Rebel TNG (version 1.0) approximately 2950....

[Eelco de Groot]

Hi Kaj,

It is just up to Ed I think, whatever he would find interesting. But it just seemed a "tad sad" that he had done most of this hard work just for an April Fool's joke. And that he thought of using Stockfish this way but not of Rebel itself. The main thing is just that, when your program is using just one processor even of a machine that has one processor with hyperthreading possibility, all that extra power is lying around free, it only costs some extra current or battery capacity to use it, even mobile phones from the Aldi today they have hyperthreading and/or four cores in the more expensive segment. The programmer of Shredder had this idea that he called 'Drei Hirn', where three top programs would look at the same position and then an arbiter program would look at the scores of all three and decide which of the different moves had the best chances and play that move, even if only one program had found it. But a drawback was always that the programs were running on one processor systems in those days and all got just one third of the available time. If you have idle processors lying around you don't have to do that. That would be a Drei Hirn or Zwei Hirn system I suppose. But Ed's april fool's joke I think already went further, it used the output of Stockfish while both programs were searching, and the best move of Stockfish, information which is just standard UCI output, then got a bonus inside Rebel's search. At least that is how I understand the idea Ed had. But supposedly Stockfish is running not 100% of the time and it still works, so why not try the same with another program? It just seemed an interesting experiment.

Regards, Eelco

[Kaj Soderberg]

Ahh...the Triple Brain kinda stuff. I had forgotten all about that. Have not touched it since Shredder 5 or so. Great idea. Would be very interesting to see what Ed can make of the concept with contemporary computing.

Cheers,
Kaj

[Henk]

Ahh...the Triple Brain kinda stuff. I had forgotten all about that. Have not touched it since Shredder 5 or so. Great idea. Would be very interesting to see what Ed can make of the concept with contemporary computing.

Cheers,
Kaj

Won't work if the output of the brains are too much correlated. I used the same idea when predicting stock quotes. Each predictor had a score of say 55 - 62%. So I thought creating many of these predictors and taking the majority as an estimate would work. But of course it did not work.

[Ponti]

Oh, I see.
I did not read the entire thread.

As a matter of fact, I'm a loyal old Rebel's customer, and I *really* did not like the joke.

I spent my time downloading it, playing against it, and even trying to configure it with Aquarium, which I could not accomplish.

Stephan already did that with Shredder ("triple brain"), but it was a serious try.

Maybe it's fun for programmers, but it is sad news for users.

Sorry for my bad mood.

Maybe I'm going to spend my time again, this time erasing all Rebel versions of my hard disk.

[Eelco de Groot]

I'm sorry you feel bad Aloisio. Please don't erase the Rebels! It's not their fault, it's their programmers fault Maybe Ed will make a real Rebel TNG one day.

Best, Eelco

[Ponti]

It is really unbelievable !

At least it could be explained in Rebel's page that stockfish is used as second engine.

We, brazilians, don't like april fool's jokes... it is only for children.

I take computer chess seriously, because I use engines to help analysing my games.

Although I really like Rebel, this episode is really disappointing.

The main idea seems interesting, though, but now that +250 means nothing.

[Adam Hair]

It is really unbelievable !

At least it could be explained in Rebel's page that stockfish is used as second engine.

We, brazilians, don't like april fool's jokes... it is only for children.

I take computer chess seriously, because I use engines to help analysing my games.

Although I really like Rebel, this episode is really disappointing.

The main idea seems interesting, though, but now that +250 means nothing.

So, o dia das mentiras is only celebrated by children?

[Rebel]

Hi Eelco,

But Ed's april fool's joke I think already went further, it used the output of Stockfish while both programs were searching, and the best move of Stockfish, information which is just standard UCI output, then got a bonus inside Rebel's search. At least that is how I understand the idea Ed had.

It's even more simpler, it's just a positional bonus applied to the root move only.

But supposedly Stockfish is running not 100% of the time and it still works, so why not try the same with another program? It just seemed an interesting experiment.

Exactly my sentiment.

When I got the idea (in January) I did not know what to expect and before putting a lot of energy in it I first contacted Marco for his approval and he was OK with the intent. So I started and a first try (a bonus of 0.25) at 40/60 gave about 120 elo, increasing the bonus to 0.50 gave approx. 20 elo more. While 140 elo is a tremendous elo gain (for just 2 hours work) it simply is not good enough for an April-1 joke. Marco said, you need at least 200 elo and I agreed with him.

And so I nearly gave up till I realized the algorithm was sensitive to scaling but I did not know to what extend. And so I started the painful process of testing on even longer time control, boy I have learned hate that. But the pay-off was amazing, could not believe my eyes. The 250-300 elo is very real. While the orginal PD version loses against SF 1.8 with 98-99% (due to outsearching mainly) it now got 44%.

The outsearching is still present of course, Rebel TNG won't be alarmed if SF evaluates Rebel's move as a losing move and in reverse, Rebel TNG won't play a deep material sacrifice found by SF. Considering all of this it's amazing what a simple positinal push of 0.50 can do. I have a parameter that can increase the bonus to a maximum of 2.00 but never tried it because then it's like dictating Rebel what to play.

[Rebel]

We, brazilians, don't like april fool's jokes... it is only for children.

Humor indeed is culture related. If you consider it as bad humor, it still remains what it is, humor.

[Ponti]

I think most of adults here (Brazil) don't care about fool's day, but children do.

Of course there are adults that celebrate fool's day...

Anyway, I was a victim of a technical joke... congratulations, Ed !