Comments on: SGPs Part III: The Problem of League Specifics

By: Mays

Mays — Tue, 23 Mar 2010 22:58:07 +0000

@slackerjack: I agree that you can come up with proper SGP denominators for any category.

My contention above is that there are SGP systems that people use without considering what league particulars were used to generate values. If those settings aren’t the same as your league, your values are going to be way off.

As you point out, that’s not a problem if you are customizing your own SGPs.

By: slackerjack

slackerjack — Tue, 23 Mar 2010 17:44:44 +0000

I’m not sure I agree with the assessment that SGP’s are inflexible WRT league size, categories, etc.

Either using real data or simulated data (as suggested in post #2) you can generate SGP denominators which are customized to your league’s statistical categories. One of my leagues uses RAT3 (which is BB+H-K/IP) and it was very easy to modify my calculations to factor in that difference.

As for the number of teams (and league depth) choosing the correct baseline takes care of that. IE once you figure out what the worst player is projected to do and assign them a value of $1, it’s relatively easy to figure out the marginal value of the stars.

By: Nick

Nick — Fri, 20 Feb 2009 04:42:35 +0000

I think that’s an average vs. replacement issue. A 1B like Pujols will have less value if compared with the average at his position rather than replacement.

Rudy and Derek Carty were sort of debating this on Razzball, and I proposed a sort of test to try and find which baseline would more accurately predict standings in a league. Proposed something similar to what Mays did earlier here, but Rudy’s not into it. Trying to think of another test to handle the “which is the superior baseline” question.

By: Confused

Confused — Fri, 20 Feb 2009 03:28:45 +0000

It’s funny because earlier today I was looking over razzball’s one and wondering what was wrong with it, using chone projections for each one they had pujols as only the 6th best player while this one correctly picked him 1st according to the stats. Seriously razzball gave an extra like 2 total points almost because arod had 4 more runs and i think 9 RBI or something it was weird.

By: Molson

Molson — Thu, 19 Feb 2009 18:08:04 +0000

Upon thinking about it, rather than use real data for different league standings, why not use fake data? You could probably approximate final stat distributions by running simulations where you take a season’s final stats for a certain league type, determine the top n players at each position, and then randomly assign them among n teams and compute the final stats. Do this a whole bunch of times.

This could allow you to approximate the final averages and ranges for a league, but I’m not sure how accurate of an approximation this would be.

By: Homer

Homer — Thu, 19 Feb 2009 15:24:36 +0000

interesting stuff. I just noticed that a piece of software I bought (which I otherwise really like) does NOT change the SGP denominators when you change around league settings. I had thought it would, but after double checking it doesn’t appear to at all. I guess that would require that the software have data on all kinds of leagues which is unreasonable.