Player Volatility and Modern Portfolio Theory (1 Viewer)

[sterjs]

I've noticed a lot of people worry about the volatility of individual players like Chad Johnson, Anquan Boldin, etc.

These concerns seem overblown to me. A fantasy football team is like a portfolio of equities with 0 correlation(exceptions: players on the same team or players facing each other). Individually, the players may be high variance, but the impact on the team isn't quite so bad since all of your players don't tend to have good or bad weeks at the same time.

Another thing to think about: Is variance something to be avoided in Fantasy Football?

If your league pays first and second place, then 3rd place and 10th place are indistinguishable. If you're playing for bragging rights, only 1st place really matters. You want to build a team that has a maximum chance of winning, not just making the playoffs or whatever. The way to do this is to always take the highest Expected Value players and ignore Variance.

 

[Phil Elliott]

I ran across this theory on another website last year. I think it was quants.com, but it may have merged with another site.

 

[Muahahaha]

The problem with your theory is this:

Week one: 0

Week two: 0

Week three: 0

Week four: 60

After three zeros, he won't be in your lineup & you will be kicking yourself as he blows up on your bench.

People always extrapolate recent performance.

 

[Captain Spaulding]

Modern portfolio theory? sounds like a quantum physics lecture or something.

Redraft league strategy: I just pick studs based on supply/demand to fill the starting positions, then when studs are gone I pick potential studs, then when all potential studs are gone I fill up the roster with serviceable journeymen. Thats my quantum physics strategy for success. Any further analysis is a waste of beer drinking time, unless you love crunching data and that sort of stuff.

 

[CalBear]

You're begging the question by assuming that there are consistent players and inconsistent players. In 2005, Boldin had at least 62 yards in every game, and had 100 yards and/or a TD in 10 out of 16. So is he a consistent player or an inconsistent one? Depends what year you're looking at.

I think you should not worry about player volatility, but not for the reason you suggest; I think you should not worry about it because it probably doesn't exist, or if it does, it's a very weak predictive factor.

 

[sterjs]

You're begging the question by assuming that there are consistent players and inconsistent players. In 2005, Boldin had at least 62 yards in every game, and had 100 yards and/or a TD in 10 out of 16. So is he a consistent player or an inconsistent one? Depends what year you're looking at. I think you should not worry about player volatility, but not for the reason you suggest; I think you should not worry about it because it probably doesn't exist, or if it does, it's a very weak predictive factor.

I actually agree with this. I think that players have different volatilies, but even if they didn't, it would appear that they did because of second order variance (Variance of the Variance!). If you put 100 children in a room and induce them pick to stocks, some of those children are going to look like geniuses even if they have no predictive ability.

 

[Butters]

I think your portfolio theory is correct statically, but that Muhahaha also has a good point above. The behavioral psychology throws a wrench in the mix. Humans just don't "dig" variance in their systems and it doesn't mesh well with our internal heuristics. Hence, the majority of mutual funds fail to keep up with their appropriate market index, the vast majority of investors chase past returns, and I bench Steve Smith the week he goes 10-120-2.

 

[CalBear]

I think your portfolio theory is correct statically, but that Muhahaha also has a good point above. The behavioral psychology throws a wrench in the mix. Humans just don't "dig" variance in their systems and it doesn't mesh well with our internal heuristics. Hence, the majority of mutual funds fail to keep up with their appropriate market index, the vast majority of investors chase past returns, and I bench Steve Smith the week he goes 10-120-2.

But don't you think you might be chasing past returns when you draft someone because he had a "consistent" year in 2007? Do you know he'll be consistent in 2008?

 

[Butters]

I think your portfolio theory is correct statically, but that Muhahaha also has a good point above. The behavioral psychology throws a wrench in the mix. Humans just don't "dig" variance in their systems and it doesn't mesh well with our internal heuristics. Hence, the majority of mutual funds fail to keep up with their appropriate market index, the vast majority of investors chase past returns, and I bench Steve Smith the week he goes 10-120-2.

But don't you think you might be chasing past returns when you draft someone because he had a "consistent" year in 2007? Do you know he'll be consistent in 2008?

Yes, that sounds right to me. I agree with your statement above, that we probably should not worry about player volatility because it is largely random and a poor predictor. (I have no evidence for that, by the way.) My point is that, even if we shouldn't, subliminally we do.

 

[Muahahaha]

I could reference countless threads of people b*&^&*ing about volatility in players and them blowing them up the week they are benched. MJD and CJ come to mind immediately.

 

[Modog814]

1) I think that total game volatility is not a good indicator of future performance simply because you only have a sample size of about 16 games (+/- a few for playoffs or injury)

2) A player like Chad Johnson or Boldin you should not worry about volatility. They'll put up their points, and their volatility probably isn't that much different than any other WR in their tier.

3) You are correct and should generally ignore variance and select the highest expected value UNLESS you are in a Best Ball league/Contest where high Volatility can be a very good thing if you aren't required to select a player before the games start.

4) I believe variance is useful in FF and I haven't really seen too many people scratch the surface of it. But I think variance is more important for individual plays than total games. For example, say Boldin is targeted 150 times in a season and catches 90 of them for a catch rate of 60%. This is a simple binomial distribution (two answers: he catches it or he doesn't) and the variance of his catch rate is equal to:

Prob of Catching X Prob of Not catching X Number of Trials -> 60% X 40% X 150 = Variance of 36 -> Standard Deviation of 6-> So if we assume he'll be targeted 150 times again this year we can expected him to catch between 78-102 Balls with 95% certainty, it gives us a ceiling and a floor. Yardage can be done in a similar manner but requires more statistical analysis (since it's not a simple binomial distribution).

So in essence using this you'd be able to determine those with more "upside", which generally means they are riskier.

 

[Muahahaha]

I should add that I have largely started drafting stud WR, because I think that the wide receivers in the league with the most targets have the least week-to-week volatility. As a result, I second guess whether I have the right WR in my lineup every week. If you've got Moss and Owens, you start them every week; there is less of this Reggie Williams vs. Nate Burleson lottery ticket approach.

In general, players whose points come mainly from yards have the least volatility and those whose points come from touchdowns have the most. The top-ranked WRs tend to have a lot more catches and yards and therefore less random noise from touchdowns than the scrubs.

Wes Welker and Greg Jennings had the exact same number of FF points last year, but I would imagine that Welker was more consistent and in a lot more team's lineups when he earned them.

 

[CalBear]

I should add that I have largely started drafting stud WR, because I think that the wide receivers in the league with the most targets have the least week-to-week volatility. As a result, I second guess whether I have the right WR in my lineup every week. If you've got Moss and Owens, you start them every week; there is less of this Reggie Williams vs. Nate Burleson lottery ticket approach.In general, players whose points come mainly from yards have the least volatility and those whose points come from touchdowns have the most. The top-ranked WRs tend to have a lot more catches and yards and therefore less random noise from touchdowns than the scrubs.Wes Welker and Greg Jennings had the exact same number of FF points last year, but I would imagine that Welker was more consistent and in a lot more team's lineups when he earned them.

Your example is poor for a number of reasons, the primary one being that Jennings only played 13 games while Welker played 16.But even given that, Jennings was significantly more reliable than Welker in 2007. In the games they played, Welker averaged 10.3 fantasy points, Jennings averaged 12.6. Welker's standard deviation was 7.35, Jennings' was 5.78--Jennings had much less volatility than Welker. Welker had two games with 24+ fantasy points; Jennings never had a game higher than 21. Welker had three games with 3 or fewer points, Jennings had one.It is important to actually do the analysis. Otherwise you just let your confirmation bias tell you what you already believe.