ke4 Re7+ 81.Kd5 Rh7 82.Ke5 Rd7 83.Rg6 Re7+ 84.Kd6 Re8 85.h7 (0:01:32) 5348kN
[+9.57]  d=19  70.Kf4 Rh1 71.Kg5 f4 72.Rc7 Rb1 73.Rc5 Ka7 74.h5 Rg1+ 75.Kxf4 Rg7 76.h6 Rh7 77.Rc6 Rf7+ 78.Ke3 Rd7 79.f4 Rf7 80.Ke4 Re7+ 81.Kd5 Rh7 (0:01:14) 4472kN
[+7.98]  d=18  70.Kf4 Rh1 71.Kg5 Rg1+ 72.Kf6 f4 73.Rc7 Rb1 74.h5 Rxb5 75.h6 Rb6+ 76.Kg5 Rb8 77.h7 Rd8 78.Kxf4 Kb8 79.Rb7+ Ka8 80.Kf5 Rd5+ 81.Ke4 Re5+ 82.Kf4 Re8 83.Kg5 Rf8 84.f4 Rf5+ 85.Kxf5 (0:00:27) 1399kN
[+7.98]  d=17  70.Kf4 Rh1 71.Kg5 Rg1+ 72.Kf6 f4 73.Rc7 Rb1 74.h5 Rxb5 75.h6 Rb6+ 76.Kg5 Rb8 77.h7 Rd8 78.Kxf4 Kb8 79.Rb7+ Ka8 80.Kf5 Rd5+ 81.Ke4 Re5+ 82.Kf4 Re8 83.Kg5 Rf8 84.f4 Rf5+ 85.Kxf5 (0:00:27) 1382kN
[+6.14]  d=17  70.b6 f4+ 71.Kxf4 Rb4+ 72.Kg5 Rg4+ 73.Kf5 Rf4+ 74.Kg6 Rf6+ 75.Kg7 Rg6+ 76.Kf7 Rf6+ 77.Ke7 Rf7+ 78.Ke8 Rf8+ 79.Kd7 Rf7+ 80.Kd8 Rf8+ 81.Kc7 Rf7+ 82.Kc6 Rf6+ 83.Kc5 Rf5+ 84.Kd6 Rf6+ 85.Kc7 (0:00:06) 226kN
[+6.14]  d=16  70.b6 f4+ 71.Kxf4 Rb4+ 72.Kg5 Rg4+ 73.Kf5 Rf4+ 74.Kg6 Rf6+ 75.Kg7 Rg6+ 76.Kf7 Rf6+ 77.Ke7 Rf7+ 78.Ke8 Rf8+ 79.Kxf8 (0:00:05) 174kN
[+6.14]  d=15  70.b6 f4+ 71.Kxf4 Rb4+ 72.Kg5 Rg4+ 73.Kf5 Rf4+ 74.Kg6 Rf6+ 75.Kg7 Rg6+ 76.Kf7 Rf6+ 77.Ke7 Rf7+ 78.Ke8 Rf8+ 79.Kxf8 (0:00:04) 153kN
[+6.12]  d=15  70.h5 f4+ 71.Kxf4 Rb4+ 72.Ke3 Re4+ 73.Kd2 Rd4+ 74.Kc3 Rc4+ 75.Kb3 Rb4+ 76.Kc2 Rc4+ 77.Kd3 Rd4+ 78.Ke2 Rh4 79.Rh7 Rb4 80.h6 Rxb5 81.Rg7 Rb2+ 82.Ke3 Rb3+ 83.Ke4 Rb4+ 84.Kd5 Rb5+ 85.Kc4 (0:00:03) 123kN
[+6.14]  d=14  70.h5 f4+ 71.Kxf4 Rb4+ 72.Ke3 Re4+ 73.Kd2 Rd4+ 74.Kc3 Rc4+ 75.Kb3 Rb4+ 76.Kc2 Rc4+ 77.Kd3 Rd4+ 78.Ke2 Rh4 79.Rh7 Rb4 80.h6 Rxb5 81.Rg7 Rb2+ 82.Ke3 Rb3+ 83.Ke4 Rb4+ 84.Kd5 Rb5+ 85.Kc4 (0:00:03) 95kN
[+6.84]  d=13  70.h5 f4+ 71.Kxf4 Rb4+ 72.Kg3 Rg4+ 73.Kf2 Rg2+ 74.Ke3 Re2+ 75.Kd4 Rd2+ 76.Kc3 Rc2+ 77.Kb4 Rb2+ 78.Ka5 Ra2+ 79.Kb6 Rxa6+ 80.bxa6 (0:00:02) 86kN
[+6.80]  d=12  70.h5 f4+ 71.Kxf4 Rb4+ 72.Ke3 Re4+ 73.Kd2 Re2+ 74.Kc3 Re6 75.Rh7 Rf6 76.h6 Rxf3+ 77.Kd4 Rf4+ 78.Ke5 Rb4 79.Rb7 (0:00:01) 46kN
[+6.37]  d=11  70.h5 f4+ 71.Kxf4 Rb4+ 72.Kf5 Rf4+ 73.Kg6 Rxf3 74.Rd7 Rg3+ 75.Kf6 Rf3+ 76.Ke6 Re3+ 77.Kd6 Rb3 78.Kc6 Rc3+ (0:00:00) 24kN
[+6.53]  d=10  70.h5 Rh1 71.Rh7 Rb1 72.h6 Rxb5 73.Rb7 f4+ 74.Kxf4 Rf5+ 75.Ke4 Re5+ 76.Kd3 Rf5 77.h7 Rxf3+ 78.Ke4 Rf8 79.Kd5 Rc8 (0:00:00) 18kN
[+5.76]  d=10  70.Kf4 Rh1 71.Kg5 Rg1+ 72.Kxf5 Rg5+ 73.Ke6 Re5+ 74.Kd6 Rd5+ 75.Kc6 Rh5 76.f4 Rxh4 77.f5 Rf4 78.Rf7 (0:00:00) 9kN
[+6.09]  d=9  70.Kf4 Rh1 71.Kg5 Rg1+ 72.Kxf5 Rg5+ 73.Ke4 Rh5 74.f4 Rxh4 75.b6 Rh8 76.f5 Re8+ 77.Kd4 Rd8+ 78.Ke5 Re8+ 79.Kd5 Rd8+ 80.Kc6 (0:00:00) 4kN
[+6.15]  d=8  70.Kf4 Rh1 71.Kg5 Rg1+ 72.Kxf5 Rh1 73.b6 Rxh4 74.Rc7 Rh8 75.Rc8+ Rxc8 76.b7+ Kb8 77.bxc8Q+ Kxc8 78.f4 Kc7 79.a7 Kb7 80.Ke5 Kxa7 (0:00:00) 3kN


Rybka 4 finds the win, although it still doesn't understand that the rook chase (after b6)  leads to a draw.
Parent - By Mark (****) [us] Date 2010-05-25 23:05
Nice!  Looks like some work was put into endgames!
- - By Dadi Jonsson (Silver) [is] Date 2010-05-25 23:34 Edited 2010-05-25 23:37
Rybka 4 seems to have more realistic evaluation of rook endgames (and a new option for rook endgames).

Dragon49 wrote: "I have a number of games in my Engine Match Database,  (most against Fritz 11) where Rybka with the white pieces, has an extra outside passed pawn.  The position is completely drawn, as Rybka's rook is in front of the pawn and the opponent's rook is in back of the pawn.  There is no way to make any progress.  The attached diagram is evidence of one of these positions.  In the position, (the game was eventually drawn) Rybka's evaluation is +2.17, where Fritz's in only +0.56.  In these positions, Rybka always over evaluates, where the other engine is more realistic."

First of all, you are now less likely to see Rybka 4 place her rook in front of a passed, isolated pawn. If it happens the evaluation will be more realistic. This is the position that dragon49 was referring to:

R7/P4p2/8/5k2/6p1/1K4P1/7P/r7 b - -


Instead of +2.17 Rybka 4 gives +1.27 (without tablebases) at depth 16:

[+1.27]  d=16  75...Kf6 76.Kb2 Ra4 77.Kc1 Ke7 78.Kb1 Ra3 79.Kc2 Ra6 80.Kd2 Ra2+ 81.Ke3 Ra4 82.Ke2 Ra2+ 83.Ke3 Ra4 84.Ke2 Ra2+ 85.Ke3 Ra4 86.Ke2 Ra2+ 87.Ke3 Ra4 88.Ke2 Ra2+ 89.Ke3 Ra4 90.Ke2 Ra2+ (0:00:01) 30kN
[+1.27]  d=15  75...Kf6 76.Kb2 Ra4 77.Kc1 Ke7 78.Kb1 Ra3 79.Kc2 Ra6 80.Kd1 Ra4 81.Ke2 Ra2+ 82.Ke3 Ra4 83.Ke2 Ra2+ 84.Ke3 Ra4 85.Ke2 Ra2+ 86.Ke3 Ra4 87.Ke2 Ra2+ 88.Ke3 Ra4 89.Ke2 Ra2+ 90.Ke3 Ra4 (0:00:00) 21kN
[+1.27]  d=14  75...Kf6 76.Kb2 Ra4 77.Kc1 Ke7 78.Kb1 Ra3 79.Kc2 Ra6 80.Kd1 Ra4 81.Kc2 Ra2+ 82.Kc1 Ra3 83.Kc2 Ra6 84.Kd1 Ra4 85.Kc2 Ra2+ 86.Kc1 Ra3 87.Kc2 Ra6 88.Kd1 Ra4 89.Kc2 Ra2+ 90.Kc1 Ra3 (0:00:00) 15kN
[+1.27]  d=13  75...Kf6 76.Kb2 Ra4 77.Kc1 Ke7 78.Kb1 Ra3 79.Kc2 Ra6 80.Kd1 Ra4 81.Kc2 Ra2+ 82.Kc1 Ra3 83.Kc2 Ra6 84.Kd1 Ra4 85.Kc2 Ra2+ 86.Kc1 Ra3 87.Kc2 Ra6 88.Kd1 Ra4 89.Kc2 Ra2+ 90.Kc1 Ra3 (0:00:00) 11kN
[+1.27]  d=12  75...Kf6 76.Kb2 Ra4 77.Kc1 Ke7 78.Kb1 Ra3 79.Kc2 Ra6 80.Kb1 Ra4 81.Kc1 Ra2 82.Kb1 Ra6 83.Kb2 Ra5 84.Kc2 Ra1 85.Kd3 Ra2 86.Ke4 Kf6 87.Kf4 Ra4+ 88.Ke3 Ke7 89.Ke2 Ra2+ 90.Ke3 Ra4 (0:00:00) 8kN
Parent - - By Banned for Life (Gold) Date 2010-05-26 04:48
OK, its gone from atrocious to just really bad. That's progress, right?
Parent - - By Dadi Jonsson (Silver) [is] Date 2010-05-26 08:13
Right, but the main improvement is that Rybka4 is less enthusiastic about placing it's rook in front of a passed, isolated pawn.

Btw. here is your recent list of Rybka 3 annoyances:

Blind bishop bug, Fixed
MP tablebase bug, Fixed for those who were unfortunate enough to encounter it (ie. non-Aquarium users)
Putting rooks in front of pawns, Fixed, or at least it's less of a problem now
Very slow to push pawns, Don't know
Poor retention when doing backward analysis, Don't know
Parent - - By oudheusa (*****) [nl] Date 2010-05-26 10:09
I understood blind bishop was not fixed, but this would be good news if it is the case.
Parent - - By Fulcrum2000 (****) [nl] Date 2010-05-26 10:58
BB was fixed, Bishop underpromotions was not.
Parent - By oudheusa (*****) [nl] Date 2010-05-26 11:10
ok, thanks for clearing that up.
Parent - - By Banned for Life (Gold) Date 2010-05-26 16:54
Progress is good, even slow progress. The question is whether R4 will be the right engine to use for CC endgame analysis. I suspect the answer will still be no, but the improvements are still valuable for late middle game analysis.

I'll reserve judgment on blind bishop until seeing if it understands this with a lot of pieces on the board. I sent Vas a position with this scenario for R3 beta testing, so I'll see if I can resurrect it. I'm betting that the fix only works with a small number of pieces, but we'll see.

Tablebase bug fix is good news. This was really, really annoying for non-aquarium analysis.

Reducing incidents of rooks in front of pawns is somewhat useful, but in all likelihood, R4 will still not be a top engine for endgame analysis anyway...

Very slow to push pawns - Maybe less of an issue with the new king safety algorithm (for both sides). We'll see...

Poor retention when doing backward analysis - I consider this to be really critical, with shredder being the leader in this area. This is also tied in with persistent hash of course, and the loss of this function is a big step backward (or it would be if the R3 version had worked better). With the availability of really cheap disk based storage, it would really be nice if the engine could store all of the analysis that is done, and allow this to be used in the search (aquarium does the former, but not the latter).
Parent - - By Uly (Gold) [mx] Date 2010-05-26 18:53
(Note: This is so far for the final version of Rybka 4 that I'm using)

> Very slow to push pawns


I've seen R4 as a pawn pusher many times, sometimes neglecting to develop her pieces, instead favoring to push the pawns, so maybe there's an overcompensation.

> Poor retention when doing backward analysis


This is definitively fixed, you can now do many things (like playing with MultiPV, excluding moves, play some moves into the board, etc.), go back to the root (move by move) and Rybka will be very responsive, remembering the main move, score and depth, like it should. It's also much better at forward propagation. If you have the a GUI that cuts results from the analysis window, shut the parameter down, so you see what Rybka think since depth 1, this will make her look much more responsible as sometimes she would look as doing nothing for a long while, while indeed she sent info to the GUI, but the GUI didn't display it.

> This is also tied in with persistent hash of course, and the loss of this function is a big step backward


Sorry about that, I bugged Vas so much about this feature that he decided to remove it instead to fix the at least 2 issues (that weren't trivial to fix); or maybe I'm giving myself too much credit and he would have killed it regardless, he just didn't have time for this due to his terrible self-imposed deadline.
Parent - - By yanquis1972 (****) [us] Date 2010-05-26 19:25
vytron, i'm sure i've asked you this before, but i've created a separate 'R4 analysis' engine with pretty much everything checked. i put a file called hash.rsh in the engine folder & directed it there. is that file for manually saving has, or has it anything to do with preserve analysis as well? and if i have preserve analysis checked, what are the drawbacks if any, and how does it work?
Parent - - By Uly (Gold) [mx] Date 2010-05-26 20:41 Edited 2010-05-26 20:45

> 'R4 analysis' engine with pretty much everything checked.


Bad idea, you should only check the upperbounds settings and the always score setting (and the large pages setting if you're using it).

> hash.rsh in the engine folder & directed it there.


The procedure is, saving the hash, then, after turning off your computer and turning it on, Rybka should react as if you never switched it off. I think this is the only use.

Preserve analysis should now be redundant with new Rybka hash usage, since it hurts games and I'd see no benefit, I'd suggest turning it off.
Parent - By yanquis1972 (****) [us] Date 2010-05-26 20:44
ah, thanks. good to know PH is useless as its one of those tempting settings.
Parent - - By yanquis1972 (****) [us] Date 2010-05-26 19:27
i feel as though i've already noticed a big differnce in rybka 4's willingness to push pawns, which i personally like quite a bit...i think until recently it seems woodpushing has been underrated at top level chess ;)  -- and i have seen a few modern engines go down that way as well. in a bullet game i posted on talkchess R4 was very fast to push her kingside pawns out of the bird opening against ivanhoe and won the game. i have never seen any engine play that opening as aggressively.
Parent - - By Banned for Life (Gold) Date 2010-05-26 19:36
Does Rybka 4 still like to lock up the black position in the Yugoslav attack in the dragon? This occurred when black played h5 and white, with pawns on h4 and g4, played g5. Black frequently ended up with an impenetrable kingside...
Parent - - By Uly (Gold) [mx] Date 2010-05-26 20:43
Could you post a FEN so one can test?
Parent - - By Banned for Life (Gold) Date 2010-05-27 02:55
Yes, I will find one of the games my engine played against Eros' engine before I gave up on the dragon...
Parent - By yanquis1972 (****) [us] Date 2010-05-27 12:28
heh - it's supposed to be -quick- suicide with black, BFL.
Parent - By yanquis1972 (****) [us] Date 2010-05-26 20:46
in the 2 positions i got R4 played classically, preferring counter on the queenside...i didn't hash the variations out enough to see which idea was better.  both were being scored pretty well for white iirc. but as vytron says a specific fen would be welcome.
Parent - By Moz (****) Date 2010-05-29 03:37 Edited 2010-05-29 03:40

>Dadi, quoting Dragon49 wrote: "I have a number of games in my Engine Match Database,  (most against Fritz 11) where Rybka with the white pieces, has an extra outside passed pawn.  The position is completely drawn, as Rybka's rook is in front of the pawn and the opponent's rook is in back of the pawn.  There is no way to make any progress.  The attached diagram is evidence of one of these positions.  In the position, (the game was eventually drawn) Rybka's evaluation is +2.17, where Fritz's in only +0.56.  In these positions, Rybka always over evaluates, where the other engine is more realistic."
>


With EGTB's my system still shows +2.17.  I took the opportunity to experiment with the Rook Scaling parameter and setting it to 0 seems to be the exact value necessary to show an eval of 0.00!



