Leveling the Playing Field
The Rule 4 draft is, without question, one of the most important events of the year for Major League teams. One great draft can change the future of a franchise. The draft gives teams an opportunity to acquire young, talented players for a relatively small financial commitment. If one of them reaches the bigs, and becomes even an average player, you’ll garner yourself a ton of value over that player’s first six years.
Naturally, then, the draft, and studying the amateur players, is a major part of each organization’s yearly workload. Consider this response from Chris Long, Padres’ Senior Quantitative Analyst, in an interview with us last year:
What's so amazing about the baseball draft, and I'm sure the draft in other sports, is the sheer number of players to consider. Different ages, sizes, polish, playing environments, growth potentials, levels of competition faced, ability components, injury tendencies, and it goes on. Then there's the information you get from the scouts. Which scouts are better? Are they looking at the right players, in the right way, the right number of times? What's the best way to integrate all of the information you have? Overlaying all of this are considerations of finance, utility, need, risk and the poker game of the actual draft. Draft the right player and he could be worth $50 or even $100 million in value to your club (see Pujols). Draft the wrong players and you'll waste millions and negatively impact your club for years. It's an extremely difficult, messy, noisy, and thoroughly insane problem to work on. It's beautiful.
We all know about scouting. It's crucial to the game, especially in college and high school, and it isn't going anywhere. But a more unexplored area (at least on the 'net), and perhaps an equally important one, is the thorough analysis of college statistics. Many times, people will bring up what Chris brought up in the above passage, saying there are too many factors to consider, too much noise in the data. There's varying levels of competition, parks, player aging, limited sample sizes, switching from aluminum bats to wood, etc. It goes on and on.
They are, of course, right on the money. Looking at the raw stats of two college players is probably a hapless endeavor. Let's look at a quick, made-up example:
Player A: .300/.480/.680
They are somewhat close, but if that's all we know about each player, we’ll probably go with Player A every time. But, let's say Player B played against the third-toughest opponents in Division 1 and also played in a big pitcher's park. Player A played in a small conference, against relatively weak competition, and a great hitter's park. Now who are you goin' with? And not to mention, this is a simplified example, which leaves out many significant factors. But it just serves as a reminder that the numbers, alone, are just numbers; they have relatively little utility in sorting out baseball players on the college level.
Anyway, as you can see, the reservations people have about college stats are real. However, there's no reason why we can't try to make some adjustments, and make some sense of the madness.
We've spent the last four months importing and adjusting collegiate baseball statistics in an attempt to neutralize the numbers to allow for cross-conference comparisons. To do this, we've discovered that Boyd's World is an invaluable tool. He gets much of the credit for accumulating a lot of the data and making it available online.
Now, our methods were actually pretty simple. We're judging the players in our system on a few things that we feel are a solid scope for the offensive skills necessary to succeed in professional baseball. They include:
All of the above are pretty self-explanatory, especially with the Wins Above Replacement explosion that happened in November and December of 2008 around the sabermetric blogosphere. However, the wOBA formula we used did not include stolen bases. Honestly, it wasn't for any particular reason, we just happened to grab the one copy of the formula that did not include it.
As for speed score, it's measuring "baseball speed," or, at least, that's the intended goal. It's actually a fairly generic speed score that is not much unlike the one Bill James used in his earlier works.
But, what do we take into account when adjusting these numbers? For us, it was park factors and level of competition faced. Those two components can vary from team-to-team in such a dramatic fashion that you'd initially swear they aren't right. For instance, Air Force had a 4-year park factor from 2005-2008 of 145. Conversely, a school like Longwood University had a park factor over that time of just 72. With such drastic discrepancies, it was important to address this. Again, drawing from Boyd's World, we have multiple-year park factors. He lists two for each team, one being a PF and one being TPF – or Park Factor and Total Park Factor. The former is just rating that team's home park, while the latter is rating all of the parks that team played in over the course of time it was tracked. So, Air Force's 145 park factor is just their home park. Playing in the Mountain West, they frequent some of the most hitter friendly parks in collegiate baseball, and their Total Park Factor was 128 from 2005-08. Basically, over those years, Air Force's team played in environments that were 28 percent more offense-friendly than a neutral ballpark, which would have a rating of 100.
To neutralize for park factors, we take the wOBA for each hitter, and simply run it through this: wOBA*square root(100/Total Park Factor). This nets us a Park-Adjusted wOBA (PAwOBA).
But that's just the first part of the components to neutralize. You also have to take into account the competition these numbers are being tested against. As mentioned previously, two stat lines, unadjusted, are not equal. Thankfully, Boyd's World comes through again with his Strength of Schedule ratings. To neutralize this, we do pretty much the same from above.
PAwOBA*square root(Strength of Schedule/100)
This gets us a wOBA for the players that are now both park and competition adjusted. We do this for IsoP's as well, using the same methods just substituting IsoP for the wOBA's. And before we jump straight to the table (even though this is going on long enough), we'd like to give a brief introduction to our "Score" category. We don't have a catchy name for it yet (although we're open to all suggestions), but what it encompasses is all of the categories that we're tracking. It weights the adjusted wOBA's, adjusted IsoP's, K and BB% and throws in our speed score, as well. But, we've rambled enough. On to the 2008 stats for the first five college bats taken in the 2008 Rule 4 draft:
The above are nothing more than just the 2008 numbers for the first five college bats taken last June. They are not meant to be a predictor of talent moving from aluminum to wood bats. Instead, it's just, at the moment, adjusting to see who had the best statistical seasons when you account for who they were playing and where. When the 2009 draft comes around, we'll have a better tool to judge player performance than just the raw stats, and hopefully it will shed some light onto who the top prospects are.
Also, don't forget that we haven't considered positional values or defense. A player's position is very important at this level. Players that start on the left of the spectrum (1b, left field, right field) have to hit a ton to make it in the bigs. Most great prospects start on the right side of the spectrum as amateurs and gradually shift to the left as they age, provided that their bats can play at those less-demanding defensive positions.
Earlier in the article, it was mentioned that this type of stuff has been somewhat unexplored on the Internet. While that may be the case, there's certainly been plenty of research into the area: