Deep Junior 12 UCI

[Uly]

For analysis Critter is the best in the opening, period, it can teach a few things in that stage of the game to top engines like Rybka or Stockfish. I really doubt the crazy moves of Junior will help humans get a better understanding of the game, if they can't understand Junior's moves themselves.

And while I agree elo isn't everything (and that's my main disliking of Stockfish that makes it very difficult to analyze with it, by design) having high elo doesn't automatically mean that you have a dull style of play, that was back in the days of Rybka 2.3.2a, nowadays engines like Critter, Stockfish or Rybka 3 Dynamic have a really high elo up there without any sacrifice of style, and going to the limits of Junior if it sacrifices a pawn for the sake of it isn't really necessary.

A position is either drawn, won or lost with perfect play, but that's not achievable anyway, so who needs an engine that says "see, you can sacrifice your Bishop for two pawns on here!" if other engines will suggest moves that are easier to play for the user, and the user can explore such sacrifices by himself?

What is the use of engines like Junior? Producing great looking games is unsatisfactory if it's just about showcasing how poorly the losing side defended.

[tomgdrums]

For analysis Critter is the best in the opening, period, it can teach a few things in that stage of the game to top engines like Rybka or Stockfish. I really doubt the crazy moves of Junior will help humans get a better understanding of the game, if they can't understand Junior's moves themselves.

And while I agree elo isn't everything (and that's my main disliking of Stockfish that makes it very difficult to analyze with it, by design) having high elo doesn't automatically mean that you have a dull style of play, that was back in the days of Rybka 2.3.2a, nowadays engines like Critter, Stockfish or Rybka 3 Dynamic have a really high elo up there without any sacrifice of style, and going to the limits of Junior if it sacrifices a pawn for the sake of it isn't really necessary.

A position is either drawn, won or lost with perfect play, but that's not achievable anyway, so who needs an engine that says "see, you can sacrifice your Bishop for two pawns on here!" if other engines will suggest moves that are easier to play for the user, and the user can explore such sacrifices by himself?

What is the use of engines like Junior? Producing great looking games is unsatisfactory if it's just about showcasing how poorly the losing side defended.

Even though I did not attack Critter (at all!) somehow you turned this into a "this engine" vs. "that engine" thing.

And I DO think Junior can be very helpful for one's play OTB (or in real time on the internet). Real time play is not like correspondence play. No one can play like an engine in real time. Time pressure, competitive pressure, blind spots all play a part. So speculation DOES work many times. That is the beauty of chess it has many different ways it can go. And Junior (along with other engines) has helped me get better at OTB play. It has helped me look at different moves and different ideas. Playing against Junior and using it (in conjunction with other engines) in post mortem analysis of my OTB games has been very helpful.

And I am sure Critter is a great engine as well. Why did you feel the need to get so excited about it??

[Uly]

This part:

Junior actually serves MY purposes better than Critter

Have you actually used Critter in your games to see if Critter isn't speculative enough for your purposes? Someone that doesn't have Junior and reads your posts may get the idea that Junior is the best engine for speculative play, but I claim that's not measurable, show me a move that Junior suggested on some position that is aggressive and puts pressure on the opponent because he has to play accurately, and I'll show you a different move that does the same thing from that postion, that I could find without using Junior.

Why did you feel the need to get so excited about it??

I get just as "excited" about all the themes that interest me, Jeremy posted that he got the impression that Junior would provide "human-like analysis" and I'm sure Junior doesn't. Junior just provides analysis that is different from other engines, but still analysis that no human would give, for the sake of shocking the opponent, but that may not even be the best way to shock (having experience with it, sometimes delaying the shock is better, but that requires human judgment), in the end the style of a player may be alienated without a good reason.

Thinker is another engine that provides such shock factor moves, maybe it would be a better comparison, it's free and 70 elo better than Junior (yes, there's more in life than elo, but why not playing OTB a move that shocks the opponent AND has a refutation harder to find? engines with better elo provide better plans).

