That is why "known ability" should be taken into consideration by a committee that has no financial stake in the matter.
Interesting FB post about FargoRate
- Thread starter u12armresl
- Start date
Players Fargo Ratings--the ones based on all their data--are a notably better measure of "true ability" or "true current skill" than are any of the following
--last 6 months
--last two years
--last 500 games
--last year
--last 1000 games
--and on and on
The uncertainly in all of these is larger than the actual deviation between the elusive "true current skill" and the players Fargo Rating. This is not to say players can't have current skill that outruns their rating or is lower than their rating. It is that we, effectively, can't measure it. What we're interested in is climate, and what we have to learn about climate is weather.
The average daily high in Denver over the last few decades might be 88 degrees F. We might be interested in whether--or at what rate--Denver is warming. The average daily high THIS year could be 91 degrees or it could be 85 degrees. Either way July 2022 data gives us no real clue about changes in climate. [expected average temperature might go up by less than a tenth of a degree in a year].
Thanks for the reply.
You do understand that the Weather man
who predicts 88 degree days everyday ,because it is the average , won't be around for long.
I do see that a fr which is somewhat slow to move with the current trend is less prone to gaming the system.
Woodshaft
Do what works for YOU!
Pool players can get better pretty quickly, even long-time players. I know players that have changed their stroke, or suddenly became mentally strong and play at prolly 40 to 60 fargo points higher than they were a few months back. And this ability is permanent. But their fargorate lags behind. For many months-- it just slowly creeps up.
I feel that, for most people, you never GET WORSE in pool. Pool is like riding a bicycle. Once you learn how, the only thing that can happen is that you will become a better rider, you're not going to get worse.
For this reason, old pool data is basically irrelevant. I don't care how good you were 2 years ago-- how good are you playing lately?
APA's Equalizer handicapping system, for example only analyzes data from the last TWENTY games to make adjustments because they know SL's can suddenly get and stay better. APA ratings tell everyone how that player is playing NOW, not 2 years ago.
Mike, you yourself state that a person isn't established on Fargorate until they have 200 games in the system.
If that is true, let me see the rating for their last 200 games. BY YOUR OWN STANDARDS, games over 200 games old aren't necessary.....
Apparently you think that pool players can't improve quickly and permanently, and so you keep using very old data for current fargorates.
Fine.
But at least let the folks who know better have access to recent data to get a more accurate picture of a player's new and improved level of play. Give us an app feature that allows us to request a player's fargorating over say a RANGE of games-- the last 200, or 500 for example.
Most big fargorate-handicapped (or capped tourneys) require at least 500 fargorated games/player. So let the tournament director access those previous 500 games' fargorate level of play in the app, compare it to the player's overall rating, and make an educated decision.
Considering Fargorate's large overall standard deviation, it's the least you can do.
Thank you!
My understanding is that FargoRate incorporates some aspect of the Exponentially Weighted Moving Average method. The 'k' value chosen in the formula affects how much the final FargoRate number is affected by recent results vs older results. FargoRate would consider their 'k' value to be proprietary. Who knows - maybe FargoRate has different 'k' values depending on their robustness number; or how many games are being inputted during various time intervals??
I would bet that FargoRate has experimented with different values for 'k' to change its sensitivity of recent results to the past results. Maybe they will show us some of these studies some day...
en.wikipedia.org
I would bet that FargoRate has experimented with different values for 'k' to change its sensitivity of recent results to the past results. Maybe they will show us some of these studies some day...
Moving average - Wikipedia
Last edited:
Woodshaft
Do what works for YOU!
All I remember from college is F=ma, and E=mc^2 lol.My understanding is that FargoRate incorporates some aspect of the Exponentially Weighted Moving Average method. The 'k' value chosen in the formula affects how much the final FargoRate number is affected by recent results vs older results. FargoRate would consider their 'k' value to be proprietary.
I would bet that FargoRate has experimented with different values for 'k' to change its sensitivity of recent results to the past results. Maybe they will show us some of these studies some day...
Moving average - Wikipedia
But seriously, let us be able to see the fargorate for a player's previous 200 to 500 games or so, is that too much to ask?
It's no secret that Fargo Ratings has no bigger fan than me, but I don't agree with using old matches in computing it.
Take a look at seeding in both pool and tennis. In both sports, the event producers want the best players matching up in the late rounds and they achieve this through seeding, and this, too, is what the fans want.
In tennis, the US Open begins today. Serena Williams, who might even be #1 if ratings were based on five-year performance, is not even one of the thirty-two seeds in the women's draw, and she shouldn't be. Yes, in the last five years, she has appeared in five grand slam finals and won one, but she hasn't reached a grand slam final for three years and hasn't shown much form in those three years. Serena is +5000 with the bookies, meaning she's a 50:1 shot to win. Nobody anywhere thinks that her long-term brilliance and fairly strong track record in the last five years make her anything less than a huge longshot to win the US Open.
At pool, Matchroom is, similarly, smart enough to understand that nine-ball tournaments must be seeded based on recent play, as measured by the Matchroom Nine-ball Rankings, not on long-term performance. That's their best chance of delivering elite matchups in the late rounds, and that's what both they and their fans want.
For pros, I think a Fargo Rating that ignored any games played more than two years ago would be more valuable than that presently computed. Similarly, it is high time that players who are, more or less, inactive, be removed from the Fargo listings. Mike Dechaine, admittedly a super-strong player once upon a time, is still rated one point higher than Skyler Woodward. Five years ago on the forum, most of us did consider them to be, more or less, equals. Now, Mike nearly never competes, has no high finishes in any high-profile event in recent years, and Skyler has played great pool over the last few years, including being Team USA's best player at the Mosconi year after year. Fargo says they are still equals, but they aren't, and Skyler would have much stronger prospects in an elite event right now than Mike.
It pains me to take any shots at Fargo, which has added something very special and important to our game, but, as noted, I am not at all on board with the inclusion of old results in one's rating and think it, too often, obscures our view of who the best players are. In my view, inclusion of old results DOES NOT give a clearer picture of current skill.
Take a look at seeding in both pool and tennis. In both sports, the event producers want the best players matching up in the late rounds and they achieve this through seeding, and this, too, is what the fans want.
In tennis, the US Open begins today. Serena Williams, who might even be #1 if ratings were based on five-year performance, is not even one of the thirty-two seeds in the women's draw, and she shouldn't be. Yes, in the last five years, she has appeared in five grand slam finals and won one, but she hasn't reached a grand slam final for three years and hasn't shown much form in those three years. Serena is +5000 with the bookies, meaning she's a 50:1 shot to win. Nobody anywhere thinks that her long-term brilliance and fairly strong track record in the last five years make her anything less than a huge longshot to win the US Open.
At pool, Matchroom is, similarly, smart enough to understand that nine-ball tournaments must be seeded based on recent play, as measured by the Matchroom Nine-ball Rankings, not on long-term performance. That's their best chance of delivering elite matchups in the late rounds, and that's what both they and their fans want.
For pros, I think a Fargo Rating that ignored any games played more than two years ago would be more valuable than that presently computed. Similarly, it is high time that players who are, more or less, inactive, be removed from the Fargo listings. Mike Dechaine, admittedly a super-strong player once upon a time, is still rated one point higher than Skyler Woodward. Five years ago on the forum, most of us did consider them to be, more or less, equals. Now, Mike nearly never competes, has no high finishes in any high-profile event in recent years, and Skyler has played great pool over the last few years, including being Team USA's best player at the Mosconi year after year. Fargo says they are still equals, but they aren't, and Skyler would have much stronger prospects in an elite event right now than Mike.
It pains me to take any shots at Fargo, which has added something very special and important to our game, but, as noted, I am not at all on board with the inclusion of old results in one's rating and think it, too often, obscures our view of who the best players are. In my view, inclusion of old results DOES NOT give a clearer picture of current skill.
You should try asking nicely. If I were Mike, I wouldn't bother responding to your post.
I am always impressed with how even keel Mike is, despite all the stupidity about FargoRate
Fargorate reduces the weight of old matches with a half-life of three years. That seems like a reasonable time to me. I suppose it would be possible to check the data for "rapid improvers" and "rapid deprovers" to see if some other decay rate is better. I think it would be hard to have an individual time constant for each player.
I suppose that if there is a player who has special circumstances as far as their rating trajectory it would be a good opportunity for a careful observer to get in a smart bet.
(A "half-life" of three years means that matches from 6 years ago are counted with 1/4 the weight of current matches and 9-year-old matches are counted with 1/8 the weight, and so on.)
Fargo Ratings based on 200 games pretty much suck. The competition they win is against FargoRate being unwilling to make a statement about how a person plays.
My understanding is that FargoRate incorporates some aspect of the Exponentially Weighted Moving Average method. The 'k' value chosen in the formula affects how much the final FargoRate number is affected by recent results vs older results. FargoRate would consider their 'k' value to be proprietary. Who knows - maybe FargoRate has different 'k' values depending on their robustness number; or how many games are being inputted during various time intervals??
I would bet that FargoRate has experimented with different values for 'k' to change its sensitivity of recent results to the past results. Maybe they will show us some of these studies some day...
Moving average - Wikipedia
This has the flavor of being right. But the notion of a "k" value is associated with Elo methods. FargoRate has no k value.
Here is a way to think about it. Consider two players, Click and Clack, who we've never seen play before. We want to know who is the better player. They today play a race to 10 that Click wins 10 to 7. At this point if we had to guess, we'd say Click is a better player. But it's a weak conclusion.
Now suppose we learn that in addition to today's match Click and Clack also played a race to 10 yesterday with the opposite result: Click LOST yesterday's match 7-to-10. Taking the two days together, the score is 17 to 17, and we'd likely conclude they play the same.
Now, suppose instead of one match yesterday, they played two, and Click lost both yesterday's matches 7 to 10.
At this point we have
Click 7 Clack 10 -yesterday
Click 7 Clack 10 -yesterday
Click 10 Clack 7 -today
The overall score is Click 24 Clack 27 and though it again be a weak conclusion, we'd give the edge to Clack, right?
If the first two matches were last week or last month, it probably wouldn't affect our conclusion. But what if the first two matches were a year ago or five years ago or a decade ago. The more distant, the less sway we might give the older matches. There must be a length of time--call it halflife--for which we'd see Click and Clack as equal players after the three matches.
Click 7 Clack 10 -halflife ago. (treated as 3.5 to 5 in "today" games)
Click 7 Clack 10 -halflife ago. (treated as 3.5 to 5 in "today" games)
Click 10 Clack 7 -today. (treated as 10 to 7 in "today" games)
The overall score in "today" games is 17 to 17, and our best guess is they play the same.
If the first two matches were 2 halflifes ago, they'd be cut in half again in influence and would not be enough to alter the conclusion from just today's match that Click is a better player.
We don't have to guess a halflife. We can use our results to see which halflife does the best job for predicting TODAY skill.
I think you're not understanding the role of older data Stu. It is not to give credit for being good for a long time or for having been good years ago. The focus is on using all data available SOLEY to best predict CURRENT skill. In other words, there is a difference between incorporating data over five years or ten years and looking at "five-year performance" or "ten-year performance." That's not what we do.
[...]
You're speculating here. Don't you think we--FargoRate--actually KNOW, without speculation, whether Fargo Ratings ignoring games more than two years old leads to better ratings?
We always welcome your thoughts Stu.
The role of older data is subtle. Here is a thought experiment for you.
First imagine we take a group of players who are playing a couple thousand games a year. The first thing we'll do is compute a RECENT RATING-400 and that will be a performance rating based on just the most recent 400 games. Then we'll compute an OLD RATING-400. That will be a performance rating based ONLY on games played 3 years ago.
How will be tell which ratings are better? We will get our crystal ball (yes, we do have one) and look into the future and see how these players actually perform over the next 8 months. We can use any ratings known now to predict those future results. The best ratings are the ones that predict future results the best.
As you might guess, RECENT RATING-400 does better than OLD RATING-400 at predicting future results.
Here is the big question. Which of the following do better at predicting future results?
(A) RECENT RATING-400
(B) OLD RATING-2000
Here (B) is based on 2,000 three-year-old games. Because these people play about 2,000 games per year, they're really games played between about two and a half and three and a half years ago and average three years old.
The answer is (B). Think about this one. (B) isn't a rating that merely weighs in or considers old data. It is made ONLY from old data. If you were an oddsmaker and had to use one list or the other to compute odds for an upcoming tournament, you are better off ignoring (A) and choosing (B). How could this possibly be?
