Let me start this out by affirming that projections are pretty wacky, in theory and practice. Particularly in baseball, where you have 30 teams in 2 leagues and 6 divisions and 162 games of sheer madness, with trades, call-ups, DL stints, fried chicken and beer popping up over the course of 6 gruesome months. You can calculate and pythagorize all you want, stuff happens that nobody can see coming and that’s that.
For the past couple of years, I’ve always enjoyed recording season projections when they are out, so that I can drag them out at the end of the season and compare. I use these projections quite profusely as part of my Fantasy draft strategy, so they are quite important to me. They are also important to many others, for all kinds of fans. This year, I thought it would be fun to share the “before and after” results with my readers, in the event you might be sitting at a bar or at your local Laundromat and a stranger should ask “hey, how about them Cards…did anyone think they had a chance to do so well in the NL Central?” Now you can answer that question.
The folks at Baseball Prospectus are making some decent money compiling and providing statistical data and analysis for use by chumps like me as well as for very important folks in the baseball business. They certainly do know what they are doing. They have ultimato respect from everyone who reads their work, whether the information is used casually or used professionally.
I purchase their handbook on an annual basis, and enjoy their work immensely. I am not a subscriber, as this is too expensive for my budget (I’m not being paid to be a Baseball Enthusiast) but if you have the cash flow, you really should consider subscribing to their exclusive content, as well as shelling out a couple of bucks for the annual handbook.
The purpose of this is not to say “look, they were way off on this” but to compare side-by-side what the projections were, and to identify some criteria by which to analyze the comparison at a very high level, and then to draw some sort of conclusion that is less than obtuse and lightly informative. There is no “right” or “wrong” in projections (predictions are a different story…); there is only the data, the distribution, and how you analyze it and use it for whatever you need it for.
The “actual” performance data presented here was compiled from Baseball-Reference. Here are the key data points I studied and compared from both sources:
Wpct (Win percentage)
RS (Runs scored)
RA (Runs allowed)
Rdiff/G (Run differential = Runs allowed – Runs scored per Game, average over 162 games)
Teams are listed in ‘ranking’ by WPct as reported by Baseball Prospectus as well as Baseball-Reference.
Focusing on the ‘slot positions’ for each team within their division is, in my opinion, a negligible analysis point. Any stat head will probably agree with me that comparing the projected ‘rank’ of two teams in their division with their final ‘rank’ is somewhat interdependent of discrete performance statistics, such as Win-Loss percentage (or, really, Win percentage). Aside from the obvious situation of a team projected to win their division that eventually comes in last in their division (or vice versa), there’s not much to discuss in regards to comparison aside from what you can clearly see…at least not by me and not in this review. I chose instead to gauge “how did they do?” by comparing WPct by utilizing a figure that I’m sure I didn’t invent but that I will call XPctDiff, or “Expectation Percentage Differential.”
Is that sexy or what?!?
XPctDiff is found by subtracting the projected WPct of a team from the actual WPct, you will get a positive or negative number based on if they did better (positive) or worse (negative) than they were projected to. Remember, final ranking is negligible!! Then you can look at how close or how far off each team’s projection was and in what direction, if you like. The smaller delta (percentage difference) defines a projection that was more spot on, the larger delta is one that was spotty. You can click on the images to see the data better in your browser window.
I will also briefly review the Rdiff/G value, in terms of improvement (positive) and deficiency (negative) over projection, for leaders and blowers in each league.
Let’s start with the National League:One thing is obvious: Nobody really can tell how to call the NL West (remember last year’s Padres?!?). This year’s Diamondbacks seemed to be last year’s Padres, with the most noticeable difference in the ranking, and also the highest XPctDiff in the entire league.
In the NL East, the Phillies (positive) and the Marlins (negative) had the largest XPctDiff, tied at .081 each. In the case of the Phillies, the difference had no bearing on their ranking; I know I said I wasn’t going to focus on ranking comparisons, but I had to point this out in order to lend some validity as to why I say “final ranking is negligible. The closest call by BP in this division was on the Braves, with an XPctDiff of .012 positive value. Among all 30 teams, this was their second most accurate projection. In regards to Rdiff/G the Phillies ended up with an 0.6 improvement over projection, the second largest improvement over this projected figure in the National League…that’s a testament to that pitching staff of theirs, as the fewer RA made the biggest difference (they scored 30 fewer runs than projected). The Marlins, on the other hand, had an 0.7 deficiency over projection, with a shift in more RA than RS yielding that value; their Rdiff/G decline was tied with the Giants for the largest deficiency over projection in the National League, who (like the Marlins), allowed more runs than they scored against projection.
The most accurate WPct projection among all 30 teams was the Pirates in the NL Central. At an XPctDiff of a scant .006, their WPct was only one game off. The Brewers were the largest value, .068 better (positive) than projected. As we enter the World Series, I should also note that the Cardinals performed .025 better than projected, a modest value but it shows you in one respect how a good run in September can benefit a team!
The NL West shows a wild finish, 2 teams (the Diamondbacks at .111 better – but still not the largest XPctDiff among all 30 teams- and the Rockies at .074 worse (negative) than projected) had abnormally large XPctDiff in the National League. BP was most accurate with the Dodgers, who finished .022 worse than projected. The Diamondbacks finished with the largest improvement over projection in terms of Rdiff/G with an 0.7 value.
And now…the American League:Here we have a beast similar to the NL West, where the AL Central seems to be the hardest division to nail down from a projection standpoint. It sheds a different light on the current fan reaction to the so-called “Red Sox Collapse” and the opposite of this reaction by fans of the Twins in regard to their performance this season, suffering even greater than the Royals were projected to.
In the AL East, the most accurate projection was undeniably the easiest call to make…for several years, the Yankees have been the easiest team in terms of performance projection, their XPctDiff was .025 better than projected. By comparison, the Orioles, who started very strong this year, finished with a rather large .074 XPctDiff. The Orioles also had the second largest deficiency in the American League over projection for Rdiff/G with -0.9.
Who had the largest Rdiff/G deficiency in the American League? We move to the AL Central for this answer, where as previously noted, was the wildest of the 3 divisions. The Twins quietly collapsed (where the Red Sox were so damn noisy about it), falling to an MLB highest .123 XPctDiff and an MLB highest 1.0 Rdiff/G deficiency over projection. OUCH. The most accurate WPct projection was the White Sox, who posted a League-lowest .018 XPctDiff value in winning only 3 games less than projected.
A few years from now, we might see a day where the AL West is as easy a division to project as the NL Central seems to be; as long as the Rangers keep playing like the Rangers, that is. They blew off the doors this year (much to Lance Berkman’s chagrin, but he’s so classy he admits his mistake after the fact), finishing better than expected, with an XPctDiff of .068, tied with the Angels for the largest value (the Angels also finished with a positive XPctDiff value). Also of worthy note, and what Berkman needs to know, is their Rdiff/G improved 0.9 over projection; like the Phillies they allowed far fewer runs than projected. The Mariners were BP’s most accurate projection in this division, with an XPctDiff of .024, only 4 wins less than projected.
At an overall high level, in light of the results of Postseason so far, the BP projections really only ‘missed’ with the NL West and the AL Central. Those of us who reviewed their data before the season began can only say we were surprised by the rise of the Diamondbacks and the fall of the Twins.
That being said, Twins fans are great people…they aren’t complaining about this very much.
Enjoy the World Series; I’ll be back soon to review the MLB Network NL/AL Predictions! I’ve dusted off the DVDs I have of the 2 specials they aired in late March, and will review their calls and what really happened. This review will be quite different; less statistics, of course…but they made predictions, which are way different than projections.
Thanks mega thanks again, to the folks at Baseball Prospectus (as well as Baseball Reference) for everything they do to give geeks like me something to write about. It’s all good.
If you enjoy my work, I encourage you to spread the word via Twitter
(I am @yoshiki89), and also please leave a comment!