Fantasy baseball is a results-oriented game. We can talk about bad luck and good fortune until we're blue in the face, but at the end of the day, the teams with the highest batting average and lowest ERA get the most points in the respective categories.
But that's the present; what about the future? Those of us in the business of crystal-balling future performance are much more process focused. What's done is done. Tell me what's going to happen next.
Whattaya mean "regression"?!
The best way to do that is to set aside the results and direct your attention to the process.
With that is a backdrop, in lieu of the normal "Round Table," I'm going to look back at the season of a couple of players for which the process is even more important than the result. Fear not, I'll gather the troops for a final hurrah next week.
The first player to be discussed is Mike Trout -- you may have heard of him. While there were pundits on either extreme, to me the most reasonable take was Trout's batting average on balls in play and home runs per fly ball would regress with regress being the more formal definition of approaching a mean.
The problem with this is Trout had not yet established a mean. Sure, using minor league equivalents added a bit of data to the sample, but MLEs aren't nearly as reliable as actual major league performance. So when we proclaimed Trout's BABIP and HR/FB would regress, it was due to the fact their 2012 levels far exceeded those which could be repeated, but we didn't have a realistic expectation of the mean -- only that it was lower than 2012's numbers.
So what happened? With a week and a half left, Trout's .384 BABIP is basically identical to last season's ethereal .383 while his HR/FB has dipped a bit from 2012's 21.6 percent to its present mark of 16.6 percent.
On the surface, it may look like the regression proponents were half right, so the process was half right. But is that really the case?
We still don't know Trout's baseline BABIP even though we now have two full seasons of data. Many will point to consecutive seasons of a BABIP north of .380 and call it Trout's established baseline. And they may be right.
But the probability of sustaining a career BABIP over .380 is extremely slim. Let's circle back to separating the process from the results and see if there is anything wonky with Trout's BABIP that may suggest we were just a year early predicting regression.
The two factors primarily fueling BABIP are line drive rates and percentage of hard hit balls. Trout's 23 percent line drive rate is the same as last season and well above the league average, so a BABIP well above league average is not a shock. Trout's hard hit percentage rose from 33 percent last season to 39 percent this year. To put that in perspective, that rate is akin to Joe Mauer's and Albert Pujols' in their primes but a bit lower that the 43 percent Miguel Cabrera has sported this season. So again, based on Trout's hard hit percentage, a very high BABIP isn't a fluke.
Further supporting an extremely high BABIP is Trout's percentage of infield hits rose this season, speaking toward his speed. Not only that, the number of infield pop-ups dropped a tad.
So while I really, really wanted to conclude that Trout's 2013 BABIP had a significant fluke element, I can't. It's fully supported by every pertinent measure. Does this mean he will sustain it? Probably not, but when he does establish a baseline BABIP, it is going to be among the league leaders. That said, I am going to regress it when I embark on 2014 player projections. I won't be regressing to his mean, but the league mean. On one hand, no one has ever sustained a BABIP where Trout sits along with the power he possesses. On the other hand, nothing has ever happened until it happens the first time.
Before we move on to the next player, something overlooked with Trout is a marked improvement in contact rate and plate patience. That is, his skills have improved as compared to 2012; he's fanning less and walking more. This very well could be a once-in-a-lifetime talent before our eyes.
The second player to be addressed is Jay Bruce. This is strictly for personal, perhaps even selfish, reasons. Back in May, I contributed to rest-of-season rankings for ESPN. I have been working on an algorithm that factors in when certain skills become reliable and incorporated that into my valuation process. Studies suggest contact rate stabilizes rather quickly. That is, if a player begins the season striking out more (or less) than normal, this pattern is likely to maintain for the season. At the time of the rankings, Bruce was fanning at a 31 percent clip, well above his career norm near 24 percent. My projection model set his rest-of-season strikeout rate around 27 percent, a bit higher than normal. That's the first part of the process.
Even though Bruce began the season fanning at a relatively exorbitant rate, he was carrying a BABIP of .366, so his average looked normal, actually better than normal in the .275 range. His early line drive rate was over 30 percent, significantly greater than his career mark of 20 percent. The same research that shows contact rate stabilizes quickly shows it takes the whole season for line drives to stabilize, so the second part of the process was regressing his rest-of-season BABIP to .305, just a little better than his career .297 mark. That is, some recognition was given to what Bruce did in the early going, but probability dictates a return to career performance.
Bruce sets lofty standards?
The final part of the process dealt with power. At the time of the rankings, Bruce had left the yard only three times. Home run rates stabilize somewhere between contact rate and line drive rate, so the expectation was for a small drop in HR rate.
So what happened? Bruce is hitting .267 with 30 HR and 100 RBI, so my rest-of-season mid-May ranking of 218 for ESPN looks pretty silly. Those trolling me on Twitter can claim a win. Or can they?
Let's break down each of the three process elements described above.
Since I crunched the numbers in May, Bruce's strikeout rate has been a little over 26 percent, so while it is still above his career norm, it is a bit better than my engine predicted.
Bruce's BABIP since that time is .320, so it indeed regressed, but not to the extent expected.
For the previous two seasons, Bruce hit a homer every 19.6 plate appearances. Since mid-May it has been once every 18, so the process was incorrect; Bruce's exceeded recent history.
There's still a week left and these percentages may change a little, but the global picture is still the same. As a whole, the process wasn't horrible. Remember, we're dealing with probabilities, not absolutes. With two of the three factors, the process pointed us in the right direction. Bruce continued to whiff at a rate above his career level and his BABIP regressed toward his career mark. The power has improved, but if Bruce doesn't hit another home this season and gets 40 more plate appearances, his PA/HR is exactly equal to his career number. Of course, Bruce is also capable of hitting five more homers by season's end.
In this case, the process under investigation is in-season projections. If someone just looks at Bruce's final 2013 line, they may chide me and insist I overhaul the process. Or just troll me on Twitter.
I'm not so sure an overhaul is required. Could it be tweaked? Sure.
But as Meatloaf tells us, two out of three ain't bad.
Focusing primarily on the science of player valuation and game theory starting in 1997, Todd Zola and Mastersball carved out an important niche in the fantasy industry. In 2006, Todd became the Research Director for fantasybaseball.com, and in 2009, he relaunched Mastersball and is now a managing partner.
Todd competes in Tout Wars and the XFL, and has been a multiple-time league champion in the National Fantasy Baseball Championship. He has been a contributor to the fantasy content at MLB.com and SI.com, is a frequent guest on Sirius/XM and Blog Talk Radio and is an annual speaker at the spring and fall First Pitch Forum symposiums.