[+0.00]  d=23  1...Kf6 2.Rf8 Rxa7 3.Rc8 Kg5 4.Rg8+ Kh5 5.Rh8+ Kg5 6.Kc4 Ra4+ 7.Kd3 Ra5 8.h4+ gxh3 9.Rxh3 Rf5 10.Ke3 Rb5 11.Rh7 Rb7 12.Kf2 Kf6 (0:00:14) 1159kN
[+0.00]  d=22  1...Kf6 2.Rf8 Rxa7 3.Rc8 Kg5 4.Rg8+ Kh5 5.Rh8+ Kg5 6.Kc4 Ra4+ 7.Kd3 Ra5 8.h4+ gxh3 9.Rxh3 Rf5 10.Ke3 Rb5 11.Rh7 Rb7 12.Kf2 Kf6 (0:00:11) 957kN
[+0.00]  d=21  1...Kf6 2.Rf8 Rxa7 3.Rc8 Kg5 4.Rg8+ Kh5 5.Rh8+ Kg5 6.Kc4 Ra4+ 7.Kd3 Ra5 8.h4+ gxh3 9.Rxh3 Rf5 10.Ke3 Rb5 11.Rh7 Rb7 12.Kf2 Kf6 (0:00:09) 820kN
[+0.00]  d=20  1...Kf6 2.Rf8 Rxa7 3.Rc8 Kg5 4.Rg8+ Kh5 5.Rh8+ Kg5 6.Kc4 Ra4+ 7.Kd3 Ra5 8.h4+ gxh3 9.Rxh3 Rf5 10.Ke3 Rb5 11.Rh7 Rb7 12.Kf2 Kf6 (0:00:09) 769kN
[+0.00]  d=19  1...Kf6 2.Rf8 Rxa7 3.Rc8 Kg5 4.Rg8+ Kh5 5.Rh8+ Kg5 6.h4+ gxh3 7.Rg8+ Kf5 8.Rh8 f6 9.Rxh3 Rd7 10.Kc2 Kg6 11.Rh1 Rc7+ (0:00:02) 563kN
[+0.00]  d=18  1...Kf6 2.Rf8 Rxa7 3.Rc8 Kg5 4.Rg8+ Kh5 5.Rh8+ Kg5 6.h4+ gxh3 7.Rg8+ Kf5 8.Rh8 f6 9.Rxh3 Rd7 10.Kc2 Kg6 11.Rh8 (0:00:02) 481kN
[+0.00]  d=17  1...Kf6 2.Rf8 Rxa7 3.Rc8 Kg5 4.Rg8+ Kh5 5.Rh8+ Kg5 6.h4+ gxh3 7.Rg8+ Kf5 8.Rh8 f6 9.Rxh3 Rd7 10.Kc2 Kg6 11.Rh8 (0:00:01) 319kN
[+0.00]  d=16  1...Kf6 2.Rf8 Rxa7 3.Rc8 Kg5 4.Rg8+ Kh5 5.Rh8+ Kg5 6.h4+ gxh3 7.Rf8 h2 8.Rg8+ Kf6 9.Rh8 h1Q 10.Rxh1 Rb7+ 11.Kc4 Rc7+ 12.Kb5 Rb7+ 13.Ka4 Ra7+ 14.Kb5 Rb7+ 15.Ka4 Ra7+ 16.Kb5 Rb7+ (0:00:01) 231kN
[+0.00]  d=15  1...Kf6 2.Rf8 Rxa7 3.Rc8 Kg5 4.Rg8+ Kh5 5.Rh8+ Kg5 6.h4+ gxh3 7.Rf8 h2 8.Rg8+ Kf6 9.Rh8 h1Q 10.Rxh1 Rb7+ 11.Kc4 Rc7+ 12.Kb5 Rb7+ 13.Ka4 Ra7+ 14.Kb5 Rb7+ 15.Ka4 Ra7+ 16.Kb5 Rb7+ (0:00:00) 132kN
[+0.00]  d=14  1...Kf6 2.Rf8 Rxa7 3.Rc8 Kg5 4.Rg8+ Kh5 5.Rh8+ Kg5 6.h4+ gxh3 7.Rf8 h2 8.Rg8+ Kf6 9.Rh8 h1Q 10.Rxh1 Rb7+ 11.Kc4 Rc7+ 12.Kb5 Rb7+ 13.Ka4 Ra7+ 14.Kb5 Rb7+ 15.Ka4 Ra7+ 16.Kb5 Rb7+ (0:00:00) 82kN
[+0.00]  d=13  1...Kf6 2.Rf8 Rxa7 3.Rc8 Kg5 4.Rg8+ Kh5 5.Rh8+ Kg5 6.h4+ gxh3 7.Rf8 h2 8.Rg8+ Kf6 9.Rh8 h1Q 10.Rxh1 Rb7+ 11.Kc4 Rc7+ (0:00:00) 49kN
[+0.00]  d=12  1...Kf6 2.Rf8 Rxa7 3.Rc8 Kg5 4.Rg8+ Kh5 5.Rh8+ Kg5 6.h4+ gxh3 7.Rxh3 Kg4 8.Rh4+ Kxg3 9.Rh5 Kg4 10.Rd5 f5 11.Rd4+ Kg3 (0:00:00) 30kN
[+0.00]  d=11  1...Kf6 2.Rf8 Rxa7 3.Rc8 Kg5 4.Rg8+ Kh5 5.Rh8+ Kg5 6.h4+ gxh3 7.Rxh3 Kg4 8.Rh4+ Kxg3 9.Rh5 Kg4 10.Rd5 f5 11.Rd4+ (0:00:00) 19kN
[+0.00]  d=10  1...Kf6 2.Rf8 Rxa7 3.Rc8 Kg5 4.Rg8+ Kh5 5.Rh8+ Kg5 6.h4+ gxh3 7.Rxh3 Kg4 8.Rh4+ Kxg3 9.Rh5 Kg4 10.Rd5 (0:00:00) 12kN
[+0.00]  d=9  1...Kf6 2.Rf8 Rxa7 3.Rc8 Kg5 4.Rg8+ Kh5 5.Rh8+ Kg5 6.h4+ gxh3 7.Rxh3 Kg4 8.Rh4+ Kxg3 9.Rh5 Kg4 (0:00:00) 7kN
[+0.00]  d=8  1...Kf6 2.Rf8 Rxa7 3.Rc8 Kg5 4.Rg8+ Kh5 5.Rh8+ Kg5 6.h4+ gxh3 7.Rxh3 Kg4 8.Rh4+ Kxg3 9.Rh5 (0:00:00) 4kN
[+0.00]  d=7  1...Kf6 2.Rf8 Rxa7 3.Rc8 Kg5 4.Rg8+ Kh5 5.Rh8+ Kg5 6.h4+ gxh3 7.Rxh3 Kg4 8.Rh4+ (0:00:00) 2kN
[+0.00]  d=6  1...Kf6 2.Rf8 Rxa7 3.Rc8 Kg5 4.Rg8+ Kh5 5.Rh8+ Kg5 6.h4+ gxh3 7.Rxh3 Kg4 (0:00:00) 1kN
[+0.00]  d=5  1...Kf6 2.Rf8 Rxa7 3.Rc8 Kg5 4.Rg8+ Kh5 5.Rh8+ Kg5 6.h4+ (0:00:00) 1kN
[+0.00]  d=4  1...Kf6 2.Rf8 Rxa7 3.Rc8 Kg5 4.Rg8+ Kh5 5.Rh8+ (0:00:00) 0kN


Increasing the Rook Endgame Scaling to 20 immediately shows a proportional increase in score (as you'd expect, but when it comes to Rybka and endgames, one never knows for sure and testing never hurts!)

