< Earlier Kibitzing · PAGE 4 OF 4 ·
|Jan-07-12|| ||James D Flynn: There are several candidates my first was d3 to open the b6-f1 diagonal for the Q. It looks quite promising e.g 22.Bxd3 Qc5ch 23.Ke2 loses to Rxh4 24.g3xh4 f4 25.Ke1 Bg4 but 23.Ke1 and I didnt see an immediate win. My next was f4 intending to take on g3 and sac the exchange to bring my Q in at g3 after thew Ng4ch. But why sac anything? Re4 looks very strong e.g 22.Bd3 Ng4ch 23.Kf3 Re3ch 24.Kg2 Bc6ch 25.Kh3 Nf2# so I settled on this and checked the game with the thought that maybe 22.h3 may hold.|
|Jan-07-12|| ||Patriot: <LoveThatJoker> Sure, no problem. First I tried to analyze it a little myself and thought if 1.Ke3, 1...Rxd1 2.Rxd1 Qxd1 3.cxb3 "might" be equal and started looking at 1...Bc5+ and thought 2.Bd4 may be the best reply. Anyway, here's what Houdini said at 20 ply:|
1. (-1.11): 1.Ke3 Rxd1 2.Rxd1 Qxd1 3.cxb3 Bc5+ 4.Kf4 Bd6+ 5.e5 fxe5+ 6.Bxe5 Qc1+ 7.Kf3 Qc6+ 8.Ke3 Qc5+ 9.Ke4 Qc2+ 10.Kd5 Bxe5 11.Qxe5 Qxg2+ 12.Kc4 Qf1+ 13.Kd4 Qf6 14.Qxf6 gxf6 15.Kd5 Kf7 16.b4 h5
2. (-2.68): 1.Qd3 Rxd3+ 2.Kxd3 Qa4 3.cxb3 Qxb3 4.Rhe1 Bc5 5.Rc1 Bf2 6.Rf1 Qb5+ 7.Kc2 Bxh4 8.Rh1 Qe2+ 9.Kb1 Qxe4+ 10.Rc2 b5 11.Rd1 Bg3 12.Rd7 Qe6 13.Rd3 Bd6 14.Ba5 Be5 15.Bc7 Qf5 16.Rd8+ Kf7
So Ke3 is best.
|Jan-07-12|| ||morfishine: Good Work <everybody>! Just love these unbalanced positions. I went with <21...d3> but had left off white's pawn at <h2>. Why do I keep doing that? |
After labeling myself a bonehead, I re-set the position, and trying to be as objective as possible, still select <21...d3> to activate Black's Queen via <c5>, so <LovethatJoker>, I'm with you here.
Really good work today. Very nice job!
|Jan-07-12|| ||LoveThatJoker: <Patriot> Seriously, man. THANK YOU! I'm going to look at this line right now. I can't begin to tell you how much this line troubled me over the past couple of days. And it is good to see that Houdini picked the 4. Bd6+ line. |
<morfishine> Hi morfishine! 21...d3 is a great move. Yes, inferior to 21...Rxh4, but White has to proceed carefully, for if he gets over confident he just might have to lose his R, his Q or just downright get mated somewhere. :) How are things, morfishine?
|Jan-07-12|| ||LoveThatJoker: <Patriot> I played through it. It leads to a won K+P endgame for Black with some wicked cool manoeuvres by the Black Q so as to win a pawn and then trade Q's off. Pure Capablanca! Pure Kramnik!|
Thanks again, man!
|Jan-07-12|| ||bischopper: today Ican not to find a move to win and have 20 minutes on board!!|
|Jan-07-12|| ||profK: Great double rook sac by the first world junior champion !!|
|Jan-07-12|| ||Fervac: <Jimfromprovidence> I think 21... Rxh4 22. Be2 - Re4 is a little pasive. In my humble opinion, I prefer to play 22... d3 is more agressive. I analyzed the position and considered it was better.|
|Jan-07-12|| ||TheBish: E Cobo Arteaga vs Ivkov, 1963|
Black to play (21...?) "Very Difficult", Black is down a rook for two pawns.
It seems pretty obvious that Black's knight belongs on g4, with his queen needing to enter the scene. How best to vacate g4 for the knight?
21...Rxh4!! 22. gxh4
Otherwise Black retains material equality with a strong attack.
22...Ng4+ and now:
A) 23. Kg1
Or 23. Kg2 Bc6+ 24. Kg1 Bxh1 25. Kxh1 Nf2+, winning the queen.
