Warrior General Games

I guess I'm just trying to figure out for in detail about how ladder rank is determined. I have heard it is like the ELO that college football BCS systems uses.

I'm just wondering why the first team below is ranked higher than the second team. Note that I am not saying we (second team) are better, just trying to figure out why one is higher when the records are similar, except the second teams losses and SOS should be much tougher than the first team. The second team has lost to two different teams this season, 4 times to #1 and 1 time to #4... the other teams losses were to people all ranked further down. Is recent record (last ten games or something?) part of the ranking criteria?

http://glb2.warriorgeneral.com/game/team/327

http://glb2.warriorgeneral.com/game/team/373

I don't know either - sometimes I wonder about strength of schedule in league games and how that affects ladder.

My rookie league (Adams) is tough (don't listen to those other guys who say their league is tough) - but winning those league games hasn't had a lot of impact over the season.

I get that someone in a crappy league would get doubly screwed if strength of league schedule played a big role -- but I don't know how it works.


Cliff notes: I wish I knew

The main thing to look at is since April 13th they have rattled off 11 straight wins while you have dropped 5 of 12.

Originally posted by bhall43
The main thing to look at is since April 13th they have rattled off 11 straight wins while you have dropped 5 of 12.

I'm well aware of our recent records as I said it above... like I said those 5 losses were to #1 4x and #4 1x... while theres were to worse teams, but longer ago

So you're saying recent records mathematically affect rankings, or are you just here to troll?

I'm saying that it really doesn't matter that they lost to some worse teams earlier in the season when they have rattled off a long string of wins and worked their way up the ladder and passed you. If the rankings functioned like you suggest it surely wouldn't make any sense to pit you and salty dogs over and over week after week driving up your SOS making it impossible to pass.

Last seasons sophmore level showed just how close teams are and how easy it is to bump a top ranked team down a peg when dd, ronin, sf, qcbc, and Maine started playing musical top 5 chairs.

College football doesn't use ELO. Even the computer rankings all have their own systems, although some start with ELO as their foundation.

Basically, think of ELO like a bag full of marbles. Everybody starts with the same number of marbles. When someone beats another person, they get to take some of that person's marbles. If you have a lot of marbles, and you beat someone with less marbles, you don't get to take as many of theirs. If you have fewer marbles than your opponent, you get to take extra. Over time, the goal is to get to where people at the same number of marbles have a 50-50 chance of beating each other.

ELO's sensitivity is determined by the K value in the formula. The bigger that number, the more marbles get taken after each game. The trick is to find a K value that is large enough to get people to the right spot, but not so large that it swings wildly past the 'correct' rating after each win or loss. Not sure if GLB2 uses it, but one trick a lot of associations use is to rank players using a known too high K value that rapidly gets that newbie into the right area on the rankings and then reduce the K value after a certain number of matches so that you reduce those massive fluctuations in ranking.

Originally posted by TDiddy8701

So you're saying recent records mathematically affect rankings, or are you just here to troll?

Recent results are definitely taken into account in ELO ratings. I don't know the math exactly, but that is how it works.

Originally posted by Xavori
College football doesn't use ELO. Even the computer rankings all have their own systems, although some start with ELO as their foundation.

Basically, think of ELO like a bag full of marbles. Everybody starts with the same number of marbles. When someone beats another person, they get to take some of that person's marbles. If you have a lot of marbles, and you beat someone with less marbles, you don't get to take as many of theirs. If you have fewer marbles than your opponent, you get to take extra. Over time, the goal is to get to where people at the same number of marbles have a 50-50 chance of beating each other.

ELO's sensitivity is determined by the K value in the formula. The bigger that number, the more marbles get taken after each game. The trick is to find a K value that is large enough to get people to the right spot, but not so large that it swings wildly past the 'correct' rating after each win or loss. Not sure if GLB2 uses it, but one trick a lot of associations use is to rank players using a known too high K value that rapidly gets that newbie into the right area on the rankings and then reduce the K value after a certain number of matches so that you reduce those massive fluctuations in ranking.

This is the right answer. As for K-value, I'm fairly certain it's constant, they're not doing anything fancy treating ratings as "provisional" or anything. And you don't really need to since you control exactly how many games each team plays - those K value tricks are usually something you do to account for the variance in accuracy from people playing wildly different amounts of games.

College football doesn't use ELO in any formula. The most common place you'll see it is in Chess ratings.

So how does your "recent results" affect your rating? Very simply - your rating was most recently adjusted based on your last game. Each game your rating goes up or down based on that game and your opponent's rating. Nothing more happens - no one goes back and says "Oh your week 1 opponent was really good you should get more credit" or "Oh man week 2 that team was basically CPU you shouldn't get much for that". It's forward moving only.

Originally posted by NiborRis
This is the right answer. As for K-value, I'm fairly certain it's constant, they're not doing anything fancy treating ratings as "provisional" or anything. And you don't really need to since you control exactly how many games each team plays - those K value tricks are usually something you do to account for the variance in accuracy from people playing wildly different amounts of games.

College football doesn't use ELO in any formula. The most common place you'll see it is in Chess ratings.

So how does your "recent results" affect your rating? Very simply - your rating was most recently adjusted based on your last game. Each game your rating goes up or down based on that game and your opponent's rating. Nothing more happens - no one goes back and says "Oh your week 1 opponent was really good you should get more credit" or "Oh man week 2 that team was basically CPU you shouldn't get much for that". It's forward moving only.

Which is exactly why your late skid and his late rise puts him ahead of you.

You had the points you had when this skid started, and he had the points he had when his streak began. Nothing was going to impact those or take those away. As stated, forward moving only. So you missed the opportunity to 'steal marbles', especially 'extra marbles' in the case of the #1 Team you faced 4 Times. He took marbles, even if he took 'less' marbles... and "maff" tells me that Less > None.

Likely you will beat him head to head, and that will help... you'll even get 'extra marbles' since he is ahead of you... and such a game will help 'correct the market'.

I just went on an ELO skid.

http://www.youtube.com/watch?v=98P-gu_vMRc

Originally posted by Robert Fripp
I just went on an ELO skid.

http://www.youtube.com/watch?v=98P-gu_vMRc


I just watched the whole 4 min vid.... Why did I do that? I couldn't look away!