THE BOOK--Playing The Percentages In Baseball

Thursday, April 02, 2009

QALPAP

By Tangotiger, 12:27 PM

This is mostly, but not strictly, a post about Fantasy valuations.

I gave my recipe a few years ago here in post 66: QALPAP (quatlus above last player at position).

In preparation of the Forecast Challenge 2009, where I was going to post the Marcel/Community Fantasy dollars, I have some doubts. This came about because, to my surprise, there was a falloff between the 39th and 40th catcher. A huge one. It was my expectation that I would always get a nice curve at the tail-end, as I’d start getting a bunch of players at the end.

This however did not happen in this isolated case.

Then I got to thinking some more: my top 40 catchers won’t be the same as everyone else’s top 40. Indeed, I’d expect that perhaps 5-10 of my 40 catchers to not be drafted (and similarly, 20% or so at each position class). So the “last player at position” should really be “last player expected to be selected at position). QALPAP is now QAlpetbsAP.

This got me thinking some more. If your opponent is doing a really crappy job at selecting players, such that perhaps the 15th, and 20th and 25th and 26th and 29th catchers ended up not being selected (your process gave you the #1 and #14 catchers), surely this must affect the entire valuation process, won’t it? After all, if you know (or could have predicted) that the 15th catcher would still be left standing, there’s no reason for you to have selected the #14 catcher as high as you did. So, it’s not only the “last player expected to be selected”, but “best player not expected to be selected”. So, we really have QAbpnetbsAP.

The problem is trying to model that. I’m at the point where what I’m thinking about doing is taking the average quatlus of the bottom half of the position pool (#21 to #40th catchers, in a 20-team, 2 catcher league), and set that as my baseline “best player not expected to be selected”.

Thoughts?

• Sabermetrics • Fantasy

#1 Drew (see all posts) 2009/04/02 (Thu) @ 13:40

I think averaging the bottom half is a reasonable way to do it, but that might be overestimating the quality of the best undrafted player. I might go with averaging the bottom quartile. Of course, I have no scientific reason for saying that. Anecdotally, it seems like the top 75% of each position pool almost always gets drafted.

#2 Tangotiger (see all posts) 2009/04/02 (Thu) @ 14:00

Well, if the top 75% of the pool gets drafted (i.e., bottom 25% that should don’t), then if I take the average of the bottom 50%, then I should get pretty much exactly what I want, shouldn’t I?

I suppose what I should do is take a probability rate of each slot starting with the 50th percentile player and going down. So, the chance that the 50th percentile won’t be picked is 5%, the 51st percentile is 7% ... the 100th percentile is 90%, etc, etc, etc, and come out with a weighted average.

#3 Zach (see all posts) 2009/04/02 (Thu) @ 18:05

The way I did mine was by looking at the number of each position ranked in the top 252 overall--the number of players drafted in a standard Yahoo! league--based on Yahoo!’s default rankings. (You could also do this for ESPN, CBS, etc., leagues.)

For instance, Yahoo! ranks 30 first basemen in the top 252, so I used the #30 1B as the “last player at position.”

#4 Tangotiger (see all posts) 2009/04/02 (Thu) @ 19:14

Yes, that’s QALPAP.

#5 Zach (see all posts) 2009/04/02 (Thu) @ 21:07

Yes, but the way I did it, though, the #17 catcher was the last player at catcher, while the #30 first baseman was the last 1B drafted, instead of the usual last player being the same.

#6 Hizouse (see all posts) 2009/04/03 (Fri) @ 13:07

I wonder if in a fantasy league, your replacement level or baseline should be the last projected starter at each position. You can draft that guy late, after everyone else has drafted a starter at that position. And you’d probably rather have that guy as starter than grab your first reserve at another position, even if the first reserve has more QA waiver wire fodder.

This sets the last projected starter at each position as having equal value, as opposed to the first undrafted player at each position as having equal value.

#7 Derek Carty (see all posts) 2009/04/03 (Fri) @ 15:46

This is something I’ve been thinking about myself a lot lately, though I haven’t come up with a great solution yet. I suppose you could assign every player a probability of selection for you own particular league and then determine the value of your last player selected based off of that, but you’re still dealing with subjectivity.

#8 Tangotiger (see all posts) 2009/04/03 (Fri) @ 16:08

Derek, right that’s what I was thinking in post 2.

Hopefully, after I run the Forecast Challenge 2009, I can come away with a whole host of new data. I can come up with how often Marcel’s #1 through #500th ranked player was selected or not. And I can do it for each entrant. Stuff that could really be fantastic for research and fantasy purposes.

#9 Derek Carty (see all posts) 2009/04/04 (Sat) @ 15:35

Gotcha, Tom. That should provide some very cool data.

Page 1 of 1 pages

<< Back to main

THE BOOK--Playing The Percentages In Baseball

Thursday, April 02, 2009

QALPAP

Latest...