You totally missed my point.
APA/Fargo cut off?
- Thread starter benjaminwah
- Start date
Yes, except that does not happen, and all you need is a handful of differently ranked players to even things out. It is very very unlikely that a pool hall happens to have everyone in it the same level. There will be a guy there that is a 600, a guy there that is a 500, 4 people that are 450, another few that are 480, maybe a couple of 700s if it's a good area, etc... They will play each other and establish the ranking there. The ONLY way this will not work is if the sample is like 2 or 3 people very close to rank and they only play each other or one is much stronger than the other which are the same level; then there may be some odd ratings if say the best player is a 500 but the other two are like 200 level and the guy keeps beating them 6-1 or 6-2. Then his starting rating may shoot up to like 700 to their 500 if no one ever visits them and points out their actual skill levels and that gets adjusted. It's like finding a hidden race of humans that never developed technology, at first they won't know what a car is, but in a few years they will be complaining about gas prices like the rest of us.
Keep in mind this is not a rating that is now just floating around with random stats, it's established over thousands of games and players what people of a certain Fargo rating can do at the table, and we can now easily sort them properly with a decent amount of accuracy, say within 50 points, and have the system go from there over time and results. This is a bit like the argument flat earthers use, they take everything we learned so far as humans and just say "well, I am going to start over cause I was taught to question things and just go by what I see with my own eyes". We don't need to start at 0 again in anything. Even if this unknown group of players that never played anyone else starts at same starting Fargo rating, say 400, eventually the better players will go up, the worse players will go down and those at that 400 level will stay there.
Players A, B and C are all 400 Fargo starting level just as a random assignment. C starts playing A and losing like 5-1, 5-2, A's Fargo rating starts to go up based on that. B plays C, C loses 5-3, 5-4 to him often, so B goes up a bit over C at 400 but since he is not as good as A his Fargo does not go up as high since his matches are closer in scores to C. B plays A and loses 5-3, 5-4 normally, Fargo sees that and adjusts them again a bit. So it goes for a bit, and these 3 players, A B C level, only playing each other, will end up in about their correct Fargo scores.
Last edited:
Yes, two players with any one of them having a solid Fargo rating is all that is needed. If it's established over 10 years that I often beat players that are 500 but loose to players that are 600 with my 2,000 matches (making me a 550), then I play a single person, they beat me by say 5-3, Fargo can say they are 600 because that is the expected and established outcome for those skills. Now every single person that other person plays can be compared to everyone I played based on how I did against that one person and how that outcome compared to the others I had. Of-course, this all needs to be based on some sort of reasonable expectation that a single match between two players is not enough to establish a rating, that is why Fargo is not set till someone reaches 200 games. No one is going to say that just because I played one player one set that the skill rating between us will be correct, but it's better than nothing since it's at least a starting point. If I stay in this same place, and only play this one player say 20 sets, that is about all that would be needed to get a good rating for him and anyone else he may play. If we play 40 sets, even better.
Have you ever ran into two players in a room that only ever played each other that are good enough to be put into the Fargo system?
It's like COVID or any other disease, it just takes one infection to spread to others exponentially. If I got Fargo, and I play you and you catch Fargo, everyone you play also has Fargo now. Only way to stop Fargo is to stick us all in a single room, kill everyone that has it and leave two people locked in a room that don't have it, then they will never get it.
Last edited:
You're still not getting it.
Bill goes to visit family in Phoenix and plays two nights of league, 10 games. He is expected to win 6 of the games but actually only wins 3, the expectation for a player 180 points below Bill.
The consequence in this made-up scenario, is EVERYBODY in Nowhereville, including people who have never played Bill, go down 180 points the next day. Let this sink in.
Now think of Henry, in nowhereville. He plays league on a different night from Bill and has never played Bill. Henry just dropped 180 points even though he played nobody in Phoenix and also none of his opponents played anybody in Phoenix.
This is very very different from what would happen if we were doing an ELO scheme.
Years ago I implemented ELO ratings for our league software that I wrote from scratch.
Gave interesting results, but they were obviously only tied to your individual performance against other league players. No ability to link it to others outside the league because they never directly competed.
Side note, we use the 17 point BCA scoring. The top 5 players had higher averages, but a good/older/crafty played had the highest ELO because he was able to win lots of close games (10-7, 11-6, etc), but rarely ran out for 17s.
Totally different system compared to the Fargo algorithm. I think the Fargo system is very accurate for anyone that plays at any competitive level.
The only small issue I see with Fargo is that if you live in NoWhereVille, you don't have much opportunity to pull off an upset against a much higher (75+) rated play.... because you never play them. It seems like in areas with a number of monster players, that the average local Fargo seems skewed a little higher. But.... over time the algorithm will smooth this out.
It would be interesting to see a "heat plot" of Fargo rating on a map of North America.
This assumes that Henry has a playing history with someone that has also played Bill. ...or numerous other degrees of separation between the two. If Henry's league is in a bubble and has zero exposure to anyone Bill has a degree of separation to, then his fargo will not adjust.
Fargo does not use mailing address as a means to adjust ratings.
I don't think people realize just how weak is the requirement to get complete coupling.
Imagine a town where the good players play on a Thursday league, and there is a casual beginner league on Tuesday. Perhaps each has 50 players, and everybody thinks of them as being separate. Nobody's really noticing, but on a typical Tuesday or Thursday one of the teams will have one substitute player from the other night. After a few months, there are hundreds of games of coupling and all the ratings line up.
You are correct it is possible for a regional bias. However if that area is aware of forgorate and progressive enough to implement them it is safe to assume they won't be a island with regional bias.
Are you using Sollato? Submitting the matches to Fargo?
That probably just means the data is thin. Perhaps one or both of you don't have that many games and/or many of the opponents you play are more or less unknown to Fargorate. This is not a regional bias but rather just poor ratings in a region with not enough data coming in.
No I’m just stating how we play and that his rating is 30 points higher than mine
He has almost 700 games in the system
The sample error is "1/sqrt-n". So in this made-up example the error would be ~30%. Seems silly to do this but ok.
What is your robustness?
He has 109 games, with 76 of them more than 3-years old and a handful of unestablished opponents.
Individuals don't need that diversity. Each can play seemingly in his own cocoon. It is the group that needs that diversity. It could be a few people moved to your town from elsewhere and a few from your town moved elsewhere. You're not even thinking about them