ARCHIVED POSTS
< Earlier Kibitzing · PAGE 826 OF 1118 ·
Later Kibitzing> |
| Jun-02-15 | | zanzibar: OK, for <Bartonicek, Zdenek> I think <CG> may have picked the wrong guy. <
317276 Bartonicek, Zdenek (1953) CZE 2072 / 0 304603 Bartonicek, Zdenek (1978) CZE 2042 / 0 >
Right now, @p70108 -> @f304603 (the younger player). The <CG> player has only 1 game/tournament <2001 Schwabisch Gmund op>. The younger player has no FIDE tournament history of this, but the other player does: http://ratings.fide.com/individual_... Lists
< 14.Staufer-Buer Schwarbisch-Gmuend GER 2002-01-02> That's a hit, even if it's the wrong year given the binary choice between the two players. The lesson is to use FIDE tournament history when possible. |
|
| Jun-02-15 | | zanzibar: Next, <Beckmann, Klaus> @p73955 Clearly trouble, with such a common name. But two/four matches don't have ratings. Our player is fairly active, so it really comes down to these two: <
4604334 Beckmann, Klaus (1968) GER 2303 / 0 4642570 Beckmann, Klaus (1959) GER 2154 / 0
>
Again, care must be exercised. But, the younger player has 150-rating point advantage. That big a differential is helpful. Looking at <CG>'s player shows a big spread in years - so <CG> could already be mixing the player's together. But some of the tournaments are recent, and if we clearly identify which of the two - maybe we should split the games. OK, time to look at FIDE - and I'm surprised not to see the EuroCup tournament for either player. Too much work... kick this one over too. |
|
| Jun-02-15 | | zanzibar: <Benes, Petr> @p63965 Just a quick look at <CG> shows a problem with the first game, from 1968. Even the oldest would only have been 8 years old. Possible? Yes, but no way!
Must be a different <P Benes>, or maybe just a <Benes>. Next, a game from 1979. This well could be the older player, almost certainly not the younger. Just trying to match up the FIDE id's allows insight into the quality of the pre-existing games. OK, we could go on, but that's the rough idea.
I suppose we could attempt some sort of tirage in player identification quality similar to the idea of doing three-levels of tournaments too. |
|
| Jun-02-15 | | zanzibar: OK, one last thought...
I like the biography for Alexander Bochkarev <Two players: The first one is from Ukraine, born on the 9th of April, 1970. The second one is from Russia, born on the 2nd of May, 1975. He is a FIDE master.> It lays it out clearly. Maybe adding two links would be helpful, but we're big boys and girls (and others?), so we can go from there without too much trouble. |
|
| Jun-02-15 | | zanzibar: <ust trying to match up the FIDE id's allows insight into the quality of the pre-existing games.> Should read <quality of the pre-existing game-player match> |
|
Jun-02-15
 | | chessgames.com: zanz, it's easy to say "let the computer make the changes where it is sure it is right, and when it's not sure, kick it over to the Bistro." But how sure are we that the computer is right, when it thinks it's sure? I fear computer mangling of data more than human error. As the old saying goes: to err is human, to really mess things up require a computer. The software I have now does some clever things, but there are some nuances it remains oblivious to, like geography. (That's a big one!) So its guesses are often right but it can be wrong in situations where any human would instantly realize the mistake. Anyhow I think we imagine mostly the same thing. Not a fully automatic system but an interactive system where an admin can give a thumbs-up to suggestions of the computer, after scanning the pertinent facts. If you ever believe you have some magic python routine that can produce output like "player #____ is fide #_____" I'd be happy to analyze its output. |
|
Jun-02-15
| | chessgames.com: On a complete different and possibly more fun subject, we're finally rewriting the horribly buggy and inefficient "Similar Game" feature. Here's a sneak peek: http://www.chessgames.com/perl/ches... (Just change the GID in the address to look at other games). The problem is, this version isn't necessarily more efficient. In fact it can be quite CPU intensive if it has to make graphics on the fly. (There is a little legend at the bottom that shows how many milliseconds it takes to produce the page. I've seen it get as high as 250 which is very bad indeed.) So this might end up being a premium-only feature just due to resource concerns. It mostly acts as an extension to the Opening Explorer so I feel somewhat justified in making it work that way. I suspect there are bugs in this new version (I'm still looking to test it on a game with a novelty on moves 20 through 21, if that's even possible.) In any case I the old version really has to go, it was based on a method of indexing openings that the Opening Explorer made obsolete almost a decade ago. Time for an update. |
|
| Jun-03-15 | | zanzibar: <chessgames> I seem to remember some of that stuff involving squares and pieces moving on a board.... the Spanish torture is nothing compared to Database torture. Anyways, looks like a nifty visual variant of a tree analysis. |
|
Jun-03-15
 | | offramp: This is not a glitch, just an oddity. If one puts in the two opponents as Lasker and Lasker, you get these 10 games:
http://www.chessgames.com/perl/ches...
