THE BOOK--Playing The Percentages In Baseball

Is this a thread where we state our interest? I am interested.

Wasn’t there a thread on this last year where the weights were all worked out?

There was a thread last year about this. I would like to get this started with only people that frequent this blog. If anybody’s interested, just email me.

Does anyone have a link to that thread?

While all of you were looking for the original thread, I already emailed him—MUAHAHAHAHAHAHA!!!

Thanks for the notice, Tango.

It was actually 2 years ago:
http://www.insidethebook.com/ee/index.php/site/comments/yahoo_fantasy_linear_weights_league/#comments

Well, that was quick. The league has filled up already. It looks like we’ve got some quality managers, all from this blog. Thanks Tango. Wish you were playing.

I literally lose focus on half the things I need to be doing. Being a Fantasy manager is no longer for me.

Feel free to ask me for anything else you need.

I’m starting a fantasy league along similar lines.

I was going to use Tango’s point system, but the problem with them is that since they are set to replacement level without a position adjustment, a number of starting SS’s and C’s show up as negative value players...meaning that, as a fantasy manager, it’s more productive to plug in an injured player or demoted AAA player into that slot than start some of these guys.

I’ve decided that the best way to address this is to use a point system that is more set to an absolute runs baseline than a replacement-level baseline.  This way, it’s the league market that sets up what will be our “replacement level”, and it’s incumbent on the manager to find players at or above replacement level when they have a player go down to injury (not doing so will have very bad consequences).

This was pretty easy to do for hitters.  Here are Tango’s replacement level points:
AB -2
H +6
2B +3
3B +5
HR +8
SB +2
CS -4
BB +3

Here they are in absolute runs-based points (they are just rounded absolute linear weights):
AB -1
H +6
2B +3
3B +6
HR +9
SB +2
CS -3
BB +3

There are rounding issues that throw things off, but this more or less will give you points = wRC*10 for hitters. The main difference is an adjustment of the value of an out from -2 to -1 points (this is also seen with the CS).  I’m not sure why my triples and HR’s also are getting an extra point here--I tried to replicate Tango’s work but was getting these same values in those cases as well.

Anyway, my problem is that I’m having trouble figuring out how to convert Tango’s pitching points into something that mirrors wRC.  Here are Tango’s points:
IP +2
H -0.5
BB -2
K 1
HR -3
2B -0.5
3B -0.5
SV +2
Hold +1

Any ideas how to make this a good approximation for absolute runs?  I could probably bump up the point value for innings pitched from +2 to +3 (or should it be +4?), but I’m not sure if that would really do what I want it to.  Part of it is that I’m not entirely sure how Tango came up with his framework, so I can’t reconstruct it.  Advice welcome!

Thanks,
Justin

P.S. If you’re interested in my league, there are still ~5 slots open:
http://www.redreporter.com/2010/2/12/1308231/sabermetric-fantasy-baseball-anyone
Send me an e-mail so I have your contact info if interested.  It probably won’t be quite as hardcore as ShaneM’s (autopick draft, transactions set to weekly), but my hope is that we have a nice balance between competitiveness and time commitments.

I’ve been fussing with this literally all day and finally came to something I’m somewhat comfortable with.  Since others might be doing the same thing, I thought I’d report back. 

Here is ultimately what I came up with for pitchers:

IP +5
K +2
BB -3
HR -13
SV +6
HD +2

I didn’t set out to have those numbers match up to the coefficients in the FIP equation.  I started with linear weights and went through a process that was as much guesswork as it was sensical.  My plan was to create a DIPSy adjustment of absolute-runs linear weights....but those numbers started out too low, so in desperation I started increasing the positive terms, eliminating anything related to hits...and the result was suddenly something that looks a lot like the FIP equation (whether by coincidence or not).  Seems to work pretty well--the rankings of players seem intuitive to me--though I don’t think my methods would hold up to much scrutiny. 

