User talk:Wesalius

Jump to navigation Jump to search

About this board

Previous discussion was archived at User talk:Wesalius/Archive 1 on 2015-11-10.

Manu1400 (talkcontribs)
Reply to "Empty list: Elo recent"
Steak (talkcontribs)

Is EloBot done with importing the Elo ratings from 2001 onwards?

Wesalius (talkcontribs)

Only this month (11/2016) is not added. But I cant work on it now, since latest update to wikibase (the software that wikidata runs on) prevents my script from adding any more statements.

Wesalius (talkcontribs)

It has been fixed. Now all the available data from FIDE since 2001 are imported.

Steak (talkcontribs)

Ok, thanks for your efforts. When are you going to import the data before 2001?

Wesalius (talkcontribs)

I think I already wrote to you in a different conversation that I dont see it anytime soon since the data in benoni and ilimpbase are formatted differently if I recall correctly.

Steak (talkcontribs)

Did you make any progress in fetching the elo ratings for the time before 2001 from the websites benoni or olimpbase?

Wesalius (talkcontribs)

Nope. The same "on hold state" as with the federation countries applies to this as well.

I may be mistaken, but from what I remember when I briefly checked the files from those sites, they were not produced as "elo ratings of players per month". If the data structure is different to fide files which I work with now, then it will be a possible problem, because the script I use now is written for "elo ratings of players per month" files.

Wesalius (talkcontribs)

Maybe you could contact people behind benoni and olimpbase if they could produce their output data in "FIDE" formatting, so it could be easily uploaded to wd using the script without major changes...

Steak (talkcontribs)

I sent a message to benoni. I doubt that the guy behind olimpbase will have time for this, he is always very busy (he is mayor of a town), so I did not ask him yet.

Steak (talkcontribs)

Benoni has answered. He told me that extracting the ratings from the old FIDE lists is quite problematic and cumbersome because the FIDE ID was not constant in these times, transcriptions from non-latin-letters may slightly change over time and players (especially women) may have changed their family name. So he does not have a complete dataset of all the ratings and he can not provide us the data as "elo ratings per players per month". He offered to help us, but at the moment he does not know how he can help us.

Wesalius (talkcontribs)

If there are such inconsistencies in the datases, they might not ne suitable for bot ať all...

Wesalius (talkcontribs)

Thank you for asking anyway.

Steak (talkcontribs)

Could you import with your bot the years when the titles (first e.g. the grandmaster title) were approved? A good source could be the german list "Liste der Schachgroßmeister" (linking does not work) where all title years have been added and checked by different sources. Not a good source is definitely the FIDE database. Many years are missing or wrong.

Wesalius (talkcontribs)

Yes, it is possible. The dates at are meant to be date qualifiers for the title values at the player items? Or what property do you think should be used?

The easiest for me would be if you would give me a list of player items with the date values so I can pass it into quickstatements.

Wesalius (talkcontribs)

Or I can actually make quickstatements figure out the items from the de.wiki articles, so no need to get the list. Just figure out what should be the appropriate structure to store it (date qualifier for title values at the player items, or something else?)

Steak (talkcontribs)

I was thinking of adding Property:P580 as a qualifier to the grandmaster statement, like here: One could also use Property:P585, but I think P580 is more appropriate because from there on the player is grandmaster (and there is the potential that the title gets revoked, so we would need Property:P582 additionally.

Wesalius (talkcontribs)

Ok. Would you please help me finish preparatory list at Benutzer:Wesalius/List?

Steak (talkcontribs)

How shall this list look like? The names as written there and the form "<Given Name> <Family Name>,<Year>", or the english transkriptions (which would be quite a workload). I think for a large part of the names the german wikidata label could be used to link the names to the Items.

Wesalius (talkcontribs)

Most of them are already done. They should like like "name of de.wiki article","year" pairs.

Wesalius (talkcontribs)

Will you work more on the list or not?

Steak (talkcontribs)

What needs to be done? The links are all correctly formatted I think. There are some years not matched to a link, what about them?

Wesalius (talkcontribs)

Those abandoned years just need the appropriate player. If you dont match them (can be found when checking the player above or below) then I will just skip them no problem.

Steak (talkcontribs)

Maybe I have time to work on the list during the weekend. You can start importing on monday. :)

Wesalius (talkcontribs)

Ok, its your request, so when you say we ready then we ready ;-)

Steak (talkcontribs)

Could you briefly sketch how you plan to match the list entries with the wikidata items? I just want to be sure that your bot will work fine.

There are at the moment 1682 items with the title grandmaster, but the correct number would be 1678, so this year import provides also the possibility to correct the statements (effectively, four are wrong, but from experience, I think there are more than four added wrongly and somewhere else the statement is missing). Your bot will hence have a residual list where he could not add any year. I would need this list also to clear up the rest.