...but it's really only 5 games. |
|
Jun-03-15
 | | Annie K.: <cg> this new feature, plus your recent tweak to the PotD, that now jumps to the featured move automatically (on pgn4web), makes me wonder if maybe we could also get straight to the exact position when we leave the Opening Explorer "gamewards"? That is, when we click on a single-game line from the OE screen? That would be a huge improvement. |
|
Jun-03-15
| | chessgames.com: Annie, that's brilliant. |
|
Jun-03-15
| | Annie K.: Thanks. :) |
|
Jun-03-15
 | | WannaBe: Ah, dang. Why didn't I think of that!! |
|
Jun-03-15
 | | MissScarlett: My games are up, finally, but I'm sure I submitted one, Alekhine vs Compton Ellis, 1913, that hasn't appeared. If so, is there a problem? I realise the Capablanca-Manhattan CC game from 1931 was rejected as a duplicate of Capablanca-Fine, and I've submitted a correction for that. |
|
Jun-03-15
| | chessgames.com: That Compton Ellis game had [Result "*"], although looking at it now it's clear it's 1-0. I just added it — Alekhine vs A Compton Ellis, 1913 About that Fine game, we'll have to look at it and do the right thing. By the way, Brooklyn Institute Chess Club and "Brooklyn Chess Club" are the same entity I presume? I updated the name of Brooklyn Chess Club based on your PGN. |
|
Jun-03-15
| | MissScarlett: No, two separate clubs. |
|
Jun-03-15
| | chessgames.com: Oh really, thanks for clearing that up. We wrongly merged them into one record. |
|
Jun-03-15
| | Annie K.: You know you spend too much time at Chessgames.com when... you start thinking a community and the institutionalizable part thereof, are the same thing? ;p But that's ok, I like it here! :) |
|
| Jun-03-15 | | zanzibar: <If you ever believe you have some magic python routine that can produce output like "player #____ is fide #_____" I'd be happy to analyze its output.> chessgames.com chessforum (kibitz #22738) I have just such a beastie...
https://zanchess.wordpress.com/2015... OK, just to be clear I'm not advocating this strategy: <You can also make the argument that sometimes to get something right, you have to start with something that's wrong and fix it. And so we should populate all of these pages with FIDE numbers and then let the editors correct from there.> chessgames.com chessforum (kibitz #22723) That's NOT what I envision...
Again, I deliberately broke up the analysis into the two camps - players where the match is almost-foolproof (only a fool would claim it completely foolproff), and candidate matches that require adult supervision. Again, the foolproof matches are those players where the name is almost an exact match, the dob matches, and the rating is approximately correct. I think I also required no degeneracies, but let's accept <CG>'s challenge and see how the list of matches matches up! Again, see the list here:
https://zanchess.wordpress.com/2015... I believe there are several hundred matches to check, enough to get an error rate down to the percent level. |
|
| Jun-03-15 | | zanzibar: By the way, I'm sticking my neck out here. I didn't go back to recheck the data in the post. So it's definitely a good test of a first pass at doing the <CG>-to-<FIDE> matching. (But I acknowledge these are pre-selected easy cases) |
|
Jun-04-15
| | MissScarlett: I want to check that I followed the correct procedure in amending <...> to the game score of Capablanca vs G A Anderson, 1926 to indicate that the final moves are missing. I notice this means the game is categorised as having been annotated. Surely that shouldn't happen. I further note that G A Anderson does not show up in the player list when <Anderson> is searched for. Why is that? |
|
Jun-04-15
 | | Tabanus: <MissScarlett> Because the limit is 42 names! I wish CG could make the list longer, say 50 or even 100 names because some players are hard to find. See Andersen where there may be more Andersen's than on http://www.chessgames.com/perl/ches.... |
|
| Jun-04-15 | | zanzibar: Now this is truly programming nerd level:
http://sourcecodebrowser.com/scid/3... The source code for SCID's <Crosstable::BestMode>. I was thinking they might have some secret sauce that I might have been missing. Not really, it looks like a good use of straight-forward logic. Specifically, I'm interested if there's a good way to automatically detect a <Knockout>, but SCID's code looks like they don't even try. |
|
Jun-04-15
| | chessgames.com: <MissScarlett: I want to check that I followed the correct procedure in amending <...> to the game score.> Yes, that's perfect. <I notice this means the game is categorised as having been annotated. Surely that shouldn't happen.> No it shouldn't and I didn't think it did. Same thing for designations like "time" or "adjudicated". <I further note that G A Anderson does not show up in the player list when <Anderson> is searched for. Why is that?> There is a limit to how many players a name search reveals and Anderson is a common name so he didn't make the cut. |
|
Jun-04-15
| | chessgames.com: <zanzibar> <I have just such a beastie... https://zanchess.wordpress.com/2015... ...
Again, I deliberately broke up the analysis into the two camps - players where the match is almost-foolproof (only a fool would claim it completely foolproff), and candidate matches that require adult supervision. Again, the foolproof matches are those players where the name is almost an exact match, the dob matches, and the rating is approximately correct.> Ah, a DOB match. OK, that explains a lot. My question all this time has been, "How can we be sure that the software is sure?" My software already make keen guesses but I wouldn't trust it to play unsupervised. Your solution is simple: the DOB is the clincher. If everything else seems to be OK, and at least the birth year is the same, we'll go out on a limb and assume it's the same. A flesh and blood editor would do the same thing. Note that when you have a match that includes DOB match, it must be true that some (flesh and blood) editor added the DOB themselves. Possibly it was added before the advent of the fide# field, or they were lazy, or whatever. That's why DOB matching never occurred to me — I was assuming the vast majority of these players would have no DOB in our database, just like that have no FIDE number. Does your software look at the geography as well? Does it have some table to tell it that "DE" is the same as "GER" etc.? <I think I also required no degeneracies, but let's accept <CG>'s challenge and see how the list of matches matches up!> Your use of the word degeneracies confuses me. Usually degenerate cases involve null-sets; like a game collection without any games. You use the term to indicate what, again? Two players with the same FIDE ID? |
|
 |
|
ARCHIVED POSTS
< Earlier Kibitzing · PAGE 826 OF 1118 ·
Later Kibitzing> |
Kibitzing
