League Ratings

[Munkey]

I am trying to work out a good way to rank the players in our league and need a formula to establish rankings based on a few criteria:

1. We do not play set fixtures and some teams play more than others so the league rating needs to reward players for the win ratio rather than a simple number of wins.

2. Some of our players may play repeatedly against each other, is there a good formula that will stop these games biasing the games played against other teams. (for example a high rating because of 10-0 wins against one player whilst losing 0-1 against all other players is not a fair ranking IMO)

3. Going against the point above I would like the impact of fluke games to be minimsed, i.e. consistency should be rewarded above fluke wins.

4. We use an Excel spreadsheet to track the league so it is not a problem if a suggested solution is slightly complex.

I realise that some of these goals may conflict with each other but would be interested to hear how other leagues are ranked and what suggestions people have.

[gallowin]

Just make each coach worth X points and you can only get those points once. Maybe make it best of the first 2 games played against this coach.

People can still play someone 10 times and get gobs of SPP building up but they'll only have 1 set of league points and be way behind the other coaches.

Our current LP system is W/L/draw plus TD & Casualty differential. Win = 4, Loss = 1, Draw = 2, TD = 0.5 & Cas = 0.25

Good luck!!

[Anthony_TBBF]

We have been testing out ranking by an adjusted win percentage. Here's the formula:

Adjusted Win% = Win% * (Total TR Against/Total TR For)

You can see it in action at http://www.tbbf.org/standings.asp

Basically it penalizes teams who constantly play lower rated TRs and rewards those who play and beat higher rated teams. The adjustment isn't all that much but it's a nice way to shake things up a bit.

[grimfang]

here is a system that i have used in another type of game.. it worked well.

1) No team may be ranked until they have played 4 games.

2)The basic point system is: win=2 pts, tie=1 pt , loss = -1pt

3) once a team has completed 4 games and is seeded in the rankings, the following point modifiers are used after each game.
If your team wins: Add and additional +1 point for every 3 places above you the oppenant was ranked. (example: if #10 beats #3, then he would recieve 2 points for the win, plus 2 points for beating a team ranked 7 places above him... for a total of 4 points.)
If your team loses: subtract an additional 1 point from your total for every 2 places below you your opponent was ranked. (example: if #3 loses to #10, then he would lose 1 point for the loss, and an additional 3 points for being defeated by a team 7 places below in the rankings.)

4) after EACH game, adjust the standings immediatley!

This system seems odd.. but trust me, it works great! It does not punish a team for losing to a higher ranked team, but rewards them for pulling of the upset. At the same time, it forces the top teams to challenge other top contenders, rather than beat up on the weaker teams all the time. (they will not want to risk losing to a team that is ranked too far below them.. it will hit them hard in the standings!)

When you use this system, you will see fun things start to happen. A weak halfling team can pull of a miracle victory and climb VERY high in the standings... but they will not stay there long if it was a fluke. And great teams will lose what should have been an easy game... they will fall to the bottom, and will climb up again quickly! This system will result in an accurate assessment of a team's skill.

[christer]

Check out the page describing the FUMBBL rankings. This ranking system is very close to what NAF will be using which in turn is an ELO-type of algorithm. Works fairly well if you ask me.

-- Christer

[Mojoshenpo]

... use a ladder! You can only move up the ladder by challenging the person above you. This ignores the padding of TR by playing teams lower in the ladder.

[Munkey]

Thanks for the sugestions, I have only just started to look at the league table again after being off for xmas and have just dug up this thread. Certainly provides some food for thought.

Our league is currently holding a cup so there is time for us to come to a consensus about the best way to rank season 2 before it starts.

[Milo]

We have been testing out ranking by an adjusted win percentage. Here's the formula:

Adjusted Win% = Win% * (Total TR Against/Total TR For)

Anthony -- is this based on TR at the time of the game, or current TR? Does it double-count the TR of teams that have been played against twice?