23...d3! 24. Qxd3 Qc5+ 25. Kg2 Bc6+ 26. Kg3 Qe5+ 27. Kh3 Nf2#.
B) 23. Ke1 Qf4 24. Qe2 Bb5! 25. Qg2 Qe3+ and Black will win White's queen after 26. Kd1 Nf2+ or 26. Be2 Qd2+ 27. Kf1 Ne3+.
|Jan-07-12|| ||sevenseaman: It seems some lightweight engine analysis goes for the intuitive <21...Rh4> but when asked to analyze in full depth the computer prefers <21...d3>.|
If that finally be the case then kudos to <LTJ> and my friend <morfishine>. I am requesting <johnlspouge> for a final verdict.
|Jan-08-12|| ||LoveThatJoker: <sevenseaman> Thanks for analyzing the position even further, man.|
I'm glad that your engine, in full depth mode, went for 21...d3.
|Jan-08-12|| ||rilkefan: <<sevenseaman>: It seems some lightweight engine analysis goes for the intuitive <21...Rh4> but when asked to analyze in full depth the computer prefers <21...d3>.>|
As I posted yesterday, stockfish thinks 21...d3 (the move I picked) is only -0.4 at 2 Gnodes.
It tries various things after ...Rxh4, settling on the sad 22.h3 Ng4+ 23.hg at a depth of 27 plies (3.6 Gnodes) with an evaluation of -6.5.
|Jan-08-12|| ||LoveThatJoker: <rilkefan> Whatever the case, give yourself at least 50% for finding a winning continuation with 21...d3. Yes it doesn't crush outright, but it is still a winning play and not a blunder or losing move by any means. |
|Jan-08-12|| ||LoveThatJoker: <rilkefan> Keep in mind that Patriot found 21...d3 to be valued at (-0.90) with Houdini at 23 ply.|
|Jan-08-12|| ||sevenseaman: <rilkefan and LoveThatJoker> I am not an engine analysis man. I have asked <johnlspouge>, the specialist to rule on d3 and Rxh4.|
He has very kindly agreed to run it on Toga 1,2 tonight. We may soon have his opinion. For me that'll be binding.
|Jan-08-12|| ||LoveThatJoker: <sevenseaman> Cool, man.|
I look forward to seeing what <johnlspouge> says on the matter.
Without a doubt though, 21...d3 is definitely a good alternative for Black to the game continuation!
|Jan-08-12|| ||johnlspouge: I ran Toga 1.3.1 on the puzzle position overnight. I give the complete best variations, although human can improve them near the end. In each best variation, this <color> denotes the last move entered by a human; this <color<>> denotes the first move different from the best variation at the previous ply.|
[plies 20 time 8:44:11 eval -5.05 P (for White)] 21…<Rxh4> 22.h3 Ng4+ 23.hxg4 Rxh1 24.Qxd4 Rh2+ 25.Bg2 Qc6 26.Rg1 fxg4 27.Kf1 Qb5+ 28.Kf2 Qf5+ 29.Ke3 Bc6 30.Qxc3 Rxg2 31.Rxg2 Qe4+ 32.Kd2 Qxg2+ 33.Kc1 Qf1+ 34.Kb2 <Qe2<>> 35.Kc1 Kc7 36.Qa5+ Kd6 37.Qc3 Ke7 38.Qd2 Qf1+ 39.Kb2 Qf6+ 40.Qc3
[plies 18 time 5:01:25 eval -1.08 P (for White)] 21…<d3> 22.Bxd3 <f4<>> 23.Be7 Rg8 24.Be2 Qb6+ 25.Ke1 fxg3 26.hxg3 Qe3 27.Qd5 Qxg3+ 28.Kd1 Bc6 29.Qc5 Nd7 30.Qc4 Qe3 31.Ke1 Ne5 32.Qc5 Qxc5 33.Bxc5 Bxh1 34.Bxa7
The evaluation for 21...<Rxh4> was less than -3 P at 13 plies and remained so thereafter. The evaluation for 21...<d3> was 0 from plies 12 to 16, -0.57 at ply 17, and -1.08 P at ply 18.
