One is basing off past finishes, the other off past performances, there's a difference. Which I'm sure is what the person you quoted meant, not that it was predicting the future.
Fargo Ratings vs WPA Rankings?
- Thread starter mikepage
- Start date
One is basing off past finishes, the other off past performances, there's a difference. Which I'm sure is what the person you quoted meant, not that it was predicting the future.
There is a difference between the most credible people, those in the industry, questioning something and those that do not understand, questioning it.
You might want to question the top people in the industry that know Mike before you make statements about his credibility.
Sent from my iPhone using Tapatalk
Banks
Banned
The biggest questioning of ratings has pertained to us regular schmucks. The pros tend to have many, many more games in.
Why? They don't have special knowledge of how other people are viewing the believability of the ratings.
Please note: I am not making a comment ABOUT Mike, I am making a comment on what people are SAYING. That is, that they don't believe in the efficacy of the ratings. Do I need to give examples?
Apologies to Mike if that wasn't clear. I am a fan of the ratings (particularly the sound mathematical basis for them) as can be seen by my previous posts. I just want to make sure that mathematical basis is not strayed from.
Thank you kindly.
I would think you would agree the small percentage of negative people on here have only recently saw their rating and are unsure of how it works That doesn't effect credibility.
Credibility is about facts
People can have their own opinions but not their own set of facts.
You are a very smart individual, and you understand people are mad because the truth of their ratings is sound and they don't like what is saying
Sent from my iPhone using Tapatalk
Yeah, Fargo ratings are interval data and WPA rankings are ordinal. But I just checked, and the WPA does have points, and the ranks are based on those points. So I think if Mike does the comparison he ought to use WPA points rather than the rankings.
Yes. The next paragraph is technical, so skip it if you're allergic or don't care
We maximize the Log-Likelihood function to get the ratings. The matrix of second derivatives of the log-likelihood function--formulas for which we have derived--is called the Fisher Information Matrix. Diagonal elements of the inverse of the Fisher Information matrix are--according to the Cramer Rao Inequality lower bounds to the variance--i.e, error bars. Off diagonal elements of the Fisher information inverse also tell us the covariance--the degree to which one player's rating is coupled to that of another. So yes, we know how to get error bars and how to get estimates of error bars.
We don't actually "call" matches. Suppose two players are rated 766 and 750. The 16-point gap says the 766 will 60% of the time win a race to 11. That means 40% of the time the 750 will win. This uncertainty in the outcome of an individual match is there even when we are certain of the ratings.
When we are uncertain of the ratings (the real rating gap might instead of being 16 points be 26 points or 6 points or might even go the other way), then we have uncertainty not only in the outcome like above but in the 60% itself. Regardless, 60/40 is still our best guess for these players, and if we are going to devise a test to figure out whether our ordering of players is superior or inferior to some other ordering in predicting match outcomes, this is what we should use. Yes some of the matches are going to be coin flips--no harm no foul on that.
I appreciate the support from some of you, and I do think we often talk past one another, but I actually don't take exception to what Mr. Corwyn_8 says.
A lot of people look our efforts and both don't know us very well and don't understand the details--understandable on both counts. When they see a list of players, either pros or players from their area, and a player rating pops out as not passing their smell test a red flag goes up. Also understandable. We live in a world with a lot of people out there making unsupported claims.
We hope, slowly, to gain people's trust. And we think we will. We have seen lot's of red flags being waived when we put stuff out there
Johnny Archer is no longer a top US player
Jiaging Wu should be nowhere near the Ko's & crew
Justin Bergman is not a top world-class player
Efren shouldn't be invited to Bigfoot
and so forth...
I think our assessment has and will continue to be vindicated.
For now, for me to claim without even knowing who the 32 players are that Fargo Ratings will doing a better job guessing match outcomes than WPA rankings, I'd say that's a pretty bold statement...
I understand, and I do think you have a legit point about close Fargo ratings. I just think it was a little over the line to say he has a credibility problem, which to me implies more than just "there are a few skeptics."
Not to derail this thread, but here are some OR players, including you, with 200+ games. I'm curious who you think is way off base. Sometimes it is just that more data is needed to get things into line (for instance if you have 200 games and a lot of them are against nearly unrated opponents, your rating might not really be well established. But a bigger issue is when games for Mike Davis league player get assigned to Mike Davis pro player or visa versa. So if there players who seem way off I can check for that...
Attachments
Banks
Banned
There were some numbers that were suspect prior to the recent adjustment. Now there are several (regional) masters about 20+ points below myself, an A. Whereas, before, i was about 20-50 below them. Actually, iirc, most of those directly below me are local masters. This may be a case of the separation of divisions causing differences in numbers.
A friend of mine did well enough in the A to move to master, so his rating was around 630 or so. The adjustment brought him down to 570s. Now, mixing those numbers into a national tournament is just as likely to cause problems. Also, there are concerns with cutoff dates, etc. For example, if i find a team that's just under the cap, do we stop playing rating events three months prior to Vegas? If we're a couple over, does someone just halfass a match? Those are a couple of reasons i have reservations over using the ratings without getting a lot of data in first. I feel the same way about arbitrarily moving people around in divisions. (They put out a local guideline saying something like an apa 5 or above could be started as an A.)
Also a little curious as to how my older performances in the B division affected my rating.
The problem with the WPA rankings is they are only accurate in ranking the most consistent participants. Many of the players lower on the list only play a few events, if that, and despite possibly having great finishes they get ranked outside of the top 50 or even top 100.
It's not like the World Snooker rankings where the majority of the players are playing consistently.
It's not like the World Snooker rankings where the majority of the players are playing consistently.
There is no recent "adjustment." On the other hand some tournaments like Western BCA come in with a lot of games, and that may impact many of your opponents, and that also impacts you. Also, some past league data has been coming in in clumps.
[...]
No. This is accounted for in the guidelines. here is a "cushion" such that if players drift up between sign-up time and tournament time they are still good withing the cushion.
Hard to say off hand. You have performance from about the last five years at western BCA, and the first few years carry considerably less weight in determining your rating. I'll contact you offline to get a sense of which of the above players seem off to you so as not to further hijack this discussion
OK now after communicating with Banks I understand. Again apologizing for derailing my own thread...
The friend of Mr. Banks has 76 games, all from one tournament from a few years ago and he performed at 646-speed for that one A-Division tournament at Western BCA in which he got 4th.
CSI considers him an "open" player according to their old categories.
We don't have enough information to say he's really a 646. Maybe he had an unusual day.
Here is what we do. We assign a "starter rating" of 525, which is the same for all formerly-classified-as-open players. And the influence of that starter rating diminishes until the player has 200 or more games and then it is forgotten.
This guy is about a third of the way to 200 games, so his performance number of 646 carries about one third weight. The 525 starter rating carries the other two thirds weight. As a result, for BCAPL in Las Vegas he would play as a 570 (the weighted average).
None of this affects the players I listed above because they all have 200+ games. Note that the old way--calling this guy "open" is equivalent to treating all open players the same, i.e., like they all have a 525 rating. The fact we have SOME information that distinguishes him from other "open" players is a good thing--and an improvement.
As always. Thanks for the thorough explanation
Sent from my iPhone using Tapatalk
Sent from my iPhone using Tapatalk
I think the problem here is that we have completely different definitions for 'credibility'. To me credibility is: "the quality of being trusted and believed in." - Google definition.
Thank you kindly.