How do we get a better prediction than either (A) or (B)? Consider (C), which is the average of (A) and (B). That does better. How do we get better still? Consider (D), which is a weighted average of (A) and (B) with the weighting optimized for predicting future results.
Last edited:
Asking whether ratings can be improved with old data is like asking whether your car can be cleaned with a bucket of dirty water.
The answer, for the car, is it depends both on how dirty the water is and on how dirty the car currently is.
You, and others, intuit both the car to be cleaner than it currently is and the water to be dirtier than it is.
Thanks for this, Mike, but we're not together here. Can old results tell you something about today's likely performance? Yes, at times they can, but I feel they don't tell you much. and that the Fargo calculation over-weights old results. Fargo also tends to forgive relative non-participation more than it should for my taste, for likely current performance is very much a function of how much and in what events one has been competing of late.
All that said, Fargo is the best measure we have in pro pool of likely performance at a moment in time, in part because the only credible rankings system right now, the Matchroom Nine Ball Rankings, consists of too little data. As I've noted many times on the forum, I have no opinion on how accurately Fargo evaluates the play of non-professionals.
You'll always have an advocate for Fargo Ratings in me, but I just don't have a complete comfort level. It's more than likely because there's something called the eye test, and, outside of the Asians still grounded by COVID, I have watched live the play of about least 90% of the Fargo Top 100 in the last twelve months and feel my grasp of who's likely to succeed in any coming event factors in certain things that are beyond Fargo's reach.
Mike, you 're a credit to our sport. Keep up all the good work, and don't think that, because I have a couple of small issues with the calculation, I am any less of an advocate for Fargo than you. In the general case, Fargo Ratings are very dependable and if I'm not inclined to give them a grade of "A+" they certainly get an "A" from this usually tough grader.
All the best, Stu
Last edited:
stewie
Active member
Maybe you are all overthinking Fargo? There will always be a difference between "being rated at" and "currently performing at". If you are a 700 in Fargo, that means you are capable of playing at that level. But it doesn't mean you should win every time you play a 600.
Going back to Mike Dechaine: his Fargo should tell you what he should be capable of, but his current results tell you he is not delivering, right now. The Fargo is not "wrong". On the pro level, that's why there are Matchroom Rankings for seeding and inviting to Matchroom events. They show better "current" performance, and should be used for that. I think World Snooker even freezes last season's rankings for seeding and has a second "current" ranking on top of it. A player might be #1 in one, but #4 in the other. All those pieces of info, give you the "complete" picture.
Now, on a more amateur level, it's hard to gauge your current performance or capabilities, if most (if not all) of the matches you play are in handicapped leagues or tournaments. I am sure competing handicapped is like putting hand brakes on the "better" players and she/he would get a better, more correct, rating, if performed more in even races. Or, at least, games-on-the-wire only and not game altering handicaps like money balls or getting points per ball in 9-Ball.
Going back to Mike Dechaine: his Fargo should tell you what he should be capable of, but his current results tell you he is not delivering, right now. The Fargo is not "wrong". On the pro level, that's why there are Matchroom Rankings for seeding and inviting to Matchroom events. They show better "current" performance, and should be used for that. I think World Snooker even freezes last season's rankings for seeding and has a second "current" ranking on top of it. A player might be #1 in one, but #4 in the other. All those pieces of info, give you the "complete" picture.
Now, on a more amateur level, it's hard to gauge your current performance or capabilities, if most (if not all) of the matches you play are in handicapped leagues or tournaments. I am sure competing handicapped is like putting hand brakes on the "better" players and she/he would get a better, more correct, rating, if performed more in even races. Or, at least, games-on-the-wire only and not game altering handicaps like money balls or getting points per ball in 9-Ball.
I don't know whether you have used this before, Mike, but I really like it.
No, it's not about "capable of."
Fargo may be based on past performance but is very frequently used to measure expected current performance. That's why we regularly see it used to project the score of an upcoming match and to project win probabilities of the contestants. It is also used on live stream to reproject the likelihood of victory for each contestant as a match progresses. When Fargo is used in this way, which is very often, it is being used as if it were a measure of "currently performing at" or, if you prefer "expected current performance." If it represented anything but that, such odds would be incalculable using Fargo.
What’s your sign? What’s your blood type? What’s your shoe size?
Fargo Rating is a joke…match up…bet high…play pool…