Robbobase origin

[benstoker]

Don't see anywhere a discussion about the origin of the robbobase code. Is it presumably "original"?

[BB+]

I have not seen anyone claim that it is not original to "Roberto Pescatore" (maybe "plus Decembrists"), except what I mentioned here regarding the computation of a suffix tree in the compression phase.

[benstoker]

I have not seen anyone claim that it is not original to "Roberto Pescatore" (maybe "plus Decembrists"), except what I mentioned here regarding the computation of a suffix tree in the compression phase.

Where you rate creation of the robbobase code on the "difficulty" scale? Hats off to the robbobase author(s)? Or, rather, nothing to sneeze at?

[ernest]

I have not seen ....

About a year ago, I read somewhere in a Forum that they traced some origin with somebody real (not any "Roberto Pescatore"), doing a research project.

[BB+]

About a year ago, I read somewhere in a Forum that they traced some origin with somebody real (not any "Roberto Pescatore"), doing a research project.

My only recollection of something like that was that Conkie(?) found an Italian guy who had been doing bitbases on/off for a number of years. I don't think there was anything definite. Maybe asking the EGTB forum would be the way to go.

Where you rate creation of the robbobase code on the "difficulty" scale? Hats off to the robbobase author(s)? Or, rather, nothing to sneeze at?

Hats off, definitely (even if it started out as someone else's research project, as conjectured above). The building code is much faster than Nalimov (especially after they added SMP, and the Larsson/Sadakane code I mentioned above), and they use less disk space in the "TotalBases" than any other such format (450GB vs 1.2TB for "standard" 6s). Even "minor" points like saving "win, draw, distance-to-loss" instead of "loss, draw, distance-to-win" (as FEG does, I think) seem to be done the "correct" way to my mind (well, after seeing their choice, now I understand why this is likely the superior way). The indexing is superior to Nalimov, but that doesn't say much. They have 5+1 done, and also a bitbase option, recently adding "dynamic loading" to the latter (previously, before they did 6s, you had to have these bitbases in memory it seems). The "Blocked" bases are another plus, though I guess they could go further and add "opposing pawns on same file" more generally.

Some negatives are: they use DTC rather than DTM, there is little documentation about how to "plug" it into an engine (though their ComStock code helps some with that), it's (yet another) format with many GBs to download, and building takes a large amount of memory (32GB, maybe you can do it in 16GB). I'm also not sure their bitbase compression is all that great, but maybe there is a space/time trade-off. There's also no "online" way to access the TotalBase info, though I guess this would not be too difficult (they have some Java GUI for this already).

[ernest]

Don't see anywhere a discussion about the origin of the robbobase code. Is it presumably "original"?

I think BB+ gives a good advice: "Maybe asking the EGTB forum would be the way to go."

http://kirill-kryukov.com/chess/discuss ... um.php?f=6

[kingliveson]

About a year ago, I read somewhere in a Forum that they traced some origin with somebody real (not any "Roberto Pescatore"), doing a research project.

My only recollection of something like that was that Conkie(?) found an Italian guy who had been doing bitbases on/off for a number of years. I don't think there was anything definite. Maybe asking the EGTB forum would be the way to go.
[...]

Given Conkie's history of investigation success, it will be prudent to take whatever he says with a grain of salt. The thread (or topic) was created by Michael D. on talkchess EO hidden (requires log-in) sub-forum. I recall downloading some source and could not verify any of his wild claims.

[ernest]

...was created by Michael D. on talkchess EO hidden (requires log-in) sub-forum.

A search in talkchess (nothing hidden...) with Michael Diosi and robbobase leads to
http://www.talkchess.com/forum/viewtopi ... 847#333847
where Michael gives this link
http://users.libero.it/rigel_g/#Inglese

...probably nothing new?

[kingliveson]

...was created by Michael D. on talkchess EO hidden (requires log-in) sub-forum.

You mean Michael Diosi?
What is talkchess EO hidden subforum?

Yes, Engine Origins to discuss clones. Hidden forum is not visible to the public.

[ernest]

Yes, Engine Origins to discuss clones. Hidden forum is not visible to the public.

In the mean time, I had edited my post!...

Note that Engine Origins is not (to my knowlege...) a hidden subforum.
"The Edge" in Rybka Forum is (to people with less than 500 posts).