[+0.44]  d=23  1...f6 2.Kb4 Ra2 3.Kb5 Rb2+ 4.Ka6 Ra2+ 5.Kb6 Rb2+ 6.Kc6 Rc2+ 7.Kd5 Ra2 8.Kc4 Ra5 9.Kd3 Ra2 10.Kd4 Ra4+ 11.Kc5 Ra2 12.Kd6 Rd2+ 13.Kc7 Rc2+ 14.Kd7 Ra2 15.Kd6 Rd2+ 16.Kc7 Rc2+ (0:00:15) 1752kN

I don't know the first thing about programming but intuitively it seems like there should be a much better way of handling these types of endgames via automatic scaling of some sort.
- - By Dadi Jonsson (Silver) [is] Date 2010-05-26 09:15
4R3/5pk1/3r2p1/P6p/7P/6P1/5P2/6K1 w - -


This is a rook ending which most engines have problems with, including Rybka 3. hal900 wrote:

> The correct move is Re1! (according to Leif), but Rybka prefers Ra8. I've tested this position on my octocore (with 6 man EGTB), and Rybka still prefers Ra8 after 10 minutes thinking time.


I analyzed this position with a single-core, 32-bit Rybka on a computer which blazed ahead at 40 Kn/s. No EGTBs. After less than 7 minutes it switched to the correct 1. Re1! After 10 minutes it still wanted to follow that move up with 2.Rb1 instead of the natural 2.Ra1. Progress, although not perfect, but many engines have more problems than Rybka 4 in this position.

[+1.61]  d=25  1.Re1 Kf6 2.Rb1 Ke5 3.f3 Ra6 4.Rb5+ Kd4 5.g4 hxg4 6.fxg4 Kc4 7.Rg5 Kd4 8.Kg2 Ke4 9.Rb5 Kf4 10.Kh3 f5 11.gxf5 gxf5 12.Rb4+ (0:10:45) 26249kN
[+1.68]  d=25  1.Re1 Kf6 2.Rb1 Ke5 3.f3 Ra6 4.Rb5+ Kd4 5.g4 hxg4 6.fxg4 Kc4 7.Rg5 Kd4 8.Kf2 Ke4 9.Rb5 Kf4 10.Rb4+ Ke5 11.Ra4 f5 12.Ke3 fxg4 13.Rxg4 Kd5 14.Kf4 Kd4 15.Kg5+ Kd3 16.Ra4 (0:10:34) 25713kN
[+1.53]  d=25  1.Re1 Kf6 2.Rb1 Ke5 3.f3 Ra6 4.Rb5+ Kd4 5.g4 hxg4 6.fxg4 Kc4 7.Rg5 Kd4 8.Kf2 Ke4 9.Rb5 f5 10.gxf5 gxf5 11.Rb4+ Ke5 12.Ra4 Ke6 13.Kf3 Ke5 14.h5 Rh6 15.a6 Rxh5 16.Ra5+ (0:06:53) 17601kN
[+1.41]  d=25  1.Kg2 Kf6 2.Rc8 Kf5 3.Ra8 Rd3 4.Ra7 Ke6 5.Ra6+ Ke5 6.Rb6 Ra3 7.a6 Ra2 8.Kf1 Kd4 9.Kg1 Kc5 10.Rf6 Kd5 11.Kg2 Ke5 12.Rb6 Ra1 13.f3 Ra2+ 14.Kh3 Ra3 15.Rc6 Kd5 16.Rf6 (0:03:36) 10066kN
[+1.41]  d=24  1.Kg2 Kf6 2.Rc8 Kf5 3.Ra8 Rd3 4.Ra7 Ke6 5.Ra6+ Ke5 6.Rb6 Ra3 7.a6 Ra2 8.Kf1 Kd4 9.Kg1 Kc5 10.Rf6 Kd5 11.Kg2 Ke5 12.Rb6 Ra1 13.f3 Ra2+ 14.Kh3 Ra3 15.Kg2 Ra2+ 16.Kh3 (0:02:49) 7783kN
[+1.41]  d=23  1.Kg2 Rd2 2.Kf3 Ra2 3.Re5 Kf6 4.Rb5 Ke6 5.Rb6+ Ke5 6.a6 Ra3+ 7.Kg2 Kd5 8.Kh3 Ke4 9.Kh2 Kd5 10.Kh3 Ke4 11.Kh2 Kd5 12.Kh3 Ke4 13.Kh2 Kd5 14.Kh3 Ke4 15.Kh2 Kd5 16.Kh3 (0:01:00) 2788kN
[+1.41]  d=22  1.Kg2 Rd2 2.Kf3 Ra2 3.Re5 Kf6 4.Rb5 Ke6 5.Rb6+ Ke5 6.a6 Ra3+ 7.Kg2 Kd5 8.f4 Kc5 9.Rf6 (0:00:43) 1995kN
[+1.42]  d=21  1.Kg2 Rd2 2.Kf3 Ra2 3.Re5 Kf6 4.Rb5 Ke6 5.Rb6+ Ke5 6.a6 Ra3+ 7.Kg2 Kd5 8.f4 Kc5 9.Rf6 (0:00:35) 1563kN
[+1.41]  d=20  1.Kg2 Rd2 2.Kf3 Ra2 3.Re5 Kf6 4.Rb5 Ke6 5.Rb6+ Ke5 6.a6 Ra3+ 7.Kg2 Kd5 8.f4 Kc5 9.Rf6 (0:00:28) 1283kN
[+1.41]  d=19  1.Kg2 Rd2 2.Kf3 Ra2 3.Re5 Kf6 4.Rb5 Ke6 5.Rb6+ Ke5 6.a6 Ra3+ 7.Kg2 Kd5 8.f4 Kc5 9.Rf6 (0:00:23) 1007kN
[+1.41]  d=18  1.Kg2 Rd2 2.Kf3 Ra2 3.Re5 Kf6 4.Rb5 Ke6 5.Rb6+ Ke5 6.a6 Ra3+ 7.Kg2 Kd5 8.f4 Kc5 9.Rf6 Kb5 10.Rxf7 Rxa6 11.f5 Kc4 12.f6 (0:00:13) 559kN
[+1.48]  d=17  1.Kg2 Rd2 2.Kf3 Ra2 3.Re5 Kf6 4.Rb5 Ke6 5.Rb6+ Ke5 6.a6 Ra3+ 7.Kg2 Kd5 8.f4 Ra2+ 9.Kh3 Kc5 10.Rf6 Kb5 11.f5 Rxa6 12.Rxf7 Kc4 13.f6 (0:00:08) 329kN
Parent - By Akbarfan (***) Date 2010-05-26 13:57

> blazed ahead at 40 Kn/s.


40 Kn/s ?  Which Processor and how much GHz ?

How much Hashtables you used?
Parent - - By yanquis1972 (****) [us] Date 2010-05-26 15:09
i'm afraid i can't replicate this behavior....

Engine: Rybka 4
by Vasik Rajlich

17.14  0:01   +1.47    1.Kf1 Kf6 2.Rb8 Kf5 3.Rb6 Rd1+ 4.Kg2 Ra1
                       5.a6 Ke5 6.Rc6 Ra3 7.Rb6 Kd5 8.f4 Ke4
                       9.Kh3 Kf3 10.Rc6 Ra1 11.Rc3+ Ke4 (543.669) 297

18.01  0:02   +1.40    1.Kf1 Kf6 2.Rb8 Rd5 3.Ra8 Rd2 4.a6 Ra2
                       5.Kg2 Ke5 6.Kf3 Kf5 7.Ke3 Kg4 8.a7 Ra3+
                       9.Kd4 Kf3 10.Kc5 Kxf2 11.Kb6 Rb3+
                       12.Kc7 Rc3+ 13.Kb7 Rb3+ (810.508) 316

18.02  0:02   +1.46    1.Rb8 Rd1+ 2.Kg2 Ra1 3.Rb5 Ra3 4.f3 Ra2+
                       5.Kh3 Ra3 6.f4 Kf6 7.Re5 Ra4 8.Rb5 Ke7
                       9.f5 Kd6 10.fxg6 fxg6 11.Rb6+ Kc7
                       12.Rxg6 Rxa5 13.Rg5 Ra2 14.Rxh5 (916.678) 312

