World Series time! Enjoy Premium-level access to most features through the end of the Series!
June 5, 2006
Doctoring The Numbers
The Draft, Part 12
If there's one indisputable fact about our draft study, it's that the dynamics of the draft change over time; the calculus of the draft from 1992 to 1999 is much, much different than it was from 1984 to 1991.
Well, there's no reason to think that change suddenly ground to a halt in 1999, and the data from a decade ago may hold little bearing on the decisions that will be made next Tuesday.
On the other hand, it's too early to analyze more recent drafts, and some data is better than none. The question is, can we extrapolate what the trends of the draft might have been based on how teams' drafting tendencies have changed?
With an assist from John Erhardt, here's a chart comparing the number of (signed) high school players vs. college players selected in the first 100 picks:
The chart is a little too spiky for my tastes--let's look at a moving three-year average to smooth out some of the random variation from year to year:
The mid-80s represented the Golden Age for college draft picks, both in terms of quantity and quality--the era of Will Clark and Barry Bonds and Barry Larkin and Mark McGwire. As we've mentioned before, one of the biggest reasons for that bountiful crop was that so many of these players were allowed to slip off to college despite being highly regarded in high school.
The emphasis on drafting collegians quieted down by 1990, and both groups were equally likely to be drafted through about 1995, when high school players suddenly spiked upwards--at around the same time that the return on high school players increased dramatically. But starting with the 2001 draft, teams have gone after college players by an ever-increasing amount. Here are the year-by-year breakdowns since 1995:
Year COL Players HS Players 1995 36 56 1996 35 57 1997 35 55 1998 43 50 1999 41 52 2000 40 52 2001 47 43 2002 49 44 2003 51 39 2004 60 34 2005 56 38More high school players were drafted in the Top 100 than college players for six straight years from 1995 to 2000, but the numbers reversed in 2001 and have not looked back. The 60 college signees in 2004 was even more than the 57 college players signed in 1984, and represents the greatest emphasis on college talent (in the first 100 picks) ever.
The "Moneyball" draft was 2002. More importantly, the book came out in the spring of 2003. It would be naïve to think that the timing was just a coincidence. Several organizations--notably the Red Sox and Blue Jays, but also less-heralded teams like the Diamondbacks and Royals--started focusing overwhelmingly on college talent at that point.
So here's the question: does the sudden emphasis on college talent change the return on draft picks? We know that college hitters returned about 25% more value than high school hitters from 1992 to 1999. But in that eight year span, there were 130 college hitters signed in the top 100 picks, compared to 227 hitters. From 2000 to 2005, there were more college hitters (138) than high school hitters (129) taken. In the 1990s, the 10th-best college hitter was being drafted, on average, after the 17th-best high school hitter. In the 2000s, when the 10th-best college hitter gets selected, more often than not the 10th-best high school hitter is still on the board. That has to mean something, doesn't it?
The answer is more complicated than you think, because such a relationship did not exist between the 1980s and 1990s. More high school players were drafted in the 1990s relative to college players, but the return on investment also increased. We've already recounted some of the reasons for this, but to recap the two biggies:
1. Signing bonuses had reached a level in the 1990s that made it highly unusual for a top high school prospect to bypass the pros in favor of college;
In 1984, ten of the top 100 draft picks did not sign; in 1985, nine did not. I don't have precise data on previous years, but in the early 80s the number of top-100 picks that failed to sign was likely much higher. But in 1986, there were only three unsigned players, which was the start of a trend: the most unsigned players in any draft since is just six.
Which is to say, the trend that has funneled the best high school talent into the pros instead of to the college ranks has not suddenly reversed course. In 2003, for the first time ever, every single player drafted in the first two rounds signed. (The first two players selected who didn't sign? Andrew Miller--possibly the #1 pick this year--and Drew Stubbs, who should go in the top 10.)
In the 1990s, teams could afford to select more high school talent, because the college talent had been weakened by the siphoning off of the best high school talent three years prior. In the 2000s, teams are selecting more college players even though the ranks of collegiate talent have not suddenly swelled with an influx of top high school players who dissed the pros.
2. The level of competition for elite high school talent had increased dramatically as a result of the development of traveling teams, the Area Code Games, etc, which made it considerably easier for scouts to evaluate the best high school players against each other.
The trend towards more competitions pitting the best high school players from around the country has continued, if not accelerated. Travel baseball teams for top players as young as 12 years old are commonplace, and by the time many high schoolers are available in the draft they've been playing against their best peers from around the country for up to six years. Showcase events in the offseason are a tremendous opportunity for scouts to observe the best incoming high school juniors and seniors in the country, and the preseason draft hype for a player is frequently dictated by his performance there.
