Yahoo Rankings

[Amazinz]

Smooth, what you're describing is basically the SGP model (Standards Gains Points). You can Google it and search here and read about the pros and cons. There are several big flaws with SGP (IMHO of course, not everyone agrees with this but many do) and it will help you to read about the pitfalls.

[rookies and cream]

Here is the problem with the SGP model, VORP, or other methods in which z-scores are calculated:







Anybody want to guess what's wrong here??

(BTW - These graphs depict last year's stats for all players that were active in my league at the end of the year)

[Fantasy Sports Genie]

While I see how that data makes the results look funny, I'm not necessarily convinced that data makes the method incorrect for evaluating fantasy value. It suggests perhaps it is unreasonable/unwise to weigh steals and homers with equal value for scoring, but since the reality is that they count equally in any league that uses both (in roto scoring), and those few outliers on your SB graph can just about single handedly win you a category...

[rookies and cream]

I'm not really referring to the outliers (in fact, I do agree that players like Reyes, Pierre, Crawford have great value), I'm referring to the distribution of the data. The fact that SB's (and saves for that matter) are not normally distributed, while the other categories are, makes it inappropriate to simply sum up the individual categories to get an overall score. All categories are not distributed equally and whichever method you utilize needs to account/correct for this.

[Amazinz]

SGP does account for this inherently because each category is being valued in context. The problem IMHO with SGP models is that they tend to overvalue the production needed to be the worst team in a category. People have added steps to the process in order to account for this but the systems end up being overly complicated scarcity models.

As far as the outliers are concerned, this is my main issue with standard deviation versus baseline. The zScore method tends to seriously undervalue the outliers, or from the other perspective baseline seriously overvalues them. Something has to give. They can't both be right.