A Questoin For FargoRate
- Thread starter watchez
- Start date
The system has two major flaws imo.
1) Incomplete data.
There is no way for the system to record every competative game on every player and in when games are left out the system suffers.
2) Not all games are equal.
Even Mike said earlier in this thread that at some point a more refined approach would be ideal. There are numerous examples for this need but I'll just give one. Suppose a world class player from the Philippines, who already has limited data on record, plays in a tournament that is getting recorded... and he is matched up against a 9 speed who catches a gear and plays a perfect 11 games on a bar box and shuts out the world class player 11-0.
Now that world class player, who already has limited data, has getting beat by 11 games from a 9 speed who played perfect on a bar box on his record. The world class player might not have even made a mistake but this one match will have an impact on his rating for a while now in this system.
It's not a perfect system imo by any means. I think most can just use good old fashion common sense and be more accurate.
1) Incomplete data.
There is no way for the system to record every competative game on every player and in when games are left out the system suffers.
2) Not all games are equal.
Even Mike said earlier in this thread that at some point a more refined approach would be ideal. There are numerous examples for this need but I'll just give one. Suppose a world class player from the Philippines, who already has limited data on record, plays in a tournament that is getting recorded... and he is matched up against a 9 speed who catches a gear and plays a perfect 11 games on a bar box and shuts out the world class player 11-0.
Now that world class player, who already has limited data, has getting beat by 11 games from a 9 speed who played perfect on a bar box on his record. The world class player might not have even made a mistake but this one match will have an impact on his rating for a while now in this system.
It's not a perfect system imo by any means. I think most can just use good old fashion common sense and be more accurate.
Last edited:
What about the TAR match, SVB's coming out party where he skull f'k'd Corey in a race to 100? I can only assume you don't have that in your data since you only show SVB winning 95 games total on a big table. You previously posted in a different thread that you go back in time to keep Archer, who never plays, up in the overall rankings but SVB match doesn't count?
SVB needs his congratulations on his fine shooting as well.
It sounds like they're considering weighting games by importance. I'm not sure if that needs to be done, but if it is done, how about this: Weight the games based on money. So a tournament like the World 9-ball with a $30000 (I think?) first prize would be weighted 3 times a big regional tournament with a $10000 first prize.
This has the advantage of being numerical and objective, and it also makes sense that 1) people probably take it more seriously if there's more money on the line, and 2) the prestige of the tournament corresponds pretty closely to the payouts.
It also might help prevent sandbagging in smaller tournaments to lower one's Fargo rating until a bigger payoff comes around.
This has the advantage of being numerical and objective, and it also makes sense that 1) people probably take it more seriously if there's more money on the line, and 2) the prestige of the tournament corresponds pretty closely to the payouts.
It also might help prevent sandbagging in smaller tournaments to lower one's Fargo rating until a bigger payoff comes around.
It is currently and always be an imperfect system. People are involved and luck to a certain extent. Just like football, no matter that the line on a game is 18 points does not mean that the favored team will always win, sometimes shit happens, end of story. Just because the favored team lost does not mean that they are not a better team though, they lost do to circumstances, the exact same thing happens in pool everyday. Fargo does give you a solid idea of who the better player is and it does a fantastic job of that. We have had mypoolstats (which is based on Fargo) here in WI for at least a few years now and I can tell you from experience that when you play someone 50 points higher than you that you will most likely have to play pretty good to beat him, same thing applies if you play a player 50 points weaker than you. It does not always mean the stronger player will win, but more often than not the stronger player wins in a longer race. Before everyone starts bashing Fargo they should wait till they become established and see how it works, I think everyone will understand better at that point.
Thanks...
I'll respond to several of the comments here.
(1) ELO-based methods. This means different things to different people; the single unifying feature is that a player's chance of beating an opponent depends only on the rating DIFFERENCE between the two players. That statement forces a certain kind of mathematical relationship that is common to all ELO approaches. This says nothing, though, about how to GET the ratings, and how the tentative ratings get informed by new results. This is where the approaches differ. And this is where FargoRate gives FIDE--the international organization that rates chess players as well as many many implementations of ELO-type schemes--the orange crush...
2 selected data I've heard before the suggestion that if we don't record ALL tournaments or ALL matches there is a bias or a problem. This is not true. We can rate players accurately by recording only every third game they play or every match they play on odd-numbered calendar days, and so forth. We don't need to get all the tournaments or all the data. What we need is to get enough data--more is always better--and to have no bias in he selection process. If we only put in tournaments for which a particular player won his first-round match, that would be a problem.
