times State of the V:EKN - October 2013

29 Oct 2013 19:22 #55883 by Amenophobis

Even in ELO system more activity means a better rating, if you are good that is. Technically the database would have to be a LOT more complex. And also it would be important that tournaments would be reported in order. Also, 5 player ELO coefficients would be very difficult to put fair coefficients on. Even worse if we need to count ratings per tournament.


A column with date of tournament should do the trick. The data selection program can then process the entries in the correct order, regardless of when specific tournament results would be uploaded.
You could run the program once each month and all reported tournaments are used for calculating the ranking. If at a later time an older tournament gets reported, that should be no problem at all.

Please Log in or Create an account to join the conversation.

More
29 Oct 2013 19:29 #55886 by fredsct

1) I thought you could create an interesting and arguably effective skill rating system by using the chess/ELO-based rating system VEKN once used (but bungled in a technical sense by using a really awful coeffecient).


Technical requirements aside, such a system would have the same issues with locality that the present system has.

...
The data points are localized, regardless of whether it's the current system or ELO. I agree with your comparison/contrasting of the current system and ELO, but the fact remains that the data points from, say, North America don't cross-pollinate with the data points from Australia. With the exception of the few players who do travel inter-continentally, the data is going to remain "siloed" whether it's ELO or the current system.


No, they're not - not a long as the two sums are compared with each other for any important purposes. If Joe from Austrailia travels to Europe and plays in the European CC, wouldn't his Austrailian ranking points matter if he could potentially have a free pass into the second day based on his total?

Furthermore, you don't have to go to a different continent to find a problem. Doesn't the disparity in travel effort make a difference just between different areas in Europe? It certainly does in the U.S., where no comparison can be made for players in Phoenix, Los Angeles, and San Fransisco, three cities in the American SW region which present very different opportunities for tournament play to an average player in each city.

If all you're saying is that ultimately, representing skill as a number isn't possible at all, again, you might be right. But if it isn't, why even have a ranking system? I don't see how implementing a thorough unfair one helps.


I would agree with you if you said the current system is imperfect. I disagree that it is unfair.


Of course it's unfair. How can you say the difference between lots of tournaments held in your home city and in nearby-driving-distance cities isn't completely unfair in comparison with someone who lives hundreds of miles from any tournaments? I'm confused about how you can be missing my point.

The discussion part is great and very valuable, but at the same time you were decrying the proposal as unfair before first knowing how it is to be implemented - and it's still not yet known.


I'm pretty sure I knew enough about it to decry where Ginés was going with
it. "...either by granting the top X players automatic qualification for the Continental Championships and/or for the Second Day of the European Championship..." is enough information to want to lodge my concerns immediately.

Please Log in or Create an account to join the conversation.

More
29 Oct 2013 21:16 #55899 by Robert Goudie

Even in ELO system more activity means a better rating, if you are good that is. Technically the database would have to be a LOT more complex. And also it would be important that tournaments would be reported in order. Also, 5 player ELO coefficients would be very difficult to put fair coefficients on. Even worse if we need to count ratings per tournament.


A column with date of tournament should do the trick. The data selection program can then process the entries in the correct order, regardless of when specific tournament results would be uploaded.
You could run the program once each month and all reported tournaments are used for calculating the ranking. If at a later time an older tournament gets reported, that should be no problem at all.


If you found an error in an old event or a late report of an old tournament then all tournaments would need ro be recalculated...not only the event with the error or missing event. Each player's elo changes between each round so a database would need to store the values as they existed before and after each round. Ultimately we did make progress on fixing the ELO ratings but had to abandon the effort because WW was unable to manage something of this magnitude.

_________________
Robert Goudie
The following user(s) said Thank You: jhattara, Squidalot

Please Log in or Create an account to join the conversation.

More
29 Oct 2013 22:19 #55909 by Amenophobis

Even in ELO system more activity means a better rating, if you are good that is. Technically the database would have to be a LOT more complex. And also it would be important that tournaments would be reported in order. Also, 5 player ELO coefficients would be very difficult to put fair coefficients on. Even worse if we need to count ratings per tournament.


A column with date of tournament should do the trick. The data selection program can then process the entries in the correct order, regardless of when specific tournament results would be uploaded.
You could run the program once each month and all reported tournaments are used for calculating the ranking. If at a later time an older tournament gets reported, that should be no problem at all.


If you found an error in an old event or a late report of an old tournament then all tournaments would need ro be recalculated...not only the event with the error or missing event. Each player's elo changes between each round so a database would need to store the values as they existed before and after each round. Ultimately we did make progress on fixing the ELO ratings but had to abandon the effort because WW was unable to manage something of this magnitude.

Of course everything would have to be recalculated every time the program runs. Since there is a very limited amount of entries in the database, this poses no problem, performance-wise.
All entries have the date of the tournament and the round number fields as key, so are historical correct. Since there are only a subset of all entries (i.d. last 18 months or so), again, performance is no factor. A periodic job can clear all entries in the database that are older than X, just to not clog the database, but OTOH, we are nowhere in 10 years expecting more then a million entries, no?
A working ELO ranking/rating system is absolutely doable with widely available SQL tools. With the caveat that in the time period between any two consecutively runs things might have changed a bit. But I think VTES players are a patient lot and can wait a month for rankings/ratings to be updated. B-)

Please Log in or Create an account to join the conversation.

More
30 Oct 2013 08:54 - 30 Oct 2013 08:55 #55937 by Lönkka

My main proposal is, "stay away from altering the structure of continental championship tournament qualifying and implementation." If you must use it, the Legends tournament proposal struck me as a reasonable outlet.

On this I wholeheartedly agree with Fred!

Finnish :POT: Politics!
Last edit: 30 Oct 2013 08:55 by Lönkka.

Please Log in or Create an account to join the conversation.

  • Lönkka
  • Lönkka's Avatar
  • Away
  • Antediluvian
  • Antediluvian
  • War=peace, freedom=slavery, ignorance=strength
More
30 Oct 2013 09:02 - 30 Oct 2013 09:07 #55938 by Lönkka

You record who the opponents are in a game or tournament and take their ratings into account.

So if you play against players with lowratings you get zero/low points? But if everyone starts off with 0 rating points how will anyone make progress unless we grate/rate players beforehand with some arbitary system (say, we decide that "Ben" is rated as A player and "Janne" is rated as "F" player -which is not that far from reality since I'm sure that at least 99,9% of players will rather have Janne as their predator instead of "Ben"...)

I'm not saying that this wouldnät work, but I'm just curious as it is not a bad idea.

But there are the problems involved:

In a multiplayer game, there really isn't any way to objectively and conclusively rank players. For example, if two players gang up and oust a third player, is that third player a lesser player for being ousted, or a greater player for it requiring two other players to oust them?



But all in all, checking the current World Rankings, I notice that in order to rank at the top you really need to perform well (=win) at BIG event(s). No casual player winning an occasional local tournament and being in the finals every now and then is ever going to threaten the top players. You could still make it to the high rankings if you manage to, like, win 8 local tournaments. So I'm not all that convinced that the current system is all that faulty either.

Finnish :POT: Politics!
Last edit: 30 Oct 2013 09:07 by Lönkka.

Please Log in or Create an account to join the conversation.

  • Lönkka
  • Lönkka's Avatar
  • Away
  • Antediluvian
  • Antediluvian
  • War=peace, freedom=slavery, ignorance=strength
More
Moderators: AnkhaKraus
Time to create page: 0.080 seconds
Powered by Kunena Forum