Meanwhile, the format for the NCAA has changed only slightly with the expansion of the tournament to 64 teams and the advent of "Super-Regionals." The typical college player might pick up a few extra games over the course of his career, generally not a meaningful impact from a scouting perspective. The best collegiate players do get to tour with Team USA in the summer (although some high school players get to play on junior national squads themselves), and wooden bat summer leagues--i.e. the Cape Cod League--are definitely useful for both players and scouts.
But overall, the trends that made high school players easier to scout (relative to college players) in the 1990s vs. the 1980s have continued into the 2000s.
So while there was good reason for teams to load up on high school players in the 1990s (even though high school players were such a bad bet in the 1980s), there has been little reason for teams to load up on college players in the 2000s except for the obvious one: in the 1990s, college hitters (not pitchers) were undervalued in the draft.
Are there any other baseball trends that have emerged over the past 5-10 years which might tilt the balance of draft strategies one way or the other? I can think of one significant one: as recently as ten years ago, if you were a 19-year-old with a howitzer for an arm, you had no safe haven. Go to college and you would be slagged by the likes of Gene Stephenson and Cliff Gustafson; go pro and you had to watch your back for Dallas Green. You were going to throw a lot of pitches no matter who you pitched for.
Today, while college teams sensibly make winning their top priority, every single major league team monitors pitch counts and has implemented pitch limits in the minor leagues. The argument has been made for years that organizations would rather direct the development of an 18-year-old kid than take the risk that the 21-year-old they just signed has a ticking time bomb in his shoulder. It's hard to argue with that logic when Tim Lincecum, who's likely to go at the top of the first round this year and stands 5'11" in cleats, throws 146 pitches in a start for the Washington Huskies.
So chalk up another advantage for high school pitchers over their collegiate brethren.
Here's a chart breaking down draft picks not only by their class, but by position--again, these are 3-year averages:
Since 2000, the number of college hitters taken in the top 100 has increased by over 50%, while the number of college pitchers has increased only slightly. The number of high school pitchers has dropped from their high in the late 1990s, but are still more popular picks than they were twenty years ago. High school hitters, meanwhile, have been trending downward since the late 1980s. In 1989 there were 40 high school hitters signed from the top 100; there were just 37 signed in 2004 and 2005 combined.
Looking at this chart, it's clear that most teams have been doing their homework. College hitters were the most valuable subset of draft picks in both the 1980s and 1990s--as a result, since 2000, teams have become much more aggressive about drafting them.
But it's also clear that in trying to address one inefficiency, teams may be creating others. After a span in the late 1990s where college and high school pitchers were selected with equal frequency, college pitchers have once again become much more popular--by a margin of 88 to 52 in the last three drafts. This, even though both groups returned almost identical value from 1992 to 1999. This would suggest that high school pitchers are now a better investment than college pitchers. Plus, the decline in selections from the ranks of high school hitters has been so steep, and sustained for so long, that they may no longer be the overpriced commodities they once were.
Let's see if we can describe this numerically. From 1992 to 1999, here is the amount of value (relative to expectations) squeezed out of the typical player drafted out of each of the four main player groups.
Group Value COL H 1.282 COL P 0.854 HS P 0.851 HS H 0.791And here's a chart comparing the average number of players selected in each group in the 1992-99 era to the 2001-05 era.
Group Avg # (92-99) Avg # (01-05) Ratio COL H 16.250 24.4 0.666 COL P 25.875 27.8 0.931 HS H 28.375 21.0 1.351 HS P 21.875 18.6 1.176Now, here's the controversial part. In order to get a ballpark measure of how each group should be valued today, we can simply multiply the Value of each group from 1992-99 with the ratio of how the desirability of that group has changed over the last five drafts. I make no claims that this method is mathematically rigorous or even particularly accurate, but it's the best we can do.
Here's what the new Value chart would look like from 2001 to 2005, if these assumptions are accurate:
Group Old Value Ratio New Value HS H 0.791 1.351 1.069 HS P 0.851 1.176 1.001 COL H 1.282 0.666 0.854 COL P 0.854 0.931 0.795The numbers get turned on their head. Now, for perhaps the first time in draft history, a compelling argument can be made that high school players are underrated. According to these admittedly fuzzy numbers, high school players render nearly 25% more value than college players--the exact opposite conclusion we reached from the 1992-99 data.
If we used the draft trends just since Moneyball was published in 2003, the discrepancy is even larger:
Group Old Value Ratio New Value HS H 0.791 1.443 1.141 HS P 0.851 1.262 1.074 COL H 1.282 0.617 0.791 COL P 0.854 0.882 0.753Using this data, over the past three drafts high school hitters were worth approximately 40% more than college players--an advantage which, if not quite as sizeable as the one college players enjoyed in the 1980s, is still incredibly significant.
That's the simplistic method. A more complicated method goes like this:
Between 1992 and 1999, an average of 16.25 college hitters were signed out of the Top 100, or roughly one every six picks. If we assume a random distribution of those hitters in the Top 100, the 10th college hitter selected would be drafted around pick #58.
Between 2003 and 2005, an average of 26.33 college hitters were signed out of the Top 100, meaning the 10th college hitter selected would be drafted around pick #36.
If we make one assumption--that the tenth college hitter selected in one draft is just as good as the tenth college hitter selected in the other--then we can give the 10th college hitter selected between 2003 and 2005 the "expected value" of the 10th college hitter selected between 1992-99. In other words, a college hitter selected with the #36 pick between 2003-05 is comparable to a college hitter selected #58 between 1992-99 (since they were both selected 10th among their cohorts), and so we should expect the the college hitter selected #36 today to only be as valuable as a college hitter selected #58 in the 1990s.
The opposite is true--the 10th high school pitcher selected from 2003-05 would be drafted around pick #55; the same pitcher would have been selected around pick #43 between 1992-99.
From our previous work, we found that the average value of each draft pick drops around 4.5% per pick for the first 38 picks, then around 1.2% per pick after that. Using this information, we can say that the 10th college hitter selected between 2003-05 provides a return on investment that is 28.5% less than the 10th college hitter selected from 1992-99, simply because he was drafted so much higher. The 10th high school pitcher, on the other hand, is 14.5% more valuable than before, because he was drafted considerably later than before.
If we run these numbers for all four groups, and look at how this affects the 5th, 10th, and 15th player in each cohort, here's what we come up with:
Player COL H COL P HS H HS P 5th -38.6% - 9.1% +36.2% +26.9% 10th -28.5% -18.0% +38.0% +14.5% 15th -33.8% - 7.7% +31.0% +23.0% Average -33.6% -11.6% +35.1% +21.5%While our method changed, the results are the same: the value of college players declines precipitously, while high school players are much more valuable than before.
If we take these numbers and combine them with our value matrix from 1992 to 1999, the results look like this:
Group Old Value Ratio New Value HS H 0.791 1.351 1.069 HS P 0.851 1.215 1.034 COL H 1.282 0.664 0.851 COL P 0.854 0.884 0.755The gap between high school and college players has shrunk, but by a very small margin--the advantage enjoyed by high school players ranges between 26% for hitters, and 37% for pitchers.
Any way you look at it, the pendulum has swung too far. Based on the data at hand, we can estimate what the ideal breakdown of draft picks should be that would make all four groups of players equally valuable in a typical year. After accounting for the fact that in a typical draft, eight of the first 100 picks are either JuCo picks or do not sign, here's what the ideal breakdown of the other 92 looks like this:
High School Hitters: 24.6
Or in percentage form:
% College Players: 51.1%
% Pitchers: 48.5%
This isn't particularly groundbreaking: the platonic ideal is about a 50/50 breakdown between college and high school talent, and a 50/50 breakdown between pitchers and hitters. But in recent years, the breakdown has, well, broken down. Over the last three drafts, the distribution was tilted 60/40 towards college players.
Last year, the Oakland Athletics surprised casual observers by snapping up three high school pitchers in a row--Craig Italiano, Jared Lansford (Carney's son), and Vincent Mazzaro--after selecting college hitters with their first two picks. Some people speculated that this was a repudiation of their Moneyball philosophy, an acknowledgment that their earlier methods were not working.
But Moneyball was never about a specific philosophy of valuing production over potential, or statistics over scouts. It was a philosophy of identifying inefficiencies in the market--whatever they are--and exploiting them. As the inefficiencies change, so do the tactics. The sudden change in the A's drafting tactics doesn't invalidate their philosophy; it simply proves that they were doing their homework, and noticed the trends in other teams' draft strategies. When everyone else was zigging, they were zagging. Now that everyone else is zagging…
The A's figured out a long time ago that flexibility and adaptability are the keys to staying ahead of the market. Because if there's one overarching Golden Draft Rule that governs everything else, it is this:
The only rule that isn't subject to change over time is that all the other rules are.