Can your chess engine see this?

First of all, If your goal is to improve, then always FIRST analyze yourself, and only then use the engine/s to support/refute your analysis. Also, you need to ask the engine the right questions in order for it to be most effective - Suggest specific lines, look at the end of them, look at the positions the engine sees at the end of the lines it is analyzing, check in depth to understand why the engine prefers some line over the others and try to see where the position goes...

Second, both Rybka 2.3.2a and Hiarcs 11see that Bxg2 is great after just a few seconds.

For example, I was wondering why Rybka gave such a high score to Qd7.

You want to know what the reason is?

Other than the cute little threat of ...Qh3 (if Bxf6 STILL Qh3), and white can't bring more defenders to the kingside (Qd7 Ne3 Rxe3!), Rybka sees that Qd7 f3 is suicidal after Bg3, and if white takes on h2, then what does black play? of course... Bxg2 again:) It's essentially the same idea, proposed in two different implementations.

For the same reason, Qc8 is also probably winning... The reason it is more complicated to win with it is a small tactical point: If black the queen leaves the d file, white can play Qd4 immediately, and join the defense. But white is still probably lost for white according to Rybka though (-1.9).

So there are probably few winning lines (Another example - Re2 may be a slower way to win. I would love it if I could not find the tactical shot immediately, and Rybka gives it as better than -1 (it gets closer to -2 as it goes deeper, but you don't need to wait for it to reach that far to form conclusions), and I think it should be winning too).

Modern engines are starting to see these things more and more.  

As to which engines to use... It depends.

Personally, I have no doubt that the latest version of Rybka is the best engine out there. Better positional evaluation, better selection (looking deeper at lines that seem more important), and an excellent search.

I'd recommend it as a first engine.

However, although Rybka is stronger than other engines in most positions (actually in the vast majority of positions), there are many positions in which it is worse than others. 

There are a few middlegame positions where Hiarcs 11 (I suppose 12 too, although it is too new for me to have formed an opinion on it yet) handles and evaluates better than Rybka, for example. However, Hiarcs is much worse than Rybka in overall playing and analysis strength.

On the rating lists, Hiarcs 11 is rated ~150 elo points weaker than Rybka.

If you want a single engine, I'd suggest Rybka. If you can afford 2, I'd suggest Rybka and Hiarcs. If you want to know which engines are better at certain kinds of endgames..... That's a whole other topic, and I don't think you'd want to use ~7 different engines:)

Naum 3, Zappa and Shredder 11 are also excellent commercial engines. 

Zappa is only very good if you have many processors. at least a quad core. It scales VERY well.  Otherwise it is just worse than other engines. 

I wouldn't look at other commercial engines of those that are currently available Junior and Fritz, while very good, are not as good as the ones mentioned above.

This may change with future versions. But, The best, by far, is the latest version of Rybka 2.3.2a. Hiarcs 11/12 simply complements it on a few positions where it does not do too well and it does, so Rybka would be my recommendation

Of the freeware engines, Toga is spectacular. and in pure strength, it is about equivalent to Hiarcs 11. Those things change with versions... But it is free, so you can easily download and test it.

 Thanks a lot. I'll think I'll go with Toga. And possobily maybe GambitFruit too.

>  Thanks a lot. I'll think I'll go with Toga. And possobily maybe GambitFruit too.

It's an excellent engine. 

Just bear in mind it is still significantly weaker than Rybka 

 Ok. But it still is significantely stronger than me and most importantly I'm a cheapster.

Significantly stronger than you shouldn't be the criteria.

Even Crafty is significantly stronger than you and me both, but it's really junk in so many positions. As an analysis tool that I expect to help me find flaws in my moves/plans and find new ideas, it is extremely bad. 

Toga is very good, but it doesn't analyze complex most positions as well as Rybka. I think it's worth the price. 

If you must go freeware though, Toga is an excellent choice indeed 

Your system also makes a difference. Erik's running Toga on two quad cores with 16 GB of RAM. Although Rybka on a single core is supposedly stronger than even that. ;)

 Good point but like I said I'm a cheapster. And I think that for the moment Toga will be enough. I've done some reading and looks like it enough for me. I also got GambitFruit to cross check things but I suppose even this will be more than I can handle.

 Yes I guess that my not not so top of the market laptop will be slow and might miss something. But I'm planning on complementing things. Me, the engine and a few books. But I don't want to lose even more time to chess than I do now it's just a way to check things I might have missed.

 That's really overestimated. even if you had a quad core, you won't be that far away from the single core version. It's more reliable, but it's damn strong anyway. (Rybka's scaling has also never been worked on seriously before, as the author (Vasik Rajlich) confirmed before, this is one of the issues to improve in future versions. The result is that for now, it doesn't scale as well as it can).

You still get something which will point out most of your problems. 

Can it just be that Qd7 is better?  With Bxg2 directly, Black arrives at Qxg5 without check allowing movements like Ne3 (preventing the black's knight to go directly to G4 and aiming to g2) followed by Qc4+ with tempo to reposition the Queen to aid the poor lonely King.  And if black takes the knight with the rook at e3, white is running low on material so now the threat is to capture the rook with the pawn and then simply trade the queen for the knight!!  White would have a rook (both!) and a knight for the queen and.  Black still stands better I think, but I also think that playing Qd7 instead of the immediate Bx2 could be stronger.

Sorry if what I say it's a nonesense, I'm using the engine MyBrain 1.0 and it's rated below 1.400...

grolich>  That's really overestimated. even if you had a quad core, you won't be that far away from the single core version.

A quad-core 64-bit system loaded with memory is often 100-200 ELO stronger than a basic system... so it's also a factor that also shouldn't be underestimated (even if we don't all run out and buy quad-core systems). But both Toga or Rybka are fine engines for post-game analysis. I think the original poster is making a fine choice.

Crafty 22.0 found the following winning combination:

  16    05:12     425.102.221    1.362.507    -3.88    Bxh2+ Kh1 Qd7 Kxh2 Bxg2 Kxg2 Qg4+ Kh2 Qh5+ Kg2 Qxg5+ Kh1 Ng4 Nbd2 Qh4+ Kg2 Qh2+

 Now, if I understand, you're sayingthe 2nd bish sac is on move 2...  If I force that variation, Crafty sees the following win:

  16    05:40     487.107.994    1.428.469    -3.90    Kxg2 Qd5+ Kxh2 Qxg5 Ne3 Rxe3 Qc4+ Kh8 fxe3 Ng4+ Qxg4 Qxg4 Nc3 Qh4+ Kg2 Qg5+ Kh2 Qxe3 Rae1 Qd2+ Re2 Qg5

 

I’m absolutely impressed beyond belief. Congratulations to all of you. Sounds like you are all the future makers of the first HAL.

Deep Junior 10.1 also saw both sacs.

1 ... Bxh2 2. Kh1 followed by:

     2. ... Bxg2+ 3. Kxg2 Qd5+ 3. Kxh2 Qxg5

next in line was:

     2. ... Qd7 3.Kxh2

 Toga is a great choice, He should just be aware that the difference between Toga and Rybka on the same system is about the same as the difference you mentioned from the 64 bit quad core system

These tournaments have been run for countless times.

Actually people are running series of hundreds of games of different engines all the time. Some rating lists are maintained on the net

(Just keep in mind that ratings are only really comparable within the same system, not between systems, so when you see one engine with a rating of 3000, and another of 2800, the difference is the same as 200 point difference in the FIDE system, but it does not say if the 3000 rated player would be rated 3000, 2990, 2900 or 3100, were it playing in FIDE tournaments).

Hundreds of games is important... Because a single tournament says nothing really. An ELO difference of even 200 over ALL other engines should give a 75% expected wins over each of the others. But if you're a single engine among ~10, and there are only ~9-10 rounds, the odds are actually not that good at all.

Make a few hundred (or a few thousand) games, however, and you have a clearer picture.

 Here's the complete rating list of one of the most well known lists (with ALL versions tested for each engine, with every number of CPU it was tested with. Time control 40 moves in 40 minutes):

http://www.computerchess.org.uk/ccrl/4040/rating_list_all.html

Rybka on a single cpu is stronger than almost anything else, except Naum and Zappa on 4 cpu machines. Rybka 32 bit, which is what most simple home users will run, on a single cpu, is stronger than almost all other engines on 4cpu (other than naum3 (on 2 cpu) and zappa (on 4 cpu).

So pay attention to the 32,64 bit issue (Even if modern processors are 64bit, the operating systems most of us have (all of us, unless we specifically requested) are 32 bit only), as well as the number of CPUs.

On that list, Deep Shredder 11 64bit 4cpu is rated 3022, while it is about the strength of Rybka 32bit on 1cpu(3027). So don't confuse them as being equal.

(Rybka 64bit on 4cpus is 3125)

Shredder 11 on 1 cpu is 2938. They don't write if it's 32 or 64 though, which is a shame, because there's a big difference.

Rybka is the best in pure playing power, and it's superior to other engines in positional play in most cases. It's also very powerful tactically in the vast majority of positoins, but weaker tactically in others.

Ratings say nothing of style though. Which is an entirely different topic. Modern engines have improved evaluations that can make even position sacrifices from time to time (Rybka is very good at that. Better than other engines, but other modern engines do this as well sometimes).

Shredder is better in some endgames than Rybka, although in othres, Rybka is better..... Hiarcs has a more attacking style, which may hurt his objectivity at times, but it means that it finds some breakthroughs in attacking positions faster than Rybka at times.... Still, Rybka can refute or defend those attacks better than other engines most of the time.

So, I'd say, first choice, Rybka, second choice, Hiarcs, third choice, Shredder (For me). (actually Naum 3 is surprisingly strong. I may change my preferences soon for second place).

That's only my opinion though.

Oh almost forgot, you mentioned Zappa...

Zappa isn't all that impressive when rum on 32 bit 1 cpu.

It's just another very strong engine. I see nothing spectacular either in style or in strength to convince me to use it.

Zappa was made to be an engine with excellent scaling to multi cpu environments, so it gains more profit than most engines (maybe all...) from extra cpus.

On the Rybka forum, Rybka's programmer (Vasik Rajlich) said that nothing really prevents Rybka from scaling the same as Zappa, if time and effort were put into this issue, but that right now, there are things that are more important for Rybka's playing strength.

 Some people had already posted the winning lines for black if black sacrifices the two bishops immediately but if you don't see it still we can play an unrated game were you can use an engine to choose your moves and  you'll see that I'll win  the same. In the game I missed Qxg5 exactly becaus it doesn't gives check. Bt it threatens a lot of things and black can  even sacrifice the exchange afterwards and still be  in a winning position.