Anyone using linear programming for SC?

[Daics35Marvel]

This one's for the nerds. Does anyone else on here use LP for Supercoach, and if so how successful has it been? I've used one for the past few years that's been successful in winning my league and gets me in the top ~5% overall. Obviously, I'm always aiming higher so am interested in sharing some info on what goes into your models.

 

[Firebat]

Wtf is linear programming?

 

[Daics35Marvel]

Wtf is linear programming?

It's a mathematical method that's useful for fantasy leagues. It finds the highest possible solution (maximum number of points) given limited resources (salary cap) and constraints (max 22 players, max 8 defenders etc.)
http://en.wikipedia.org/wiki/Linear_programming

It's a long shot that others are using it but thought it was worth asking as I'm grappling with a few formulating ideas.

 

[Ingall]

Wtf is linear programming?

Not sure, but it seems to be something to help you finish 17,000th

 

[TimC]

It's a mathematical method that's useful for fantasy leagues. It finds the highest possible solution (maximum number of points) given limited resources (salary cap) and constraints (max 22 players, max 8 defenders etc.)
http://en.wikipedia.org/wiki/Linear_programming

It's a long shot that others are using it but thought it was worth asking as I'm grappling with a few formulating ideas.

I haven't had the idea of using it in the past, but would be interested to see your previous formulas, if you don't mind sharng that is.

 

[Resch]

The trouble is it relies on past data being a reflection of future performance and also Rookies have no past data.

Also with the downgrading of price for injured players it will be biased to them based on the points to $ link

 

[Daics35Marvel]

The trouble is it relies on past data being a reflection of future performance and also Rookies have no past data.

Also with the downgrading of price for injured players it will be biased to them based on the points to $ link

Agreed, but really we're all using past data to predict the future. At the end of the day, the success/failure of all our SC teams comes down to how well we can predict the number of points that players will accumulate. LP is just a useful method for taking those predictions and quickly finding out which combination of players gives the maximum total points. If my points predictions for each player sucks, the result will suck.

 

[SJ]

I am aiming to use it this year. Have set up a table in Excel (not all players complete yet) where I will put all players in consideration, my expected average for them in 2010, and their price etc.

Except Solver does not like changing that many cells...

 

[Falchoon]

The trouble is it relies on past data being a reflection of future performance and also Rookies have no past data.

Also with the downgrading of price for injured players it will be biased to them based on the points to $ link

It's potentially useful if you massage the data.

Set up a spreadsheet with a number of headings - age, role change, team +/- etc and assign percentage changes to each player using past data.

Then apply your programming.

I've thought about it but at the end of the day it still comes down to selecting the Sam Giberts of the world.

 

[GUMBLETRON]

The question is whether those initial estimations are any better than your regular selection hunches, and whether applying it to every player in the competition is better than simply closely comparing key picks.

 

[Impromptu]

LOL watda?

 

[SJ]

I am aiming to use it this year. Have set up a table in Excel (not all players complete yet) where I will put all players in consideration, my expected average for them in 2010, and their price etc.

Except Solver does not like changing that many cells...

Does anyone know of other ways to do it in Excel that don't involve Solver?

 

[Daics35Marvel]

Does anyone know of other ways to do it in Excel that don't involve Solver?

I haven't used Excel Solver much, but from what I remember it doesn't cope with integer variables very well. Integer problems, e.g. you can't select 0.22 of player x, are much harder to solve than normal LPs. Although I haven't used it, I've heard that What'sBEST is a good solver that works in Excel. Outside of Excel, there are some great free solvers available (e.g. GUSEK), but formulating the problem becomes more complicated.

 

[4#Didak#4]

Linear programming intermingled with a variety of complex algorithms referenced across some conditional probablilities and mapped & coordinated with some other large nonsense mathematical words is generally how I get my best results.

WTF people! Keep it simple pick the best players you can under the salary cap. Too many variables to think that you can work out some formula for success.

 

[SJ]

Linear programming intermingled with a variety of complex algorithms referenced across some conditional probablilities and mapped & coordinated with some other large nonsense mathematical words is generally how I get my best results.

WTF people! Keep it simple pick the best players you can under the salary cap. Too many variables to think that you can work out some formula for success.

Out of interest, in your simple world, how do you define and quantify "best"?

 

[thomas_m18]

I know I take Supercoach seriously...

but I couldn't take it this seriously.

It's a cool concept, but it's not for me.

 

[pendlebury_10]

A very interesting read.

 

[4#Didak#4]

Out of interest, in your simple world, how do you define and quantify "best"?

That is completely subjective, everyone has different opinions on who are going to be the "best" otherwise everyone would have the same squad. My best are based purely on watching football & judging who is a good/bad player in relation to their price. Simple! If you are relying on writing computer programs to work things out for you are in serious trouble. I don't care how smart you are you cannot factor in all the variables that matches football knowledge.

 

[SJ]

That is completely subjective, everyone has different opinions on who are going to be the "best" otherwise everyone would have the same squad. My best are based purely on watching football & judging who is a good/bad player in relation to their price. Simple! If you are relying on writing computer programs to work things out for you are in serious trouble. I don't care how smart you are you cannot factor in all the variables that matches football knowledge.

There is no subjectivity when it comes to this game.

Unfortunately for you, the SuperCoach scoring system does not take into account who is a "good/bad player".

Basically the whole premise of the game is finding the way to maximise the number of points scored, with a given starting salary of $10,000,000 (and a given amount of trades).

Thus, when selecting players, one already needs to estimate the amount of points scored by a player, and find the relative cost of a player in relation to the amount of points he will score.

Basically whacking a few of these estimates into a linear programming system is no different to comparing a player's price to his 2010 estimate (which is what every good player should be doing anyway, consciously or subconsciously), but it is able to do many at once rather than one at a time.

 

[DanA]

That is completely subjective, everyone has different opinions on who are going to be the "best" otherwise everyone would have the same squad. My best are based purely on watching football & judging who is a good/bad player in relation to their price. Simple! If you are relying on writing computer programs to work things out for you are in serious trouble. I don't care how smart you are you cannot factor in all the variables that matches football knowledge.

If you watched every game and concerntrated on every player that would be a brilliant method. The problem is it's unlikely that you watch every game (particularly NAB Cup) and even if you did it's unlikely that you would notice every players contribution.

But linear programing sounds complicated. Every year I don't have the time. I just cut and past them into a spreed sheet and divide ave points per game in the NAB cup by their price and rank them. Pick off the ones I don't like and look up the unknowns .

Players like Ablett, Reiwoldt, Goddard etc I don't care about their scores, All I care about is their role and preseason.