rybka randomizer versus gm joe ghallager analyses

[tano-urayoan]

c3 is not a weak move just doesnt bring any advantage to white as opposed to dxc5 or the wild c4 as Mischa suggested. The question will be how good is an engine to help find an opening suggestion in a very early phase of the game. Using Joseph method with the Rybka tool didnt brought anything astonishing to this position. Maybe is engine another engine will have suggested c4 or dxe5 as someone here have said: is better to have multiple opinions that just following the inconditional leader.

[ozziejoe]

Yes, it would be fascinating to see what other engines come up with (particularly engines that are more aggressive and/or contrast with rybkas style)

Maybe the future gui for rybka will allow you to randomize different engines (I hope)

best
J

[Marek Soszynski]

Joseph,

The Randomizer is relatively new technology for all of us. Isn't the popularity of 4.Nf3 merely a function of how the Randomizer works at 7 plies over "about" 1,200 games? Wouldn't 12,000 games give a different popularity of moves (with different success statistics)?

My hope is that computers can help with the analysis, even early in the opening.

My problem is that i don't trust theory, even early on.

Then why not run the Randomizer on the start position itself, before any move has been played? Surely someone has already done this...

[Ovyron]

Then why not run the Randomizer on the start position itself, before any move has been played? Surely someone has already done this...

I did:

Preliminary conclusions: Nf3 was the best move from the starting position, but then, it was very easy to refute it by throwing theory at it, and lowering its score. And then d4 was the best move, but after doing more deep randomizers on moves that are theoretically good against it, it proved to be not as good, and f4 (!?) took the lead...

At some point, it was revealed that black had a better score than white from the starting position!

Final conclusions: it was a never-ending story that proved that 500 years of theory were way better than the randomizer. Also, the randomizer scored really bad with e4, so, by blindly looking at it, e4 would have been ignored completely.

[Marek Soszynski]

Final conclusions: it was a never-ending story that proved that 500 years of theory were way better than the randomizer.

Thanks Ulysses! You have confirmed my suspicions. The Randomizer may yet prove its worth in middlegame positions, but we shall see.

[Mark]

Then why not run the Randomizer on the start position itself, before any move has been played? Surely someone has already done this...

I did:

Preliminary conclusions: Nf3 was the best move from the starting position, but then, it was very easy to refute it by throwing theory at it, and lowering its score. And then d4 was the best move, but after doing more deep randomizers on moves that are theoretically good against it, it proved to be not as good, and f4 (!?) took the lead...

At some point, it was revealed that black had a better score than white from the starting position!

Final conclusions: it was a never-ending story that proved that 500 years of theory were way better than the randomizer. Also, the randomizer scored really bad with e4, so, by blindly looking at it, e4 would have been ignored completely.

Would it speed things up if you had black playing with a strong book, or are you already doing that? Also, I was thinking that it's too bad that the randomizer has to play out the entire game. It would be good if it could adjudicate the game based on the eval after say 20 ply or so. You could get many more games that way.

Regards,

Mark

[Ovyron]

Would it speed things up if you had black playing with a strong book, or are you already doing that?

A strong book would probably destroy the randomizer... So, both sides need a book, and then your results depend on the positions coming from the book, but then, starting from these positions instead would be better than beginning from the starting position.

It would be good if it could adjudicate the game based on the eval after say 20 ply or so.

I think that you can tell the GUI to do this? (Adjudicate the game after a certain score threshold.)