[tomgdrums]

This part:

Junior actually serves MY purposes better than Critter

Have you actually used Critter in your games to see if Critter isn't speculative enough for your purposes? Someone that doesn't have Junior and reads your posts may get the idea that Junior is the best engine for speculative play, but I claim that's not measurable, show me a move that Junior suggested on some position that is aggressive and puts pressure on the opponent because he has to play accurately, and I'll show you a different move that does the same thing from that postion, that I could find without using Junior.

Why did you feel the need to get so excited about it??

I get just as "excited" about all the themes that interest me, Jeremy posted that he got the impression that Junior would provide "human-like analysis" and I'm sure Junior doesn't. Junior just provides analysis that is different from other engines, but still analysis that no human would give, for the sake of shocking the opponent, but that may not even be the best way to shock (having experience with it, sometimes delaying the shock is better, but that requires human judgment), in the end the style of a player may be alienated without a good reason.

Thinker is another engine that provides such shock factor moves, maybe it would be a better comparison, it's free and 70 elo better than Junior (yes, there's more in life than elo, but why not playing OTB a move that shocks the opponent AND has a refutation harder to find? engines with better elo provide better plans).

Well unfortunately I couldn't use Critter because both times I tried to download it I got a virus...and I got a PM on the Rybka forum that I wasn't the only person that happened to. So no I haven't used Critter. But Junior does have limit strength (which I like) and a good opening book that comes with it. And I like it. AND in my previous post I even stated that I was sure Critter was a great engine! I never said Junior was better or Critter worse, I just said that Junior serves me better for what I need. Most of the free engines don't have enough of the features that I want in an engine. That is just the way it is at this time.

And I don't know why you dislike Hiarcs (our opening book argument) and Junior so much?!?

But I happen to like both Hiarcs and Junior AND Shredder...A LOT. They are fun to play against and give great and different analysis from each other and all come with great opening books.

And I have nothing against Critter. I wanted to use it and tried twice.

[Uly]

I just said that Junior serves me better

You can't say that if you've never used Critter (you'd say something like "Junior serves me well, and I've never used Critter", actually, you shouldn't mention Critter at all).

I never said Junior was better or Critter worse

Yes, you did, I repeat:

Junior actually serves MY purposes better than Critter

And I don't know why you dislike Hiarcs (our opening book argument) and Junior so much?!?

Who said I dislike Hiarcs? Hiarcs Paderborn 2007 is one of my favorite engines and for the same reason I love Critter! Original good moves, great playing style.

I also have nothing against Junior, I just have to jump in and defend Critter when you keep repeating that Junior is better than Critter even when you've never used Critter (and apparently because Junior doesn't come with a virus? I've downloaded all releases of Critter and all of them have been clean).

In conclusion, I've used both and Critter comes with more original good moves than Junior, love Critter's style, and speculative play can't be measured.

[Jeremy Bernstein]

Let's turn this around. Which engine with reasonable strength provides the most human-like analysis? I tend to use HIARCS for analysis of my games (although at the level I play, and the time give for game analysis (~15 minutes for a game after I've marked it up myself), the difference between what HIARCS says and what Rybka says is negligible). What are other good options?

Jeremy

[kingliveson]

Let's turn this around. Which engine with reasonable strength provides the most human-like analysis?
Jeremy

You hear this all the time (human like), but I have never been able to figure it out...I dont even know what it means. More blunders?

[Jeremy Bernstein]

Let's turn this around. Which engine with reasonable strength provides the most human-like analysis?
Jeremy

You hear this all the time (human like), but I have never been able to figure it out...I dont even know what it means. More blunders?

I guess I'm imagining some Turing Test for chess computers. When I play engine X, can I be convinced that I am playing a human? Computer move choice is always logical, but not, for lack of a better word, idiomatic for human play. This could have plenty of reasons, but the fact that even the best human players have blind spots, rule out certain lines without a second thought, which a computer might successfully investigate, makes a difference in game play. Sure, the computer move might be "better", but it's not a move that would pass this "Turing Test". I personally don't care if my engine plays well against other engines.

