The elostat program was great for generating ratings from one tournament but I'm not sure how to handle them going forward. For example, as new people join the ninja pool I don't want to re-estimate average ratings.

I was thinking about for all new games using the ICC formula which works as:

I think the ICC formula is similar to how USCF calculates ratings. People with less then 20 games are provisional and I have a general understanding how that works but I don't understand exactly how it works.

Anyone have thoughts about this? How to handle ratings going forward? How exactly are provisional ratings calculated? Anyone ...

One thought I had is this one (from a practical point of view).
If we can generate ratings from the first tournament, then we put them in that order for the pairings in Tonc2 or Letonc1.
If we get new people (as is the case), then we work like this:
If those new people have an official rating, then we put them in the list according to that rating.
If they don't have a rating, then we put them in the middle of the bunch, assuming that the odds are best served there.

This is a good enough formula. The provisional ratings have to do with the k constant, which changes (!!) after a certain amount of games, making rating (de)increments smaller with time. I think the world standard provisional k is 40, and it's lowered to 20 (or should it be 10?) after 20 or so games.

And Obi1, the rating idea doesn't sound too good for me, as there are ninjas all around the world, each one with a different federation and different kind of rating calculation. Unless we only accept FIDE ratings, which I doubt many of us have.

I think we should always consider "ninja" ratings. If the player doesn't have a ninja rating yet (at this time this means he didn't play in TONC), then he's listed, in alphabetical order, at the end of the rating list. Then we make the pairings.

What do you think?

Greetings everyone! I have returned from a vacation in Utah to visit my sister.

I flew back home yesterday, conncecting at JFK airport at 9 AM. Phew! Twelve hours later and I might still be there

Ha ha ha, I am the Average Ninja. Base rating of 1400 gives me 1406, I get ranked exactly in the middle... hehehe, cool.

Surf, what formula was used for generating the original ratings? I don't see anything wrong with it.. or is ICC one a lot better? I don't see a problem with the one you've been using.

As for provisional rating, I'd say all new people should get a rating of 1400. Those who are vastly better will shoot up fast, as Obi1/Rimfaxe did.

I'm not sure what formula the program, uses but I assume it is something similar. It might be in the documentation somewhere. It uses an iterative process where everyone starts with the same rating then it runs through all the games and calculates new ratings, then does this over and over again using the newer ratings until the ratings stabilize. I think it did this 47 times to generate the ninja ratings. So essentially it just calculates the relative difference between the players and centers it around the mean that I specified (1400).

I think in typical ELO type rating lists the mean is 1600, I used 1400 since it was mostly a white belt tournament. What I don't want to start doing is continually try to guess the average rating as new players enter the tournaments. And at some point with enough players and enough games it isn't practical to generate ratings like this, it would take even a very fast computer a long long time. However for our needs maybe using this program will be sufficient for quite a while. So maybe what I will do is continue using it with a mean rating of 1600 and forget about how our pool matches up with other rating lists. Everyone gets a start rating of 1600, so you can just add 200 points to whatever your rating was that I posted. And I will just maintain a master PGN list of all Ninja games. This will work for quite a while and means I won't have to do anything new. Have I mentioned that I'm lazy? Sometime in the future when there are hundreds of ninjas competing and tens of thousands of games, ratings will need to be calculated differently.

Is everyone okay with this?


Edit:
Welcome back Jsweet. Utah is the only place in the US that I've been where I feel like I'm in a foreign country

HAH.. riiiight. :p

Seriously though, TONC's will happen every 4-6 months or so. I don't think it'd be too bad to put in a few new people with an average rating every tournament. You lost me on that other stuff though, so hopefully someone who knows ratings better can answer..

I just like using 1400 better. I am 1200 OTB and I was middle of the pack in the tourny, so I think 1600 would be too generous.