Wednesday, June 17, 2009

Standardized Scores

So, the past few weeks I've been mathing up the couple of fantasy leagues I manage. Mostly, I've been converting weekly head-to-head play results into rotisserie league scores. In the league I've been in the longest, we are already rotisserie, so converting to head-to-head isn't all that easy or informative so I didn't do it. What I did think about was the ranking system.

In most rotisserie leagues, your accumulation of a certain stat is ranked against your peers. You are then assigned a number to go along with that rank, but the current system does not account for incremental changes at all. In a ten team league, the person with the best whatever gets 10 points and the next gets 9. If doesn't matter if the person in first is ahead of the next place by 1 run or 200 runs, they can only get 1 more point than the second place person.

I decided to look at what would happen if you gave out points based on standardized (z-scores) instead of ranks and here are the full results (click pic to see full pic).

And here are the much more managable summary results (no individual categories) (again click for full somewhat readable table).

First, it is amazing how much movement there would end up being in the standings. The current 5th place team rises, and convincingly, to first. The 1st and 3rd place teams both plummet. I suppose this means those dropping a lot may be winning several categories by a little and losing a few others by a lot. It could also mean that they are tanking a few categories along with other teams. If 3 or 4 teams give up on saves, then only 1 team appears at the true bottom of the rankings and gets 1 point, the others would get 2-4 points and not suffer than much from the tanked category.

On the other hand, if a category was particularly close and bunched together, someone may suffer and get a low ranking despite being close to the league average or even the league leaders. In the standardized system, they wouldn't lose many points and wouldn't be anchored trailing slightly in a tight race.

There's a lot of stuff to this system, and I find it pretty fascinating. One interesting note is that of the 100 possible spots (10 categories by 10 teams), I only caught 3 that would be considered outliers (past + or - 1.96) and all 3 were averages and not pure numbers. That is to say, based on the number of times per something else and not just the total result.

It would be interesting to see a league run this was for a season to see how you could exploit the system. I think if I were playing in this league, I would make a go at drafting an all steals offense and not even play a pitcher.

If you could get enough steals, your players would probably stay near average in runs and batting average. The gain from being so far above in steals would most likely more than make up for what you lose in HR and RBI.

As for the pitching, by giving up on wins, K's and saves, you'll be a low outlier. But first, you'll drag the overall average way down making you less of an outlier. Second, if you have a 0.00 ERA and WHIP, not to mention 0 losses in this league, you're likely to lead those categories by at least as much as you lose the stats.

Of course, all of that would be for not if someone else used the same strategy. Or would it?

No comments: