This one turned out to be much more difficult to word than I was anticipating...
Prerequisites for understanding: None
Prerequisites for deriving: N/A; conceptual.
Just One Number
A reasonable understanding of value is a critical tool in the sabremetrician's arsenal. While may many argue that you can't put a number on every little thing a player does, when a team is making a transaction they must assign some value to players lest they make completely arbitrary decisions. When a player signs a contract with a team, he's signing for what the team is willing to pay him. This may or may not jive with what fans, analysts, or even other teams may think he's worth, but no team gets a bad deal on a player deliberately, nor do they make stupid trades on purpose. In order to be even remotely efficient with their resources, teams must have processes for determining player value. How?
The Currency of Baseball
This is actually the big question in baseball analysis: How much is a player worth? It is impossible to have an informed opinion on any transaction that occurs without subscribing to a logical definition of player value. This value is essentially the currency with which baseball operates. Are those two prospects a good return for that star slugger about to walk in free agency? Was that ace pitcher worth tying up ten percent of the team budget for the next seven years? Clearly the idea of salary as value is inadequate here - there's no reasonable way of comparing prospects with an established player using salary alone. Ironically, the rules regarding club control of young players means that the dollar is not, and never has been, the currency of baseball.
What is it then? The answer is simple: wins. A baseball team is trying to get as many wins as possible both now and in the future given the resources at hand. More wins means more people at the ballpark, which means more revenues. More wins also means the chance of a playoff spot, which has a huge impact on team finances. In other words, value can be measured by how many wins a player adds to a team in any year. Acquiring wins drives the whole system, and they're worth a certain amount on the market - over the last few years, an expected win has typically been worth around $4 million per season when signing free agents.
Teams often differ wildly in terms of how they calculate expected wins. This means that they'll value players very differently, leaving some to make head-scratching moves while others gobble up apparently undervalued assets. Our goal as analysts is to provide a coherent, consistent methodology for evaluating players in terms of wins. If a statistic cannot be translated into wins, it is of minimal utility to us. Fortunately, most can.
What Follows
2 recs | 38 comments
I've been meaning to raise this point for awhile now...
…and maybe I ought to wait until the next post, but why is the conversion from runs to wins necessary at the player level? Given the variable environments that baseball players occupy, it seems like converting to wins can, in some way, actually dilute the meaning behind a player’s worth, especially since the reliability of defensive stats is (in my opinion at least) not nearly as good as that of offensive stats.
After all, a big bat added to a poor defensive team, and that same bat added to a stellar defensive team should yield different win results, but likely the runs added by the bat will have less variance.
Maybe I’m misunderstanding the necessity of the run/win conversion, but it seems like an added layer of confusion, like quoting oil prices in Yen, then converting Yen to Dollars to gauge the price of a barrel of oil.
Thanks for this series by the way, I’m very much enjoying it.
Andersean - February 16, 2010
You raise some good points
But instead of ignoring the run-win conversion (which is a post down the road, of course), it is probably better to simply look at where the runs are actually coming from.
Like I said, we need one measure of player value to make any sense. However, we have a completely separate idea in our heads while we look at this one number: our confidence that we’ve evaluated the player correctly.
Graham MacAree - February 16, 2010
At U.S.S. Mariner that dude gets shut down for disagreeing and being an 'idiot.'
Thanks for the series. We’re not all statisticians and this series has been really, truly, fucking great as an introduction to the technical derivation of the numbers that a lot of us take for granted.
THolt - February 16, 2010
?
Find me even a single politely worded disagreement or well-articulated question on USSM that received a harsh response.
Milendriel - February 17, 2010
There's not really a need to slam USSM here
But I appreciate the appreciation nonetheless.
Graham MacAree - February 17, 2010
It might also be useful to discuss the difference between
Wins above Replacement and Wins above Bench (or whatever you wan to call it). People always say that a 2 win player is worth around 8 million but if he only gives you a 0.5 win upgrade over what already have then he really isn’t worth 8 million to you. We often use WAR to evaluate the $/win in a contract when we sign a player to tell if our GM did a good job but maybe we should use wins above bench instead of WAR. I can see both sides.
Edgar for Pres - February 16, 2010
Just to be clear
If I haven’t talked about a concept like replacement level at all, I’m probably planning a post once the basics are all wrapped up. Right now, I’m exploring the statistical toolbox and first principles. There’s a lot yet to come.
Graham MacAree - February 16, 2010
Sorry for skipping ahead. I'll hold my tongue.
I don’t know if you want to do this but a list of planned topics might be cool.
Edgar for Pres - February 16, 2010
Not to be overly pithy
But basically: everything.
This is going to be a lonnnnng series.
Graham MacAree - February 16, 2010
How much are you going to split things up?
Are you going to profile each of the DIPS metrics in different posts? How about the defense metrics?
vivaelpujols - February 16, 2010
I think you're going to have to wait to find out ;-)
baetown415 - February 17, 2010
My plan is to split them up by data source.
Graham MacAree - February 17, 2010
Speaking of toolbox
Is it possible to put these in a section separate from “The More You Know” like “Sabermetrics 101?” If it’s going to be a long series it could benefit from having a section that contains exclusively this material!
ThundaPC - February 17, 2010
Done
Graham MacAree - February 17, 2010
Sweet!
Thanks!
ThundaPC - February 17, 2010
Good idea
Jeff Sullivan - February 17, 2010
This is another reason why I have long wondered about the necessity of conversion to wins
Wins, naturally, require some starting point, either replacement or bench, whereas offensive run measurements, based on my understanding of wOBA, linear weights and the like, don’t require some agreed upon value.
Also, to be clear, I’m not saying that I think it’s wrong to use wins as the currency, I’m simply questioning whether its more useful to use wins instead of runs.
Andersean - February 16, 2010
Wins can be wins above average, which is self defining.
So that isn’t really a valid indictment against wins.
vivaelpujols - February 16, 2010
For a team, yes
but as you add players to a roster I would imagine it’s not a simple linear addition of their wins, if you were using wins above average.
Andersean - February 16, 2010
Also, to be clear
I’m not saying that wins are bad, I’m just trying to understand why wins and not runs are the preferred currency.
Andersean - February 17, 2010
It comes down to logic (as all sabremetrics should)
There is no real reason that a team would be paying for runs. It’s pretty clear how wins themselves benefits teams though, and so the only way we can use runs as the currency is to find a transformation between runs and wins.
Graham MacAree - February 17, 2010
When you frame it like that
it makes more sense, but I still question the need to translate, since you have to go through runs anyway to reach the win value. That is to say, if there were a stat that went directly from some observable phenomenon to wins without going through runs, I’d concede it makes more sense to have all stats converge on wins.
Thanks for taking the time to respond to my questions, looking forward to the rest of the articles.
Andersean - February 17, 2010
There isn't really a need to translate but you're doing it implicitly when you're thinking about value
Does that make any sense? If the wins/runs relationship didn’t work, we wouldn’t really be able to have a logical reason to measure players in terms of runs above or below whichever benchmark we chose.
And you’re very welcome. If I wasn’t getting questions, then I would worry.
Graham MacAree - February 17, 2010
Makes perfect sense
and the more I think about it the more I’m realizing that even mentally you need to have some way of drawing meaning from run total anyway. However, I’m left musing over whether the conversion needs to be explicit, but I don’t believe that’s a winnable or worthwhile argument since we’re essentially on the same page here.
Thanks again.
Andersean - February 17, 2010
There is still a basis for runs above (something)
The something can be zero or it can be average or it can be replacement. For wins the natural benchmark that makes the most sense tends to be replacement or at least that is the most common.
Edgar for Pres - February 16, 2010
pitching is the currency of baseball
-Dayton Moore
Freneau - February 16, 2010
Models - the mathematical kind - and validity
As a statistician, my big concern about many of these metrics and models is the validity and reliability of them. Conceptually these ideas are great, and as both a fan and a stat guy, I’m very interested to see in where this discussion is going. We can build great statistical models of practically anything, but if they don’t really estimate what they intend to estimate, they aren’t of much value.
To me, this PECOTA prediction seems to be a good example of a system that predicts wins every year, but I have never seen it “back checked” to see how well it worked, which would/should lead to refinements.
Thanks for the “course”
New England Fan - February 17, 2010
Oh don't worry I'm definitely going to be talking about reliability and validity.
Graham MacAree - February 17, 2010
And Baseball Prospectus's black boxes
Graham MacAree - February 17, 2010
Tango's done some back checking on the various projection systems
I’m sure Graham will get to that.
(PECOTA is interesting in that it seemed to be much better for pitchers than hitters, and overall has come in a bit worse than Marcels recently.)
marc w - February 17, 2010
Yeah I'm interested in PECOTA too
The methodology is interesting but in practice it was awesome and now kind of a disappointment.
Edgar for Pres - February 17, 2010
Graham
I noticed in the Sabermetrics 101 file that the posts are sorted in ascending order. Is it possible, just for simplicity’s sake, to sort them in descending order as opposed to the format you are using now? It seems to me that it might be simpler for the new folks if it were a “start at the top and work your way down” kind of format.
ToddK - February 17, 2010
I honestly have no idea
I’ll look into it.
Graham MacAree - February 17, 2010
I don't know if that's possible
Jeff Sullivan - February 17, 2010
Could just start a FanPost and continually bump it
with links in descending order in it. I’m sure we could rec it to the top of the FanPost list
seattlebruin - February 17, 2010
At the bottom of each one there can be a "previous posts" list that keep updating?
That might be annoying though or eventually very long, but it could work.
CapSea - February 17, 2010
Well, if I think something needs to be referred back to, I link it in the intro as well as where it comes up in the post
It’s happening in the one I’m writing right now.
Graham MacAree - February 17, 2010
Works for me.
Besides, there is no real reason to worry about how people access it until its over.
CapSea - February 18, 2010
You must Login with your SB Nation account and be a member of Lookout Landing to post a comment.