The saves issue is tricky.  I hate saves with a passion, but we have nothing else to work with that relates to reliever leverage.  This gets elite closers in 2009 equivalent in value to starters like Aaron Harang, Johnny Cueto, Brad Penny, etc.  That feels about right to me.  I’m throwing setup guys a bone with the +2 holds, but it will take a lot of holds to get anywhere close to a closer’s value.  ::shrug::

Anyway, the system has good pitchers at 800+ points, great pitchers at 1000+ points, and elite pitchers at 1250+ points.  This matches up pretty well with the hitter scoring, which more or less is wRC*10.  Based on the number of slots for pitchers (8) vs. position players (9) in my league, I’m crudely estimating a 55%/45% split between hitters and pitchers in terms of team point totals.  Not quite the 60/40 Tango’s system provides, but it feels close enough to me.

What I like about the wRCish system I’m employing is that negative players are hard to come by, so it’s always better to have an active player in the lineup than someone who isn’t playing at all.  Managers in the league will essentially create a replacement baseline by plugging in the best talent they can when their starters go down.  And we’re still, at least with hitters, tied to something that is somewhat similar to reality.
-j

My philospophy for a hardcore sabrmetric league would be “reward skill ignore luck”

I would take two elements of “The Book” the components of wOBA and the calculations of skill vs luck.

I would regress the woba components of an event by the amount of skill involved

so for pitchers you only have w k hr if you assume no skill in babip.

W and K are more consistant than hr so HR shoudl be downgraded slightly and I would use the average value of a bip for all other events which takes care of IP.

For batters it is a bit more complex but I would also regress the woba values of s d t to the mean of a bip event as there is less skill there than in w K or HR.  I’ll dig out my estimates and try and make some scores although I am sure Tango has better estimates.

I would use absolute rather than relative tuns and create a replacment level value for pitchers and batters so if you dont have enough innings or PAs for a position you must use this to avoid the problems of weak hitting c or SS being more productive on the bench

Each week you would have your runs scores and conceded I would then use a pythagorous/log 5 like method to give you your wins head to head against the rest of the league.

Beyond the Box Score needs two more participants in a WAR-based salary cap league.  12 teams, picks spread out over two weeks, no transactions after the draft, $60M salary cap using real 2010 player salaries, scoring by fWAR.  Email me for more info () or check out…

http://www.beyondtheboxscore.com/2010/2/17/1313906/boab-zack-greinke-division-draft (A different league that’s ready to roll)

http://www.beyondtheboxscore.com/2010/1/25/1269273/btbs-ball-on-a-budget-fantasy (some rules specifics)

Jinaz 9/10,

If you are around, could you post the decimal places on your linear weights?  Thanks.

Hi,

I just used these linear weights:
http://www.tangotiger.net/bsrexpl.html
I’m using lwts_RC column since I’m doing things in absolute runs.

Since a single is an AB+H, and a double is an AB+H+2B, some combining of these stats is necessary so that the totals come out right.  Here are the “combined” linear weight values I’m using, followed by the rounded point values for each term:

Event lwts_rc Pts
AB -0.1 -1
H 0.563 6
2B 0.287 3
3B 0.57 6
HR 0.939 9
SB 0.193 2
CS -0.282 -3
BB 0.3 3

Hope that helps.
-j

Yahoo allows you to use Assists and Putouts and Errors. What would reasonable values be for those categories?

You can’t change the relative point value by position but maybe there’s still some reasonable values?

If you use putouts, C’s and 1B’s will break whatever you try to do.  And without putouts, outfielders are nearly worthless.  I don’t think it’s workable.
-j

In my head I was assuming strict position requirements, i.e. 1 at each position and no Util players, or something like that. Doesn’t that solve the problem of PO meaning completely different things for 1B and C?

Ugh, I didn’t know my email address would be “published.” Should have read the form more closely but I was assuming it wouldn’t be published based on previous comment forms.

Anyways I guess having strict position requirements doesn’t fully solve the problem because the scales are different at the different positions. An outfielder who gives you 50 more PO may be very valuable in real-life, but 50 more PO at 1B at C is not a big deal. Is this the problem?

Matt: I deleted it.