[tomgdrums]

I just said that Junior serves me better

You can't say that if you've never used Critter (you'd say something like "Junior serves me well, and I've never used Critter", actually, you shouldn't mention Critter at all).

I never said Junior was better or Critter worse

Yes, you did, I repeat:

Junior actually serves MY purposes better than Critter

And I don't know why you dislike Hiarcs (our opening book argument) and Junior so much?!?

Who said I dislike Hiarcs? Hiarcs Paderborn 2007 is one of my favorite engines and for the same reason I love Critter! Original good moves, great playing style.

I also have nothing against Junior, I just have to jump in and defend Critter when you keep repeating that Junior is better than Critter even when you've never used Critter (and apparently because Junior doesn't come with a virus? I've downloaded all releases of Critter and all of them have been clean).

In conclusion, I've used both and Critter comes with more original good moves than Junior, love Critter's style, and speculative play can't be measured.

You really have to LOOK and READ what I said. I said Junior serves ME better. That is different than categorically saying Junior is better than Critter. Much different. And I also stated that Junior has limit strength and a good internal and ctg book which I can use. So yes even IF I hadn't contracted viruses when downloading Critter...Junior would still serve me better because of everything I just noted. And to repeat: Saying it serves me better is not the same as saying it is the better engine. Better is subjective and objective, depending on what the user wants and needs.

People in the computer chess world really need to get a grip on the fact that people use different engines for different reasons and purposes and end results.

In fact other than glass and pro-deo most of the commercial engines serve me better right now because of features that I like to have in engines. Those are my purposes and reasons. Other people like to do other things with engines. So other engines will serve THEIR purposes better. Why is that not ok with you?

And how on earth do you see it as an attack on Critter? I never criticized Critter's play, strength or move choice. In fact, in my initial post, I didn't even bring up the two viruses I caught from downloading two different versions of Critter. And YES that does mean something to me!! Sorry but it hasn't happened with any other program..free OR commercial...that I have ever downloaded.

[tomgdrums]

Let's turn this around. Which engine with reasonable strength provides the most human-like analysis?
Jeremy

You hear this all the time (human like), but I have never been able to figure it out...I dont even know what it means. More blunders?

I am not totally sure either but I do think that certain engines will give clearer or more easier to follow variations for humans to consider and learn from in OTB or real time situations.

I have noticed that Hiarcs and Junior (and Shredder sometimes) give variations that I understand and are still winning. Sometimes an engine at the strength of the super top engines will give variations that are just slightly better but one needs to understand 10 plies forward to get why it is that extra centipawn of a better move.

NOW that extra centipawn is CRUCIAL in engine vs. engine and/or corr. play where engines are used. Absolutely crucial.

I do not think that extra centipawn is as crucial when trying to use engine analysis to understand a post mortem of an OTB or real time game between humans. Strategy does play a factor.

For instance I have been going through the tests in Silman's "The Amateur Mind" and on one test I got the right move and plan according to Silman (yay me!) and then just to see I checked it with an engine. Hiarcs liked Silman's choice and then Rybka showed a different move immediately. When I showed Rybka's choice to Hiarcs it then chose Rybka's move BUT the difference in evaluation was minuscule for both engines. So Hiarcs at first chose the more obvious, and strategically sound and winning move that the human master chose (Silman)

Rybka instantly went for a slightly better move. I can't remember the difference in evaluation but it was small. So both engines chose winning moves and Hiarcs' choice was clearer and based on some "human" strategic thinking.

That is why Rybka will win against Hiarcs in most engine vs. engine matches...but I am not always going to get Rybka's analysis as to why something is better especially in murky situations. Hiarcs and Junior like to pick more obvious and still "correct" moves that help me to learn for OTB.

I am not saying that I NEVER learn from what Rybka gives for evaluations! But quite often Hiarcs gives the strategically sound and clearer version so that my human brain can see why it is winning.

That is what I think of when I think of "human-like" for an engine. And of course how they implement the limit strength fuction as well!

I also firmly believe that user's purposes have A LOT to do with what engines best serve his or her needs. That is something that is often overlooked in the computer chess world.