19.01  0:03   +1.43    1.Rb8 Rd1+ 2.Kg2 Ra1 3.Rb5 Ra3 4.f3 Ra2+
                       5.Kh3 Ra3 6.f4 Kf6 7.Re5 Ra4 8.Kg2 Ra2+
                       9.Kh3 Ra3 10.Kg2 Ra2+ 11.Kh3 Ra3
                       12.Kg2 Ra2+ 13.Kh3 Ra3 14.Kg2 (1.003.256) 312

19.03  0:03   +1.46    1.Rc8 Rd1+ 2.Kg2 Ra1 3.Rc5 Kf6 4.Kf3 Ke6
                       5.Rb5 Kd6 6.Ke3 Ra3+ 7.Kf4 Ra2 8.f3 Kc7
                       9.Rc5+ Kb7 10.Re5 Ka6 11.Re7 (1.012.220) 311

20.01  0:03   +1.46    1.Rc8 Rd1+ 2.Kg2 Ra1 3.Rc5 Kf6 4.Kf3 Ke6
                       5.Rb5 Kd6 6.Ke3 Ra3+ 7.Kf4 Ra2 8.f3 Kc7
                       9.Rc5+ Kd6 10.Rb5 Kc7 11.Rc5+ Kd6
                       12.Rb5 Kc7 13.Rc5+ Kd6 14.Rb5 (1.122.728) 311

21.01  0:04   +1.40    1.Rc8 Rd1+ 2.Kg2 Ra1 3.Rc5 Kf6 4.Kf3 Ke6
                       5.Rb5 Kd6 6.Ke3 Ra3+ 7.Kf4 Ra2 8.f3 Kc7
                       9.Ke4 Ra4+ 10.Kd5 Ra3 11.Ke4 Ra4+
                       12.Kd5 Ra3 13.Ke4 Ra4+ 14.Kd5 (1.397.692) 315

22.01  0:07   +1.39    1.Rc8 Rd1+ 2.Kg2 Ra1 3.Rc5 Ra3 4.Re5 Kf6
                       5.Rb5 Ke6 6.Rb6+ Kd5 7.a6 Ra5 8.Rf6 Ke5
                       9.Rc6 Ra3 10.Kh3 Kd5 11.Rf6 Ke4
                       12.Rb6 Kf3 13.Rf6+ Ke4 14.Rb6 (2.362.245) 322

22.04  0:13   +1.40    1.Re5 Rd2 2.Kg2 Ra2 3.Kf3 Kf6 4.Rb5 Ke7
                       5.Ke4 Kd7 6.Ke3 Kc7 7.f3 Ra3+ 8.Ke4 Ra4+
                       9.Kd5 Ra3 10.Ke4 Ra4+ 11.Kd5 Ra3
                       12.Ke4 Ra4+ 13.Kd5 Ra3 14.Ke4 (4.525.024) 331

23.01  0:15   +1.40    1.Re5 Rd2 2.Kg2 Ra2 3.Kf3 Kf6 4.Rb5 Ke7
                       5.Ke4 Kd7 6.Ke3 Kc7 7.f3 Ra3+ 8.Ke4 Ra4+
                       9.Kd5 Ra3 10.Rc5+ Kb7 11.Ke4 Ra4+
                       12.Ke5 Ra3 13.Kf4 Ra4+ 14.Kg5 (5.013.734) 328

24.01  0:23   +1.42    1.Re5 Kf6 2.Rb5 Rd1+ 3.Kg2 Rd3
                       4.Rb6+ Ke5 5.f3 Rd2+ 6.Kh3 Rd3 7.f4+ Ke4
                       8.Rb4+ Ke3 9.Ra4 Rd7 10.a6 Ra7
                       11.Kg2 f5 12.Ra5 Kd4 13.Kf3 Kc4
                       14.Ke3 (7.643.580) 320

25.01  0:30   +1.39    1.Re5 Kf6 2.Rb5 Rd1+ 3.Kg2 Rd3
                       4.Rb6+ Ke5 5.f3 Rd2+ 6.Kh3 Rd3 7.a6 Ra3
                       8.Kh2 Ra2+ 9.Kg1 Kd4 10.g4 Kc5
                       11.Rf6 Kb5 12.Rxf7 Rxa6 13.gxh5 (9.693.798) 320

25.04  1:25   +1.54++  1.Kf1 Kf6 2.Rb8 Rd3 3.a6 Ra3 4.Ra8 Kf5
                       5.Ra7 Ra1+ 6.Kg2 Ke5 7.Kf3 Ra3+
                       8.Ke2 f6 9.Ra8 Ra2+ 10.Kf3 Kf5 11.a7 Ra3+
                       12.Kg2 Ra2 13.Kf3 Ra3+ 14.Kg2 (25.702.244) 301

25.01  1:37   +1.69++  1.Kf1 Kf6 2.Rb8 Rd3 3.a6 Ra3 4.Ra8 Kf5
                       5.Ra7 Ra1+ 6.Kg2 Ke5 7.Kf3 Ra3+
                       8.Ke2 f6 9.Ra8 Ra2+ 10.Kf3 Kf5 11.a7 Ra3+
                       12.Kg2 Ra2 13.Kf3 Ra3+ 14.Kg2 (28.459.126) 292

25.01  3:19   +1.84    1.Kf1 Ra6 2.Re5 Rc6 3.Rb5 Rc2 4.a6 Ra2
                       5.Rb6 Kf8 6.Kg2 Ra3 7.Kh3 Ke8 8.f3 Kf8
                       9.Kg2 Ra2+ 10.Kf1 Ke8 11.g4 Kd8
                       12.gxh5 gxh5 13.f4 Kd7 14.f5 (55.936.357) 280

25.05  3:36   +2.09++  1.Ra8 Kf6 2.Kf1 Rd2 3.a6 Ra2 4.a7 (60.090.292) 278

25.01  4:48   +2.31    1.Ra8 Kf6 2.Kf1 Ke5 3.a6 Re6 4.a7 Re7
                       5.Ke2 Kf5+ 6.Kf3 Rd7 7.Kg2 Re7 8.f3 Ke5
                       9.Kh2 Kf5 10.Kh3 Ke6 11.Kg2 Ke5
                       12.Kh2 Kf5 13.Kh3 Ke6 14.Kg2 (81.087.117) 281

26.01  6:28   +2.43++  1.Ra8 Kf6 2.f3 Ke5 3.a6 Re6 4.Kg2 Kf6
                       5.Kf1 Re3 6.Kf2 Ra3 7.a7 Ra2+ 8.Ke3 Ra3+
                       9.Kd4 Kf5 10.Kc5 Ra6 11.Kb5 Ra2
                       12.Kb6 Rb2+ 13.Kc6 Ra2 14.Kc7 (105.047.148) 270

26.01  6:57   +2.47    1.Ra8 Kf6 2.f3 Ke5 3.a6 Re6 4.Kg2 Kf6
                       5.a7 Re2+ 6.Kh3 Re7 7.g4 Ke5 8.Kg3 f5
                       9.g5 f4+ 10.Kg2 Ke6 11.Kh1 (111.681.628) 267

27.01  7:56   +2.47    1.Ra8 Kf6 2.f3 Ke5 3.a6 Re6 4.Kg2 Kf6
                       5.a7 Re2+ 6.Kg1 Re7 7.Kf1 Kf5 8.Kf2 f6
                       9.Kf1 Rd7 10.Ke2 Rb7 11.Kd2 Rb2+
                       12.Ke3 Rb3+ 13.Kd4 Ra3 14.Kc5 (125.679.069) 264

28.01  8:40   +2.47    1.Ra8 Kf6 2.f3 Ke5 3.a6 Re6 4.Kg2 Kf6
                       5.a7 Re2+ 6.Kg1 Re7 7.Kf1 Kf5 8.Kf2 f6
                       9.Kf1 Rd7 10.Ke2 Rb7 11.Kd2 Rb2+
                       12.Ke3 Rb3+ 13.Kd4 Ra3 14.Kc5 (137.462.071) 264

28.01  8:59   +2.47    1.Ra8 Kf6 2.f3 Ke5 3.a6 Re6 4.Kg2 Kf6
                       5.a7 Re2+ 6.Kg1 Re7 7.Kf1 Kf5 8.Kf2 f6
                       9.Kf1 Rd7 10.Ke2 Rb7 11.Kd2 Rb2+
                       12.Ke3 Rb3+ 13.Kd4 Ra3 14.Kc5 (142.955.315) 264

best move: Re8-a8 time: 8:59.857 min  n/s: 264.801  nodes: 142.955.315
Parent - - By yanquis1972 (****) [us] Date 2010-05-26 15:11
shredder finds the correct continuation instantly:

4R3/5pk1/3r2p1/P6p/7P/6P1/5P2/6K1 w - -


Engine: Deep Shredder 12 x64 (1024 MB)
by Stefan Meyer-Kahlen

16/25  0:00   +1.10    1.Re1 Kf6 2.Ra1 Ra6 3.Kg2 Ke5 4.Kf3 Kd4
                       5.Kf4 Kc4 6.Kg5 Kb5 7.f4 f5 8.Ra2 Kc4
                       9.Ra3 Kd4 10.Ra4+ Ke3 11.Ra1 Kd4 (1.946.072) 2772 SB:363

17/34  0:01   +0.90    1.Re1 Kf6 2.Ra1 Ra6 3.Kg2 Ke5 4.Kf3 Kd4
                       5.Ra4+ Kc5 6.Ke4 Kb5 7.Ra1 Kc4 8.f4 (5.166.268) 2904
                       SB:11.039

18/34  0:02   +0.90    1.Re1 Kf6 2.Ra1 Ra6 3.Kg2 Ke5 4.Ra4 Kd6
                       5.Kf3 Kc5 6.Ke4 Kb5 7.Ra1 Kc4 8.f4 (8.367.190) 3013
                       SB:17.423

19/32  0:03   +0.90    1.Re1 Kf6 2.Ra1 Ra6 3.Kg2 Ke5 4.Ra4 Kd6
                       5.Kf3 Kc5 6.Ke4 Kb5 7.Ra1 Kc4 8.f4 (9.503.816) 3046
                       SB:20.634

20/35  0:03   +1.01    1.Re1 Kf6 2.Ra1 Ra6 3.Kg2 Ke5 4.Ra4 Kd6
                       5.Kf3 Kc5 6.Kf4 Kb5 7.Ra1 Kb4 8.Kg5 Kb3
                       9.Kh6 Kb2 10.Ra4 Kb3 11.Rf4 f5
                       12.Rd4 Kc3 13.Ra4 (11.797.032) 3086
                       SB:26.438

21/39  0:05   +1.02    1.Re1 Kf6 2.Ra1 Ra6 3.Kg2 Ke5 4.Ra4 Kd6
                       5.Kf3 Kc5 6.Kf4 Kb5 7.Ra1 Kb4 8.Kg5 Kb3
                       9.Kh6 Kb2 10.Ra4 Kb3 11.Rf4 f5
                       12.Rd4 Kc3 13.Ra4 Kd3 (16.114.598) 3130
                       SB:41.505

22/37  0:06   +1.09    1.Re1 Kf6 2.Ra1 Ra6 3.Kg2 Ke5 4.Ra4 Kd6
                       5.Kf3 Kc5 6.Kf4 Kb5 7.Ra1 Kb4 8.Kg5 Kb3
                       9.Kh6 Kb2 10.Ra4 Kb3 11.Rf4 f5
                       12.Rd4 Kc3 13.Ra4 Kd3 (20.745.572) 3189
                       SB:53.508

23/43  0:10   +1.06    1.Re1 Kf6 2.Ra1 Ra6 3.Kg2 Ke5 4.Ra4 Kd6
                       5.Kf3 Kc5 6.Kf4 Kb5 7.Ra1 Kb4 8.Kg5 Kb3
                       9.Kh6 Kb2 10.Ra4 Kb3 11.Rf4 f5
                       12.Rd4 Kc3 13.Ra4 Kd3 14.Ra1 (35.014.817) 3210
                       SB:177.876

24/43  0:15   +1.06    1.Re1 Kf6 2.Ra1 Ra6 3.Kg2 Ke5 4.Ra4 Kd6
                       5.Kf3 Kc5 6.Kf4 Kb5 7.Ra1 Kb4 8.Kg5 Kb3
                       9.Kh6 Kb2 10.Ra4 Kb3 11.Rf4 f5
                       12.Rd4 Kc3 13.Ra4 Kd3 14.Ra1 (50.694.792) 3182
                       SB:301.593

25/43  0:22   +1.12    1.Re1 Kf6 2.Ra1 Ra6 3.Kg2 Ke5 4.Ra4 Kd6
                       5.Kf3 Kc5 6.Kf4 Kb5 7.Ra1 Kb4 8.Kg5 Kb3
                       9.Kh6 Kb2 10.Ra4 Kb3 11.Rf4 f5
                       12.Rd4 Kc3 13.Ra4 (70.633.631) 3116
                       SB:471.783

best move: Re8-e1 time: 0:26.395 min  n/s: 3.130.509  CPU 381.2%   n/s(1CPU): 821.224  nodes: 82.582.842 SB: 508.369
Parent - - By Moz (****) Date 2010-05-29 04:26

>shredder finds the correct continuation instantly:


Stockfish 1.71 finds Re1 instantly but seemed reluctant to follow it up with Ra1 until I enabled the pawn endgame extensions at which point all became clear:

[+2.50]  d=35  1.Re1 Kf6 2.Ra1 Ra6 3.f3 Ke5 4.Ra4 Kd5 5.Kf2 Ke5 6.Ke3 f6 7.g4 hxg4 8.fxg4 Ke6 9.Kf4 Kf7 10.Ra1 Ke7 11.Ke4 Ke6 12.Kd4 f5 13.gxf5+ gxf5 14.Ke3 Kd5 15.Kf4 Rh6 16.Kg5 Rh8 17.a6 Ke5 18.Re1+ Kd5 19.h5 f4 20.Kxf4 Rxh5 21.Rd1+ Kc6 22.Rc1+ Kd5 23.a7 Rh8 (0:01:45) 863382kN

[+2.30]  d=34  1.Re1 Kf6 2.Ra1 Ra6 3.f3 Ke5 4.Ra4 Kd5 5.Kf2 Ke5 6.Ke3 f6 7.Ra3 Kd5 8.Ra1 Kc4 9.g4 Kd5 10.gxh5 gxh5 11.Kd3 Kd6 12.Ke4 Ke6 13.Ra4 Ke7 14.Ra3 Ke6 15.Ra1 Ra8 16.a6 Ra7 17.Kd4 Kd6 18.Ra5 Ra8 19.Ke4 Kc7 20.Kf5 Kb6 21.Ra1 Ka7 22.Kxf6 Rf8+ 23.Kg5 Rxf3 24.Kxh5 (0:01:14) 611236kN

[+2.30]  d=33  1.Re1 Kf6 2.Ra1 Ra6 3.f3 Ke5 4.Ra4 Kd5 5.Kf2 Ke5 6.Ke3 f6 7.Ra3 Kd5 8.Ra1 Kc4 9.g4 Kd5 10.gxh5 gxh5 11.Kd3 Kd6 12.Ke4 Ke6 13.Ra4 Ke7 14.Ra3 Ke6 15.Ra1 Ra8 16.a6 Ra7 17.Ra4 Ra8 18.Kd4 Kd7 19.Ra1 Kd6 20.Ke4 Kc6 21.Kf4 Kb6 22.Kf5 Ka7 23.Kxf6 Rf8+ 24.Kg5 Rxf3 25.Kxh5 (0:01:08) 562395kN

[+2.30]  d=32  1.Re1 Kf6 2.Ra1 Ra6 3.f3 Ke5 4.Ra4 Kd5 5.Kf2 Ke5 6.Ke3 f6 7.Ra3 Kd5 8.Ra1 Kc4 9.g4 Kd5 10.gxh5 gxh5 11.Kd3 Kd6 12.Ke4 Ke6 13.Ra4 Ke7 14.Ra3 Ra8 15.a6 Kd6 16.a7 Kc6 17.Ra4 Re8+ 18.Kf5 Ra8 19.Kxf6 Kb5 20.Ra1 (0:00:57) 476776kN

[+2.30]  d=31  1.Re1 Kf6 2.Ra1 Ra6 3.f3 Ke5 4.Ra4 Kd5 5.Kf2 Ke5 6.Ke3 f6 7.Ra3 Kd5 8.Ra1 Kc4 9.g4 Kd5 10.gxh5 gxh5 11.Kd3 Kd6 12.Kd4 f5 13.Kc4 Kc6 14.Kb4 Ra7 15.Rc1+ Kd5 16.Rc5+ Ke6 17.f4 Rb7+ 18.Kc4 Ra7 19.Re5+ Kf6 20.Kd4 Rd7+ 21.Kc4 Ra7 (0:00:22) 176641kN

