Fargo theoretical rating with two players

[easy-e]

Well why would you want that?

If you had no FARGO RATE what should Bev belles you AVATAR work LOST? Because you caN FIND FARGO, ND?

You have a way with words, my friend.

 

[markjames]

remember what john steinbeck said about fargo, ND

 

[ribdoner]

Why can't we have your Fargo rating posted below your Avitar?

if we also used our name some truth would bubble up esp in view of the handicapped events which have massive payouts

 

[mikepage]

[...]

If there are only 2 players in a bubble, we assign them a 350 rating to start with. After 200 games Player 1 shows to be winning games at a 2-1 ratio, a 100 point difference. Would Player 1 end up a 450 and Player 2 stays a 350, would player 2 go down to a 250 with 1 staying at 350 or would one go to 300 and the other to 400?

It is none of these. There really is no starting rating or starting point in FargoRate. As you suggest, what the data tells us is only that the players seem near 100 points apart. But we cant tell if they're pros or beginners or somewhere in between. There is, of course, an answer to the question. But I would not have known what that answer is without doing it. I would have expected the players would end up 50 points below and 50 points above some averagish/middleish rating. So I created two fake players with 200 to 100 score and the result is 424 and 524. These ratings are not meaningful other than being 100 points apart.

 

[mikepage]

I think so also, but I would love to see Mike plug in the two players in and see how the ratings progress. Set 1 , 10-5, what is the Fargo after that? Set 2, 10-5, what is the Fargo after that, and so on untill we reach the 200-100 game result

I am also wondering if we plug in a single set of 200 games, with a 200-100 result, would the final rating be different than if we put in 20 sets of 10-5 results.
[...]

I put in 200 to 100 and got ratings 524 and 424.

I could have instead put in a 10-5 match every day for 20 days. It would have went to 524 and 424 on the first day and then not changed after that.

 

[Icon of Sin]

So did you ever live in Arizona under the Arizona System?

I have, I did, I had a AZ Rating, did you?

What's your Fargo?

 

[Icon of Sin]

Well how can you fairly handicap a person on their TRUE SKILL LEVEL sayin a 9 Ball Tournament, on a Valley Bar box. If their only experienced is playing of 4.5 x 9 Gold Crowns 8 ball. IMHO this would be like a guessing game at best.

IMHO All Handicap System are BULLS***, and the only true was to judge person Talent or lack of is OPEN PLAY. Forces people to put out effort or be a looser. If Fargo is so GREAT when don’t DCC, use it to provide an edge to lessor players?

Handicapping System were conceived by Pool Bar, Sports Bar & Room owners to identify their Recreation Player, who spend money, and keep better player from coming in and win TOURNAMENTS. Chasing the Money spending recreation player AWAY, who pay bills

Serious Player most do not pay the bills, they consume less alcohol, then the BANGERS DO. They are the people who are the upper 25% of Pool Skill or more.

AGAIN JMHO

 

[mikepage]

If they never play against anyone in the system to connect them with the main set of players, they have no real rating. It may be that they are each somehow weakly and fictitiously connected to the average Fargo rating, but that would be a completely arbitrary reference. (Example: give them a 1-1 record against an average player but with a weight of one thousandth of a game.)
[...]

This is an interesting point. With Maximum likelihood--what we do--the system is just as happy with those two players being at 1200 and 1100 as it is with those players being at 500 and 400. As a result we can get some goofiness.

Suppose for instance we have this isolated pair of players with the 200 to 100 score, and all we really know is they're around 100 points apart. Now this part isn't going to happen, but imagine the weaker player plays 300 games against Gorst and the score is 150 to 150. We'd be pretty much forced to conclude the stronger player is around 950.

Here. though, is a related thing that CAN happen. The weaker player plays not 300 games but 2 games against Gorst and wins one of them. Maybe Gorst breaks dry and leaves a 1-9 combination. The system is going to do the same thing--put the two players at 850 and 950, even though we know that's an absurd result.

Bob's suggestion sounds reasonable--have every player connected by a weak rubber band to the middle of the distribution. Once they have a lot of games that rubber band is overwhelmed by stronger forces and no longer matters. But for players with weak connection to the group or nearly isolated groups there is a little pull toward reasonable. In other words our prior knowledge is that the players are human and absent other data, they're far more likely to have ratings near 400 than near 1000.

FargoRate with rubber bands is replacing Maximum Likelihood Estimation (MLE) with Maximum A Posteriori (MAP) Estimator. It sounds like a good idea, but in practice it's not for us. MLE, warts and all, is better.

 

[hang-the-9]

It is none of these. There really is no starting rating or starting point in FargoRate. As you suggest, what the data tells us is only that the players seem near 100 points apart. But we cant tell if they're pros or beginners or somewhere in between. There is, of course, an answer to the question. But I would not have known what that answer is without doing it. I would have expected the players would end up 50 points below and 50 points above some averagish/middleish rating. So I created two fake players with 200 to 100 score and the result is 424 and 524. These ratings are not meaningful other than being 100 points apart.

Thanks! I know the ratings were not real, but I was curious how the math would work with players when the computer only has results to work from without being able to compare skills. Logically it would be the safest assumption to make to give both players some credit vs assuming one was better or worse by a large margin.