The NERV: North East Region (7/31/09 update)

[Jexik]

Originally Posted by Shades fan

Maybe someone would start one for each area! I would enjoy seeing one for the West coast,and Texas,because it is so big of a Tournament state.

I'm not sure if all of the data is still handy for all of that stuff. Don't forget IA, MN, and OH. They run some big tournaments up here too. The size and number of the events is what makes creating an ELO system very difficult to do at this point in time I'd imagine.

[Shades fan]

Originally Posted by Jexik

I'm not sure if all of the data is still handy for all of that stuff. Don't forget IA, MN, and OH. They run some big tournaments up here too. The size and number of the events is what makes creating an ELO system very difficult to do at this point in time I'd imagine.

That is why he said that he was only doing it to tournaments he goes to,along with only in his area,it wouldn't be as manageable,or as correct as it is today.

[ollie]

Thanks Doustin (and others) for helping me scrape up some older info. One problem I have is when the tournament ranked by points rather than win/loss.

To do the Elo properly, I need the results of all of the individual match-ups. To do anything at all, I need to know the number of wins and losses for each player. n the second case, I'm thinking that I'll add 8 for each win and subtract 8 for each loss. This will be more conservative than the Elo changes, but still get people moving towards their more settled position on the scale. (I'm only happy to do this because it will be for the events at the birth of the scale---any events from here on in will need the full pairings info.)

Oh, and it isn't only tournaments I go to. Any tournament that has a sizeable proportion of NERVers is fair game. For example, if NERVers are going to GenCon and play each other in a tournament there, I'll include that game on the rankings (in that case it won't be worth doing the whole tournament(s) and so games between NERVers and non-NERVers won't be counted). Another thing that might be cool is to have semi-competitive game days. 4-6 people playing a full round-robin for example. With a high proportion of NERVers involved, I'd be happy to include such events in the ratings.

As for other areas---I don't think it is that big a job to do one single area. As I briefly explain in the FAQ it's not really the size of the project that makes a national system unworkable, it's the lack of mixing among different groups of players. Elo is a comparative system and needs a high level of mixing to stay accurate.

[Jexik]

Originally Posted by ollie

As for other areas---I don't think it is that big a job to do one single area. As I briefly explain in the FAQ it's not really the size of the project that makes a national system unworkable, it's the lack of mixing among different groups of players. Elo is a comparative system and needs a high level of mixing to stay accurate.

I'm just saying that the paper records of two-year-old tournaments with 30-60 people are probably lost by now.

[ollie]

Originally Posted by Jexik

I'm just saying that the paper records of two-year-old tournaments with 30-60 people are probably lost by now.

That's not a huge issue. Just start now and go forwards. The further back in time you can access, the more quickly players will reach a fairly settled rating. However, there is nothing wrong with declaring today to be the first day of the system and everyone moving on from there.

[RobertDD]

When will you be updating the first page ollie?

[guido]

Originally Posted by Shades fan

That is why he said that he was only doing it to tournaments he goes to,along with only in his area,it wouldn't be as manageable,or as correct as it is today.

Which gives him a HUGE edge in his scoring. I suggest if anyone holds a Tourney that Ollie can't go to in the New England aria, that they send him the results. I don't mean to imply that you are cheating in any way, ollie, and you seem to go to most Tourneys anyway. I just think it's more fair for those who live in the outskirts like Main or New York (not part of New England, but they have some NERV invading our Tourneys). If there is an open Tourney in their aria that some of us souther New Englanders can't get to, and Ollie can't make for whatever reason, they should still be recognized.

[RobertDD]

Originally Posted by guido

[...]I just think it's more fair for those who live in the outskirts like Main or New York (not part of New England, but they have some NERV invading our Tourneys).[...]

Last time I checked, the outskirts of Maine were still New England? Or are you talking aboot Canada?

[ollie]

Originally Posted by guido

Which gives him a HUGE edge in his scoring. I suggest if anyone holds a Tourney that Ollie can't go to in the New England aria, that they send him the results. I don't mean to imply that you are cheating in any way, ollie, and you seem to go to most Tourneys anyway. I just think it's more fair for those who live in the outskirts like Main or New York (not part of New England, but they have some NERV invading our Tourneys). If there is an open Tourney in their aria that some of us souther New Englanders can't get to, and Ollie can't make for whatever reason, they should still be recognized.

Going to more tournaments does not actually give someone an advantage. Each player's rating will go through an initial period of fast change and then settle in some fairly small region*.

Going to more tournaments means that the initial period happens faster. So if two people are both play at the 1400 level and the first goes to a lot more tournaments, then that first player will reach the 1400 region faster, and so have a higher rating until the second player catches up. However, one of the cool things about the Elo system is that once you've reached your 'true' rating, playing more games does not improve your rating (unless your playing skill improves during those games). At the moment, and for a while, we're in the initial fast change phase for everyone. I'm expecting at least a 500pt spread from top to bottom once things have settled down.


*This makes various assumptions, many of which are debatable. I hope we'll have the debates at some point, but I don't have time to lay out the issues at the moment.

EDIT: I said above somewhere that I don't have to be at the tournament for it to count. I will be contacting the organisers of any such event to get the info I need for the ratings.

[ollie]

Originally Posted by RobertDD

When will you be updating the first page ollie?

I've done all of the initial prep work for the BatB tournament. Before I can run the numbers through I need to include any data I'm going to from earlier tournaments. Uprising is seeing what info he can find regarding Chaos in the Capital; I have a bunch of scores from Doustin from Roll Shields or Die---I need to see if I can recreate win/loss records from that; I'm waiting to hear back from Pilgrim on past Beatdowns.

Does anyone know of any other tournaments? And if anyone can post their own records from those tournaments (or particular results of others) that would be a huge help.

Let's say Sunday. Any information from before Marlboro vs. The World needs to be collated by Sunday morning. I'll then run everything up to and including BatB and post it on Monday morning. Does that sound like a reasonable plan? Suggestions are always welcome.

[guido]

Originally Posted by RobertDD

Last time I checked, the outskirts of Maine were still New England? Or are you talking aboot Canada?

I was implying that Mainiacs might have a hard time getting to a CT Tourney and that New Yorkers (which there are a few of in the NERV) might have difficulty making it to a NH or Maine Tourney.

[ollie]

More thoughts on past tournaments.

The point of including the results of past tournaments is to accelerate people's progress towards their 'true' rating. I've been trying to think of a suitable way to do this when the only available data is the final rankings, without having the additional win/loss info. This is what I've come up with:

Suppose a tournament has n players. Award the winner 2n points, second place 2(n-2) points, third place 2(n-4) place, and so on down to halfway in the standings. Similarly, dock last place 2n points, second-to-last 2(n-2) points, and so on up to half way.

Thoughts? None of this is really a big issue. Getting this part as accurate as possible merely means we'll reach our stable state sooner. Nothing goes wrong in the long run if we assign everyone totally arbitrary (and different) scores at the start and work from there.

Another problem I realise I'm going to have is identifying people. Before I run the numbers I'll post a list of names I have. There will be repeats on it because I won't know that "Y" and "X's Son" are the same person. Help with condensing the list will be appreciated.