I then backslid on the variation 21…<d3> 22.Bxd3 <f4<>>:
[plies 17 time 03:53 eval -1.38 P (for White)] 21…d3 22.Bxd3 <f4> 23.Qe2 Qc5+ 24.Kf1 Rg8 25.Be4 Bb5 26. <Bf5+<>> Kb8 27.Bd3 Qd5 28.Kg1 Bc6 29.Be4 Qd4+ 30.Kf1 Bxe4 31.Bf6 f3 32.Bxe5+ Qxe5 33.Qe3 Re8 34.Kf2
My opinion (in case it is of any value) is that 21...d3 is likely just sufficient to win, and that a computer analysis could easily undervalue it. On the other hand, 21...Rxh4 is clearly superior to 21...d3, and even a human could analyze 21...Rxh4 to a clear win without real difficulty.
|Jan-08-12|| ||johnlspouge: Toga now gives
[plies 18 time 22:25 eval -3.02 P (for White)]
21...d3 22.Bxd3 <f4> 23.Qe2 Qc5+ 24.Kf1 Rg8 25.Qg2 Nxd3 26.cxd3 Bc6 27.Qh3+ Kb8 28.g4 Bxh1 29.Bf6 f3 30.d4 Qb6 31.Be5+ Ka8 32.Kg1 Qb2 33.Qf1 Bg2 34.Qc1 Rxg4 35.Qxb2 cxb2 36.Re1
There is no doubt that 21...d3 wins.
|Jan-08-12|| ||Jimfromprovidence: <Fervac> <I think 21... Rxh4 22. Be2 - Re4 is a little pasive. In my humble opinion, I prefer to play 22... d3 is more agressive. I analyzed the position and considered it was better.>|
I don't agree that 22...d3 is better, but it sure looks like it's winning.
It looks like black also has 22..Qc5 first and then 23...d3, so those two moves are probably interchangable.
<Patriot> posted this Houdini analysis of the position.
<1. (-5.92): 3...Rxh4 4.Be2 Re4 5.Rf1 d3 6.cxd3 Ba4 7.dxe4 Qb6+ 8.Kg2 Bxd1 9.Rfxd1 fxe4 10.Rdb1 Qc6 11.Ra2 Nc4 12.Re1 Ne3+ 13.Kh1 c2 14.Raa1 f5 15.Rec1 Qc3 16.Re1 Qa5 17.Kg1 Qc5 18.Rac1 Qxa3 19.h4
It starts on move 3 because I put the position in at move 19 and forwarded to the game position.>
|Jan-08-12|| ||LoveThatJoker: <johnlspouge> Great engine analysis, JLS!|
|Jan-08-12|| ||sevenseaman: <johnlspouge> Thanks for a highly detailed opinion on the <21...d3> line. You have gone to great lengths in providing us an elaborate and thorough analysis.|
It settles all matters between Rxh4 an d3. Its clear both are technically sound. Seeing the incidence of its selection, intuitively Rxh4 appeals more.
One tiny question: You are, as it is, using good depth; does a single computer change its advice from time to time w/o any significant alteration to input parameters?
|Jan-08-12|| ||johnlspouge: < <sevenseaman> wrote: [snip] One tiny question: You are, as it is, using good depth; does a single computer change its advice from time to time w/o any significant alteration to input parameters? >|
Yes. See, e.g., Hiarcs [ http://www.hiarcs.com/pc_uci_option... ] under "Hash". The computer remembers previous calculations to speed up future ones. Most programs have a feature to clear the hash table, for exactly the reason I think you are asking this question :)
|Jan-09-12|| ||rilkefan: <<johnispouge>: There is no doubt that 21...d3 wins.>|
Well, not enough to worry about, but I suspect one can't know the landscape well enough to avoid the pertinent free lunch theorem (sf for example changed its evaluation a good deal this afternoon following up 21...d3 22.Bxd3 <f4> 23.Qe2 Qc5+ 24.Ke1 [which it prefers] going from a depth of 29 to 30 [22 Gnodes]). My sense from your data and my less informed check this afternoon is that ...d3 wins only if one finds a series of box or anyway very difficult moves, hence the slow convergence, and ...Rxh4 is not in that category or the above.
|Jan-09-12|| ||johnlspouge: < <rilkefan> wrote: [snip] My sense from your data and my less informed check this afternoon is that ...d3 wins only if one finds a series of box or anyway very difficult moves, hence the slow convergence, and ...Rxh4 is not in that category or the above. >|
I followed the evaluations for another ply, when it went back down to -1.7 P. I have never seen a computer evaluation oscillate so much at 19 plies. Does 21...d3 win? It appears so. Is 21...Rxh4 better and simpler? Unquestionably so.
|Jan-09-12|| ||rilkefan: <I followed the evaluations for another ply, when it went back down to -1.7 P.>|
This line might serve well for evaluating engines. Though as noted above I want to be careful about distinguishing between some nonexistent absolute algorithmic strength and measurable strength in practical positions.
< Earlier Kibitzing · PAGE 4 OF 4 ·