Wesalius (talkcontribs)

For players that have their article written, I dont have to match anything, its pretty error proof. Go to http://tools.wmflabs.org/wikidata-todo/quick_statements.php and try yourself. Put dewiki and then follow this example for any player that has article on dewiki:

"Jacob Aagaard,2007" -> "Jacob Aagaard    P2962    Q105269    P580    +2007-01-01T00:00:00Z/09    S143    Q48183"

Steak (talkcontribs)
Wesalius (talkcontribs)

Has to work, just put the "Jacob Aagaard    P2962    Q105269    P580    +2007-01-01T00:00:00Z/09    S143    Q48183" in Calc/Excel columns like

Jacob Aagaard P2962 Q105269 P580 +2007-01-01T00:00:00Z/09 S143 Q48183

then copypaste from there to quickstatements. Dont forget to put dewiki into the little window above

Steak (talkcontribs)

Ok, worked, the tabstops were the problem.

Wesalius (talkcontribs)
Steak (talkcontribs)

What about looking for the date of birth? This would narrow it down to maybe three possible items for each list entry, and the correct one could then choosen by hand.

Wesalius (talkcontribs)

Probably the easiest is to create a query that looks up item that:

  • matches the label
  • matches the birthday
  • has P2962:Q105269

and then to recycle such a query with different labels.

Steak (talkcontribs)

Would it be possible if I run the quickstatements first for the articles that exist, and then you run the query which compares the birthdays for the rest?

Wesalius (talkcontribs)

How about you do both (since you know how)? Or you want to do just the no-work part?

Steak (talkcontribs)

I dont know how to run this query and then match the list entries to the results. I am already cleaning up the statements for the future year adding, this is also work :)

Wesalius (talkcontribs)

It is work that you came up with ;-) Go ahead with the quickstatements ;-) I will look into the queries later...

Steak (talkcontribs)

Why is it in you example above "2007-01-01"? We only know the year, so stating it at day precision would be wrong.

Wesalius (talkcontribs)
Steak (talkcontribs)

Yes, the precision is year, but internally, it is saved 2007-01-01. I think it would be more appropriate to add the data as "2007-00-00", to avoid any possible confusion.

Wesalius (talkcontribs)

You are the one who is going to add the data, so do it that way if you think it is better.

Steak (talkcontribs)

The quickstatements-list has run through. Now you can go on with the query. If you want me to help, please ask. :)

Steak (talkcontribs)

Hi! At the FM Mix'n'match catalogue, there are a few titled players (mostly FM, a few IM) remaining without item. Can your bot maybe create these items?

Wesalius (talkcontribs)

So all of these in the catalogue yo are linking are notable and should have their item?

Steak (talkcontribs)

Yes.

Wesalius (talkcontribs)

Done

Steak (talkcontribs)

Thanks!

Steak (talkcontribs)
Wesalius (talkcontribs)

Ooooh, thank you! They look great! I might reuse them on cs.wiki.

PokestarFan (talkcontribs)
Steak (talkcontribs)

What about importing the Elo ratings since January? Is there a problem?

Wesalius (talkcontribs)

No problem, I am just busy in real life, graduated from an university a month ago, everything is settling down.

Wesalius (talkcontribs)

I am working on it, one of the scripts which reformats the data from fide refuses to work with me, but hopefully I will make it work soon and the uploading will start again.

Wesalius (talkcontribs)

I resolved the issues, the elo ratings are being added.

Steak (talkcontribs)

Nice :)

Steak (talkcontribs)

There are some items where the bot skipped adding the month qualifier in the first run in November/Dezember 2016. I removed such Elo statements because it does not make sense without month. Is it possible that your bot checks at all items if all Elo statements are complete for all available months, and adds the elo rating of missing months?

Wesalius (talkcontribs)

Yes, it will add them this time or in another iteration. It goes through all the previously added months and checks if the value is already added, if it is not (or have been deleted) then it adds it, so if it will be added now or next time depends on if that month has already been checked.

Steak (talkcontribs)
Wesalius (talkcontribs)

Hmm, it seems there were some errors when parsing the xml source file. Since there is just couple of them I will fix them after the run is done, thank you for noticing.

Reply to "Elo ratings"
Steak (talkcontribs)

In Q24959896, April 2008 and April 2009 are wrong. What happened?

Wesalius (talkcontribs)

I have no idea where are these coming from. Did you see some values like these anywhere else?

Steak (talkcontribs)
Wesalius (talkcontribs)

Good, I will fixem now.

Wesalius (talkcontribs)

Now, they have this ±1 at the end, I do not know how to remove them...

Steak (talkcontribs)

[https://www.wikidata.org/w/index.php?title=Q24959896&type=revision&diff=391109268&oldid=391108463] Fixed. Just type "2333+-0" ;)

Wesalius (talkcontribs)

Thanks, I will keep that in mind for next time.

Steak (talkcontribs)

Hi. Your Bot is again adding birthyears as Elo ratings.

Wesalius (talkcontribs)

Thank you for noticing, the source fide ratings file for April 2007 had inconsistent number of spaces compared to some other files. I thought I catched all the inconsistent files, but this one escaped me. I will repair my mistake soon.

Wesalius (talkcontribs)

I double checked now the rest of the data and there should be no more birthyears mixed in ratings. In case of the error (unlikely, but always possible) happening again do not hesitate to block the bot.

Steak (talkcontribs)

Thanks for checking. But I am not an admin, so I cannot block the bot. ;)