Next year, we are totally going to win a playoff game.
I would be very interested to hear more about your idea though gooner. I am sure there is something to what you are saying, and I am sure there are teams looking at that.
But I think teams don't necessarily want to be well diversified (with the exception of relievers). I would want a team, built around a single model or mission statement. If your idea is to get on base, then get guys that get on base. Adding a Scott Hairston doesn't help our the model. Now I think there may be something to diversifying yourself a bit, or in essence getting the cheapest player available per win, but I think that is something that teams currently do like for instance the Rays and A's and to some degree the current Chicago Cubs (though I don't see this being a long term model for the Cubs its just a way to go from awful to being mediocre to get people off their backs a little as we rebuild).
Its been a while since I have dealt with a CAPM or an SML, but I would be intrigued by your thoughts such as what is a risk free rate or what is beta and would it be different for different positions? Or maybe I just misread your thoughts since you said the CAPM has its shortcomings.
I am not suggesting teams should be well diversified. Teams certainly take on specific types of risks. These risks are sometimes conscious decisions, and sometimes decisions based on necessity (payroll, stadium, etc.). End of day, I fundamentally believe two things:
1) It is possible, over time, to determine not only the relative value of these risks, but to determine the future value of these risks by measuring the volatility associated with said risks (by, in turn, understanding the volatility of the various assets (players) that represent said risk.
2) Based on the above, it should be possible to not only optimize HOW an organization builds a team around specific types of risk (the various pieces that go into it) but also to understand the relative value of taking on one kind of risk vs. another kind of risk.
edit: put another way, yes, I do fundamentally believe there is a risk free rate of return. It is similar to (but not equivalent to) replacement player value. Different players may exceed such risk free rate of return but they do so with varying degrees of volatility. This is true at the player level, but is probably also true as you start to aggregate players by various characteristics. By understanding the volatility (beta) you should be able to arrive at a more realistic wins contributed number. Because not all wins are the same. Some are more risky than others.
Last edited by gooner; 01-30-2013 at 11:31 AM.
Next year, we are totally going to win a playoff game.
Every time gooner posts, I realize that I am stupid. Not because he makes me feel stupid, but I realize just how many things I never could even imagine considering on my very own.
Miggy Montero BelievesDo you?
You didn't understand that? And to think 1908 is molding young minds
And as far as your beta I am not sure how much of a sample size you are going to want to use for your beta. I mean using only 3 years is going to be huge and have a huge margin of error, but using 5 or so years isnt going to be all that useful because of age and what not.
I guess I am trying to figure out what these "risks" are that you are looking at. To me risks would be like injury risks. Kinda what we are doing currently with people like Scott Baker. He has what I would call I suppose a high beta because its risky. Granted we dont have a market to go off of persay. Lets say he comes with a beta of 3. So he could be worth 3 times the market value if he does well or none if he does poorly.
Where as someone like a Feldman to me would have a very small beta, because no matter what the market (again just a term of expression for this) does he is going to be pretty much the same.
Is this what you are talking about?
My mind couldn't be any more blown if I was Kurt Cobain.
The Final Words of Dale Horvath. Wise to the end......
First, what I mean by risk is uncertainty. Variability of performance. Volatility. Beta.
I don't think you want to be too literal about how we apply CAPM to this, but I think there is a lot of value in the quantifying volatility approach to asset valuation. Basically, CAPM suggests that the required rate of return for an asset is the risk free rate of return plus a premium that compensates you for taking on the additional volatility. In today's markets, US Treasuries are used as the risk free rate of return because ultimately, as you suggested, there is no such thing as a true risk free rate of return. It's the closest thing we have, but even it has risk (i.e. volatility and variability).
In baseball terms, the analogy is that the true production of a team is its production (lets use Wins for now and not open a second can of worms) multiplied by a fraction that represents the volatility of that production, indexed for overall volatility of that particular type of production. I am not necessarily warranting the ability to use this tool to judge an individual person (not in the context of this conversation at any rate - lets save that for another day). Rather, my interest is in evaluating overall strategy and team composition.
There are two things to consider here. There is production, and there is variability of that production (volatility). Consider the replacement player. Given you are required to field a team, the replacement player represents the league minimum level of production. All else being equal, you can be assured (albeit with a base level of risk) of this level of production. In actuality, even this production has some variability to it (as do US treasury bills), but it is easy enough to calibrate that base level noise out of the equation. Let's call it systemic risk.
Above this replacement level player, there are players who provide superior return (more wins). However, year over year, the production of those players is certainly not consistent. It varies. Now, consider the nature of that variation in production. There are two aspects to it (in the context of wins). First, there is variation in how much the components that constitute win shares contribute to said production. Defense, base running, hitting, etc. Second, there is variability within each of those components year upon year.
Here is my basic hypothesis: as you start to dimensionalize and aggregate the variability of performance, you will find that there are patterns in variability across categories and over time. If you can quantify the relative volatility of these various components (and their cross correlations) you can develop more efficient strategies around team construction (i.e. how to build your portfolio of assets).
The current paradigm of player valuation is predicated around assigning a dollar value per win. This is a highly flawed approach because it doesn't take into account two things. First, it doesn't take into account the overall volatility of that win share number (and doesn't account for the variability of the relative components of that win share). Some wins are more valuable than other wins because there is less volatility in them.
Second, it doesn't contextualize the win share based on the type of volatility you are looking for. As a team, you make determinations about what kinds of volatility you want to be exposed to, and the types of hedges you are seeking. Some win shares should be more appealing to you than others because of the composition of those win shares.
In reality, GM's make these types of strategy decisions all the time. I think there is a way to apply a strict numerical basis for those decisions.
Next year, we are totally going to win a playoff game.
I kinda see what you are saying, but I think it still comes back to what I was saying with the injuries. I don't think you are going to find a statistical way to show the volatility of a player because in order to get a sample size you are going to need several years, and unless you are going to go to a monthly basis (which honestly might work) you are going to have such a small sample size that even if you did find the variation you are basically just throwing out a random number that means nothing. Now going to a monthly type variation you may be on to something, because you can say he will be worth this much within a reasonable amount of certainty.
You could use that to find a beta and thus load up your team with players that are both higly volatile (to me like a soriano) or someone who isnt volatile (someone like an Adam Dunn pre 2011). You could do it with injuries as well, and the whole point would be to try to get your beta to somewhere like what 1.1 or so? Or 1.5? Granted again I am using the numbers you would use in the financial world, but if you are looking for something like that I would think you wouldn't want to be at a 1, because at that point whats really the point?
I think if you are a big market team like say the Yankees yeah you probably want to be at a one or so because you want to have a pretty stagnant result. With the higher beta you are going to be looking at a much wider range of your victories and as the Yankees you don't need that. You need stability and to just get into the playoffs on a consistent basis. So personally I think your model wouldn't work for a large market team.
As far as a small market team like the A's I think this is a decent idea, but I am not sure its really much different than what is being used now. With limited resources you have to take some chances or buy some riskier stocks. With this though you are going to be looking at a much wider array of win totals. You could be looking at a 70 win team or possibly a 95 win team if the pieces fall into play.
It's an interesting idea, but I am not sure how well it would really play out. The rRF would have to be really really low IMO, much like it is in real life.
Maybe slightly OT, but was there ever a statistical reason for Adam Dunn falling off a cliff? Would that be something that you would try to predict gooner? Or is that just too random of an anomaly to take into account?
Links to two email interview with Tom Tango are below. Excerpts are all Tom Tango quotes. You can also try contacting Tom at firstname.lastname@example.org.
1/30/13 - Jon Greenberg's interview, "Q&A: New Cubs 'saberist' Tom Tango."
I can't provide any consulting services to any other MLB team, and I'm limited to what I can do on my blog. Because of my (non-disclosure agreement), I don't discuss any of the particulars of my work.The issue with the fielding stats that we have now is that we have to infer a lot simply because we aren't recording enough. You'd rather record the fielder's positioning rather than infer it. You'd rather know how many hops a ball takes to get to the shortstop rather than infer it. Basically, all the things we see and we know and we take for granted as a baseball fan isn't being recorded. Even something as simple as hangtime took forever to finally get recorded. You and I know looking at a seven-second lazy flyball is going to be caught by every outfielder in MLB, and is therefore noise. But, if the systems aren't being told that it was a seven-second flyball, it tries to guess based on other parameters on its difficulty, and therefore might suggest it had a 90 percent of being caught rather than 99.9 percent. Instead of that data being treated as noise, the fielding system treats it as valid useful data.
But, just because a metric has bias or noise doesn't mean we should discard it altogether. We need SOMETHING. As long as the bias and noise isn't too extensive, then something is better than nothing....the one thing that is most important is to make sure you understand the context of the stat. Don't just look at RBIs, look also at how many runners are on base. And not just how many runners, but where are they on the bases. And not just where, but how many outs, especially for runner on third. Once you realize there's so much bias in a player's RBI totals you learn to move away from it, and focus on the skills that lead to RBIs.8/24/11 - Justin Bopp's interview, "Q&A with SaberWizard Tom Tango."Every team always tries to make the best moves possible, given the constraints at hand. The issue is always how to best value each piece, each move, for both the short- and long-term. And Cubs fans should feel confident that the front office has and will have the best people in place to make those decisions.
Every team is always looking to make itself better, and they are all trying to manage these various moving parts, some of which have more value today and others have more value tomorrow. So, try to understand why a team is doing what it's doing: use their perspective, rather than use your own perspective as to why something is good or bad.I don't know how things work with most of the teams, but you have to presume that a healthy majority of them have alot of smart people working on the valuations, and they are reasonable from their point-of-view. Occasionally, you'll get a high-profile blunder, or perhaps a team or two simply shoots from the hip. But, those are in the minority.
Last edited by Yagyu+; 01-30-2013 at 03:49 PM.
Eat your heart out Jeffy?