Think of it like this. Somebody gives you a weighted coin that gives heads 61% of the time when flipped. You don't know that number but you want to figure out whether it is a fair or a biased coin and if biased by how much. You can get insight into it by flipping the coin many times. I've got excel in front of me, so I actually did this experiment. After 10 flips results are 5,5 for heads,tails. You can't tell much from that, After 100 flips it is 64,36. It is looking like the coin is biased in favor of heads, but you can't really tell by how much. 64% would be your guess. After 1000 flips results are 609,390. Now you can bet your lunch money the coin is biased. After 10,000 flips results are 6081, 3916. Now if that coin was actually flipped 30,000 times, it doesn't matter if you chose the first 10,000 or the second 10,000 or every third flip. The quality of your assessment is characteristic of 10,000 flips.
(3) Why don't we give the "formula."
There is not a formula to give. It is a mathematical optimization process. You can think of all the ratings of every player in the world as variables. And you can think of specific game/match results among those players as outcomes that are either likely or unlikely depending upon on the values of the ratings. Then you can imagine there is a set of ratings for which every match in the system coming out the way it did is most likely. This is a technique in statistical inference called maximum liklihood. The results are well defined. We just can't point to a simple formula.
4. Watchez's comments You said you don't like the font or design of our website. That kind of feedback is important to us. You also said you didn't watch the videos. That disappoints me. If you are interested, then I encourage you to start by viewing the videos. If you are not interested, then...well.. scratch that...you started a thread about it....
You said something about "Archer, who never plays." We have, just in 2015, the following tournaments for Archer: US Open 9-Ball, Jamaica Joes $2000 entry tournament, Carom Room tournament, Turning Stone, World Pool Masters, US Open 8-ball, US Open 10-ball, and Memphis Open. It is hard to imagine what are your criteria for being an active player...
not necessarily...
Winning a race to 5 5-0 may have more of an effect than winning a race to 100, 100-99...
Jaden
Winning a race to 5 5-0 may have more of an effect than winning a race to 100, 100-99...
Jaden
JC
Coos Cues
I checked out a bunch of players here in our Western BCA and am surprised how good of a job the fargo ratings seem to be doing as far as accurately rating players.
My hope is that the WBCA has in mind to use the Fargo ratings to take the decision of who plays in what division out of the league operators and competition committee's hands. Some thing like if your rating is above 595, you play in the masters. 675 you're a grand master. Below 425 and you're a B etc. My guess is that 's what they are up to since they are now scoring the team events for Fargo.
The other thing I noticed is I saw a couple of players who's names had changed either with a different first name or through marriage and both names were rated about the same. A good example is my friend Mary Coffman who was Mary Hopkin before getting married. She has massive robustness under both names and is rated about the same. Another example I noticed was Larry Sowell and Lawrence Sowell who are in fact the same person and are rated almost exactly the same with a different set of games and opponents being considered.
JC
My hope is that the WBCA has in mind to use the Fargo ratings to take the decision of who plays in what division out of the league operators and competition committee's hands. Some thing like if your rating is above 595, you play in the masters. 675 you're a grand master. Below 425 and you're a B etc. My guess is that 's what they are up to since they are now scoring the team events for Fargo.
The other thing I noticed is I saw a couple of players who's names had changed either with a different first name or through marriage and both names were rated about the same. A good example is my friend Mary Coffman who was Mary Hopkin before getting married. She has massive robustness under both names and is rated about the same. Another example I noticed was Larry Sowell and Lawrence Sowell who are in fact the same person and are rated almost exactly the same with a different set of games and opponents being considered.
JC
With apologies to Mr. Page, here it is for all to see:
If Player A has a rating of R_A and Player B a rating of R_B, the exact formula (using the logistic curve)[12] for the expected score of Player A is
E_A = \frac 1 {1 + 10^{(R_B - R_A)/400}}.
Similarly the expected score for Player B is
E_B = \frac 1 {1 + 10^{(R_A - R_B)/400}}.
This could also be expressed by
E_A = \frac{Q_A}{Q_A + Q_B}
and
E_B = \frac{Q_B}{Q_A + Q_B},
where Q_A = 10^{R_A/400} and Q_B = 10^{R_B/400}. Note that in the latter case, the same denominator applies to both expressions. This means that by studying only the numerators, we find out that the expected score for player A is Q_A/Q_B times greater than the expected score for player B. It then follows that for each 400 rating points of advantage over the opponent, the expected score is magnified ten times in comparison to the opponent's expected score.
__________
I'm glad we could clear that up.
lol... That is for going from the rating difference to the chance of winning. That's the easy part. I assume when people are asking for the formula they're interested in how the ratings change with new information.
Incidentally, for the expressions you posted, two players win games in the following ratio:
RATIO = 10^(ratingdifference/400)
(That's 10 raised to the power [ratingdifference/400])
For Fargo Ratings, two players win games in the ratio:
RATIO = 2^(ratingdifference/100)
Note the "2" vs the "10" and he "400" vs the "100." These differences are inconsequential and are just a matter of taste.
It is my understanding, when talking about "ELO-based methods", that they all use wins and losses to attempt to measure the probability of any one player beating another player. If that's true, the underlying algorithm for computing the ratings should be similar. You have variables that can differ such as the k-factor, and variations of the system that add variables to attempt to make it more accurate, such as the glicko system. However, the same basic computation is involved in getting the initial ratings.
People still seem to be confused about what ELO ratings are in general. It isn't the subtle differences of your particular system that are causing the confusion. This is why there's no way anyone is going to understand what you're talking about when you try to explain your system, because they first need to understand ELO ratings.
People still seem to be confused about what ELO ratings are in general. It isn't the subtle differences of your particular system that are causing the confusion. This is why there's no way anyone is going to understand what you're talking about when you try to explain your system, because they first need to understand ELO ratings.
The "K factor" and the Glicko modifications are the same issue. Basically the current ratings create an expectation and then if the results differ from that expectation, one player exceeded the expectation and the other fell short. All agree the first guy's rating should go up and the second guy's rating should go down. But by how much? That's the role of the "k factor." FIDE uses a very unsophisticated choice for k. The're just guessing. Glicko does better, recognizing your rating shouldn't move as much if your opponent's rating is not well know or if your rating is well known.
In deriving the equations for the maximum liklihood approach, we derived as a byproduct a theoretical "k factor" that naturally takes these things into account. So we don't have to impose the ratings deviation stuff. It is in there.
Regardless, all those--FIDE, Glick, Glicko-2... are sequential approaches to updating the ratings. We used to do that. Now we have taken a big leap to the global optimization...
Fair enough....
Utterly untrue. No system, and no statistical model can possibly have ALL data. Nor do they need to in order to produce accurate results. I suggest you do some research on sampling in statistics.
Sigh, ignorance is bliss.
The more limited the data there is for a player, the less accurate the rating will be. This is true of any system, and this is why there is a metric displayed beside the rating that indicates how limited the data is.
The more matches you play, the more accurate your rating becomes, and the less significant any one result is. So the flukes become insignificant, however they are still reflected in the rating as they should be.
This system doesn't measure skill, or who is necessarily better. It includes the element of luck that affects the result of matches. It attempts to answer the question, if player A plays player B, what's the probability of either player winning? If player A is a world class player much better than player B, what's the probability that player B catches a gear and wins 11-0? That's why those "fluke" matches should be included, because we are interested in calculating the probability of those fluke matches happening.
Do some research before assuming you have it all figured out.
EDIT: And yes, the system can be improved upon, but that's a delicate process. You have to be careful not to introduce any subjective factors. For example, if you wanted to weight the matches differently for bar box vs. 9-foot, coming up with how to weight each isn't easy to do, since it's entirely subjective as to what the difference in significance should be between bar boxes and 9 foots.
Last edited:
Can someone tell me how to use this site from my iPhone? I looked in the FAQ under "What options are available in the FargoRate mobile application" It says to use Find a Player, however I am unable to see that anywhere. By using the menu in the top right, I can see Home | Videos | Top Players | FAQ | Contact Us Nowhere can I see, "Find a Player". I also thought maybe there was a separate app for this, but I cannot find it in the Apple App Store under FargoRate or 'Fargo Rate'.
The APP is close, but not quite available yet. You can go to fairmatch.fargorate.com and play around with sort of a beta version of it...
Thanks Mike, it appears that requires you to register/login. However the "Join Now" button doesn't function. I'll try again some other day.
Appreciate the response.
Last edited:
I looked at your site at work where they frown on me watching videos.
Your site is pretty but simply not easy on the eyes to view info. IMO
And it should be able to click on a player and see the records so the result of how the FargoRate was put together can be seen.
People are worried about players sandbagging to keep their rating down. Here is my thought on it -- there will be a pro event sometime, somewhere and it will be seeded based on FargoRate. Johnny Archer will get a bye in the first round cause of his high rating based on data from 1998 and a player like Oscar Dominguez that plays every weekend against top flight competition will be playing a match in the first round.
So in 2015 Archer went to a whopping 7 events (US Open 8 ball and 10 ball were essentially the same event). Compare that to other players you have ranked high on your rating such as SVB, Bergman, DeChaine, Woodward, Deuel to name a few. Yes I don't think a pool player leaving his house 7 times is active. In 2014, 9 events -- slightly more but still not active. If, to you, 7 events is active then what is inactive - 1? 0? :thumbup:
It's based on games... not events, or even matches. How many games do you suspect he played in 7 events?
Far less than someone that played in 27, 37 or 47 events.