[+2.10]  d=30  1.Re1 Kf6 2.Ra1 Ra6 3.f3 Ke5 4.Ra4 Kd5 5.Kf2 Ke5 6.Ke3 f6 7.Ra3 Kd5 8.Ra1 Kc4 9.g4 f5 10.gxh5 gxh5 11.Kf4 Kb3 12.Re1 Rxa5 13.Re5 Ra4+ 14.Kg5 Kc4 15.Kxh5 Kd4 16.Rxf5 Ra2 17.Kg5 Rg2+ 18.Kf4 Rh2 19.Kg5 (0:00:13) 102162kN

[+1.97]  d=29  1.Re1 Kf6 2.Ra1 Ra6 3.f3 Ke5 4.Ra4 Kd5 5.Kf2 Ke5 6.Ke3 Kd5 7.g4 Ke5 8.Ra1 hxg4 9.fxg4 f6 10.Ra2 f5 11.gxf5 Kxf5 12.Rg2 Kf6 13.Rg5 Ke7 14.Ke4 Kd8 15.Kd3 Kd7 16.Kd4 Ra8 17.Rb5 Kc7 18.Rg5 Rd8+ 19.Ke4 Ra8 (0:00:07) 53929kN

[+1.97]  d=28  1.Re1 Kf6 2.Ra1 Ra6 3.f3 Ke5 4.Ra4 Kd5 5.Kf2 f6 6.Ke3 Re6+ 7.Kd3 Ra6 8.Ra1 Kc6 9.g4 hxg4 10.fxg4 Kd5 11.Ke3 Ke5 12.Ra2 f5 13.gxf5 Kxf5 14.Rg2 Kf6 15.Rg5 Ke7 16.Ke4 Kd7 17.Kd4 Ra8 18.Rb5 Rh8 19.Rb7+ Ke6 20.Rb6+ (0:00:05) 38173kN
Parent - By yanquis1972 (****) [us] Date 2010-05-29 04:30
interesting -- it seems like they're already enabled (default setting 2) -- what change did you make?
Parent - - By bnc (***) Date 2010-05-26 16:29
I am also getting this same bad endgame behavior in Deep Rybka 4. Just like with R3, R4 doesnt understand that rooks belong behind passed pawns, not in front of them.

This is very disappointing  :( :( :(
Parent - By yanquis1972 (****) [us] Date 2010-05-26 16:30
im about to make a post on the main forum about this; i think the rook endgame scaling parameter may help.
Parent - - By Moz (****) Date 2010-05-29 04:08 Edited 2010-05-29 04:14

> In Response to Dadi Jonsson concerning the following position:


4R3/5pk1/3r2p1/P6p/7P/6P1/5P2/6K1 w - -


> This is a rook ending which most engines have problems with, including Rybka 3.


With EGTB's and default Rook Endgame Scaling on a multicore system R4 never found Re1.  Changing the Rook Endgame Scaling to 50 was enough to clarify the position however Rybka still has trouble with the subsequent moves!  Ridiculous.  But at least it's progress.

[+0.83]  d=25  1.Re1 Ra6 2.Re5 Rc6 3.Rb5 Rc1+ 4.Kg2 Ra1 5.Kf3 Ra4 6.Ke3 Kf8 7.Ke2 Ra2+ 8.Kf3 Ke7 9.Rb7+ Ke6 10.Rb6+ Kd5 11.a6 Kd4 12.Kf4 Rxf2+ 13.Kg5 Ra2 14.Kh6 Ra3 15.Rf6 Rxg3 16.Rxf7 (0:00:21) 6506kN
[+0.83]  d=24  1.Re1 Ra6 2.Re5 Rc6 3.Rb5 Rc1+ 4.Kg2 Ra1 5.Kf3 Ra4 6.Ke3 Kf8 7.Ke2 Ra2+ 8.Kf3 Ke7 9.Rb7+ Ke6 10.Rb6+ Kd5 11.a6 Kd4 12.Kg2 Ke5 13.Rc6 Kd5 14.Rf6 Ke5 15.Rb6 Kd4 16.Kf3 (0:00:12) 4054kN
[+0.83]  d=23  1.Re1 Ra6 2.Re5 Rc6 3.Rb5 Rc1+ 4.Kg2 Ra1 5.Kf3 Ra4 6.Ke3 Kf8 7.Ke2 Ra2+ 8.Kf3 Ke7 9.Rb7+ Ke6 10.Rb6+ Kd5 11.a6 Kd4 12.Kg2 Ke5 13.Rc6 Kd5 14.Rf6 Ke5 15.Rb6 Kd4 16.Rc6 (0:00:09) 3085kN
[+0.83]  d=22  1.Re1 Ra6 2.Re5 Rc6 3.Rb5 Rc1+ 4.Kg2 Ra1 5.Kf3 Ra4 6.Ke3 Kf8 7.Ke2 Ra2+ 8.Kd3 Ke7 9.Ke4 Kd6 10.f3 Ra4+ 11.Ke3 Ra3+ 12.Kf4 Kc6 13.Rb6+ Kc5 14.Rf6 Rxa5 15.Rxf7 Ra6 16.Kg5 (0:00:07) 2369kN
[+0.83]  d=21  1.Re1 Ra6 2.Re5 Rc6 3.Rb5 Rc1+ 4.Kg2 Ra1 5.Kf3 Ra4 6.Ke3 Kf8 7.Ke2 Kg8 8.f3 Kg7 9.Ke3 Ra3+ 10.Ke4 Ra4+ 11.Kd5 Ra3 12.Kc4 Rxf3 13.Rb3 Rf6 14.Rb6 Rf1 15.a6 (0:00:05) 1738kN
[+0.83]  d=20  1.Re1 Ra6 2.Re5 Rc6 3.Rb5 Rc1+ 4.Kg2 Ra1 5.Kf3 Ra4 6.Ke3 Kf8 7.Ke2 Kg8 8.f3 Kg7 9.Ke3 Ra3+ 10.Ke4 Ra4+ 11.Kd5 Ra3 12.Kc4 Rxf3 13.Rb3 Rf6 14.Rb6 Rf1 15.a6 (0:00:05) 1455kN
[+0.83]  d=19  1.Re1 Ra6 2.Re5 Rc6 3.Rb5 Rc1+ 4.Kg2 Ra1 5.Kf3 Ra4 6.Ke3 Kf8 7.Ke2 Kg8 8.f3 Kg7 9.Ke3 Ra3+ 10.Ke4 Ra4+ 11.Kd5 Ra3 12.Kc4 Rxf3 13.Rb3 Rf6 14.Rb6 (0:00:04) 1306kN
[+0.82]  d=18  1.Re1 Ra6 2.Re5 Kf6 3.Rb5 Ke7 4.f4 Kd7 5.Kf2 Kc6 6.Re5 Ra7 7.Kf3 Rd7 8.f5 (0:00:03) 1037kN
[+0.83]  d=17  1.Re1 Ra6 2.Re5 Kf6 3.Rb5 Ke7 4.Kg2 Kd7 5.Kf3 Ra7 6.Kf4 Kc6 7.Re5 Rd7 8.f3 Rd4+ 9.Re4 Rd7 10.Re8 Kb5 11.Re5+ (0:00:03) 1011kN
[+0.82]  d=17  1.Kg2 Rd4 2.Re5 Ra4 3.Kf3 Kf8 4.Rb5 Kg8 5.Ke3 Kg7 6.f3 Ra3+ 7.Ke4 Ra4+ 8.Kd3 Ra3+ 9.Ke4 Ra4+ 10.Kd3 Ra3+ 11.Ke4 Ra4+ 12.Kd3 Ra3+ 13.Ke4 Ra4+ 14.Kd3 Ra3+ 15.Ke4 Ra4+ 16.Kd3 (0:00:03) 937kN
[+0.74]  d=17  1.Ra8 Rd4 2.Rb8 Ra4 3.Rb5 Kf8 4.Kf1 Ke7 5.Rb7+ Kf8 6.Ra7 Kg7 7.Ke2 Kf6 8.a6 Ke6 9.Ke3 Kd5 10.Ra8 Kc6 11.f3 Ra3+ 12.Kf4 Ra4+ 13.Ke5 Ra5+ 14.Kf4 Ra4+ 15.Ke5 Ra5+ 16.Kf4 (0:00:02) 665kN
[+0.83]  d=16  1.Ra8 Rd1+ 2.Kg2 Rd3 3.Kf1 Rd2 4.a6 Ra2 5.a7 Kf6 6.Kg2 Ra3 7.Kh2 Ra4 8.Kg2 Ra3 9.Kh2 Ra4 10.Kg2 Ra3 11.Kh2 Ra4 12.Kg2 Ra3 13.Kh2 Ra4 14.Kg2 Ra3 15.Kh2 Ra4 16.Kg2 (0:00:00) 177kN
[+0.83]  d=15  1.Ra8 Rd1+ 2.Kg2 Rd3 3.Kf1 Rd2 4.a6 Ra2 5.a7 Kf6 6.Kg2 Ra3 7.Kh2 Ra2 8.Kg1 Ke7 9.Kg2 Ra3 10.f4 Kf6 11.Kf2 Kf5 (0:00:00) 92kN
[+0.83]  d=14  1.Ra8 Rd1+ 2.Kg2 Rd3 3.Kf1 Rd2 4.a6 Ra2 5.a7 Kf6 6.Kg1 Ra1+ 7.Kg2 Ra3 8.Kf1 Ra2 9.Kg1 Ra1+ 10.Kg2 Ra3 11.Kf1 Ra2 12.Kg1 Ra1+ 13.Kg2 Ra3 14.Kf1 Ra2 15.Kg1 Ra1+ 16.Kg2 (0:00:00) 55kN
[+0.83]  d=13  1.Ra8 Rd1+ 2.Kg2 Rd3 3.Kf1 Rd2 4.a6 Ra2 5.a7 Kf6 6.Kg1 Ra1+ 7.Kg2 Ra3 8.Kf1 Ra2 9.Kg1 Ra1+ 10.Kg2 Ra3 11.Kf1 Ra2 12.Kg1 Ra1+ 13.Kg2 Ra3 14.Kf1 Ra2 15.Kg1 Ra1+ 16.Kg2 (0:00:00) 31kN
[+0.83]  d=12  1.Ra8 Rd1+ 2.Kg2 Rd3 3.Kf1 Rd2 4.a6 Ra2 5.a7 Kf6 6.Kg1 Ra1+ 7.Kg2 Ra3 8.Kf1 Ra2 9.Kg1 Ra1+ 10.Kg2 Ra3 11.Kf1 Ra2 12.Kg1 Ra1+ 13.Kg2 Ra3 14.Kf1 Ra2 15.Kg1 Ra1+ 16.Kg2 (0:00:00) 20kN
[+0.83]  d=11  1.Ra8 Rd1+ 2.Kg2 Rd3 3.Kf1 Rd2 4.a6 Ra2 5.a7 Kf6 6.Kg1 Ra1+ 7.Kg2 Ra3 8.Kf1 Ra2 9.Kg1 Ra1+ 10.Kg2 Ra3 11.Kf1 Ra2 12.Kg1 Ra1+ 13.Kg2 Ra3 14.Kf1 Ra2 15.Kg1 Ra1+ 16.Kg2 (0:00:00) 14kN
[+0.83]  d=10  1.Ra8 Rd1+ 2.Kg2 Rd3 3.Kf1 Rd2 4.a6 Ra2 5.a7 Kf6 6.Kg1 Ra1+ 7.Kg2 Ra3 8.Kf1 Ra2 9.Kg1 Ra1+ 10.Kg2 Ra3 11.Kf1 Ra2 12.Kg1 Ra1+ 13.Kg2 Ra3 14.Kf1 Ra2 15.Kg1 Ra1+ 16.Kg2 (0:00:00) 11kN
[+0.84]  d=9  1.Ra8 Rd1+ 2.Kg2 Rd3 3.Kf1 Ra3 4.Ke2 Ra2+ 5.Kf3 (0:00:00) 8kN
[+0.84]  d=8  1.Ra8 Rd1+ 2.Kg2 Rd3 3.Rb8 Ra3 4.Rb5 Ra2 5.Kf3 Kf6 6.Ke3 Ke7 (0:00:00) 4kN


RES is going to be invaluable to corr players.
Parent - By Uly (Gold) [mx] Date 2010-05-29 04:44

> RES is going to be invaluable to corr players.


Invaluable to corr players that rely exclusively on R4.
Parent - - By yanquis1972 (****) [us] Date 2010-05-29 18:16
i noticed that very early on, but it's by the slimmest of margins. if you check the MPV i believe it scores multiple moves at exactly that (or very, very close).  however i have my RES set to 60 by default atm.  i find it very hard to believe 100 is the best for analysis, & even 60 may be too high.
Parent - By Moz (****) Date 2010-05-30 03:16

> i noticed that very early on, but it's by the slimmest of margins. if you check the MPV i believe it scores multiple moves at exactly that (or very, very close).  however i have my RES set to 60 by default atm.  i find it very hard to believe 100 is the best for analysis, & even 60 may be too high.


I hope it's a good sign that you and I are on the same page.  I've got mine tentatively set at 50 since 100 "felt" too high.  Having said that, I think there's more to RES than I initially thought.  Take a look at the following position where the side with the rooks is clearly winning and notice how little RES impacts the eval. 

8/1kp3pp/p1p2pb1/P1P5/4rB2/1K5P/1PP2QP1/3r4 w - -


I assume that RES only come into play in positions with fewer than X pieces?  Or some defined set of positions that qualifies as endgame? Does anyone know what that is exactly?  Any info would be appreciated.
- - By Dadi Jonsson (Silver) [is] Date 2010-05-26 10:16
Blind bishop example.

r7/8/8/5bk1/8/5B2/5RPP/6K1 b - - 0 1


Domelius wrote:

> Rybka cannot find the line 1...Ra1+ 2. Rf1 Rxf1+ 3. Kxf1 Bh3!!. The move 3... Bh3!! is the only move which saves the game for back


Rybka 4 1-CPU 32-bit without EGTB finds the drawing line and gives the correct evaluation:

[+0.00]  d=18  1...Ra1+ 2.Rf1 Rxf1+ 3.Kxf1 Bh3 4.Kf2 Bxg2 5.Ke3 Bh3 6.Be4 Be6 7.Kf3 Kh6 8.Bd3 Bh3 9.Kf4 Kg7 10.Kg5 Bd7 (0:00:29) 1085kN
[+0.00]  d=17  1...Ra1+ 2.Rf1 Rxf1+ 3.Kxf1 Bh3 4.Kf2 Bxg2 5.Ke3 Bh3 6.Be4 Be6 7.Kf3 Kh6 8.Bd3 Bh3 9.Kf4 Kg7 10.Kg5 Bd7 (0:00:28) 1056kN
[+0.69]  d=17  1...Ra1+ 2.Rf1 Rxf1+ 3.Kxf1 Bh3 (0:00:27) 994kN
[+1.49]  d=17  1...Ra1+ 2.Rf1 Rxf1+ 3.Kxf1 Bh3 (0:00:26) 969kN
[+1.89]  d=17  1...Ra1+ 2.Rf1 Rxf1+ 3.Kxf1 Bh3 4.Kf2 Bxg2 5.Kxg2 Kh6 6.Kg3 Kg7 (0:00:26) 957kN
[+2.04]  d=17  1...Ra1+ 2.Rf1 Rxf1+ 3.Kxf1 Bh3 4.gxh3 Kh6 5.Kf2 Kh7 (0:00:26) 947kN
[+2.16]  d=16  1...Ra1+ 2.Rf1 Ra4 3.Re1 Ra2 4.Re2 Ra5 5.Kf2 Bg4 6.Re4 Bxf3 7.Kxf3 Ra3+ 8.Re3 Ra5 9.Rc3 Rf5+ 10.Kg3 Re5 11.h4+ Kh5 12.Kh3 Re4 13.Rc5+ Kg6 (0:00:13) 434kN
Parent - - By Vempele (Silver) [fi] Date 2010-05-26 11:13
Congratulations on your crowning. :)
Parent - By Dadi Jonsson (Silver) [is] Date 2010-05-26 11:52
Thanks. Like some other crowned members, I hadn't noticed this myself :)
Parent - - By Uly (Gold) [mx] Date 2010-05-26 18:56
Oh, +1

I also didn't notice ;)
Parent - By Geomusic (*****) Date 2010-05-28 07:02
vytron 14k posts? lol
Up Topic The Rybka Lounge / Test Positions / Rybka 4 vs Rybka 3

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill