Baseball statistics measure performance, not ability. Performance obviously results in part from ability, but baseball statistics also reflect randomness, or luck. A player might bat .310 one season and .290 the next without any appreciable decline in ability. In 600 at-bats, the difference between .310 and .290 is 12 hits, one every 50 at-bats or about every two weeks. Clearly chance alone might explain such a small discrepancy.
In statistics, it’s assumed that the larger the sample size, the more reliable the results. Statisticians have tools to measure random variation. These tools estimate with a 95-percent level of confidence that a player who bats .300 in 600 at-bats has a level of ability between .263 and .337, while a player who bats .300 in 6,000 at-bats has a level of ability between .288 and .312. In other words, the role of chance shrinks considerably as the sample size grows.
Besides measuring past performance, baseball statistics also have predictive value. Based on their track records, Jeff Bagwell stands a better chance than J.T. Snow of batting .300 or hitting 40 home runs. Their career performances indicate different levels of ability beyond mere randomness. How precisely do baseball statistics tell the future? A few principles are evident.
First, it should be easier to project performance for established than unestablished players, since veterans offer a greater sample size through which to isolate random variation, gauge ability and devise forecasts.
Second, a player’s level of ability, and hence his performance, should fluctuate with age, injuries, the adoption of superior equipment and improvements in fitness, skills and experience.
Third, and perhaps most important, the role of luck in a season is substantial, and the direction of that luck is unpredictable.
Consider Bagwell and Snow. With a .303 batting average in 5,949 at-bats, Bagwell’s level of ability likely falls between .291 and .315. But in a season of 600 at-bats, a player with Bagwell’s statistics might bat between .255 and .315. Snow’s .264 batting average in 4,132 at-bats yields a likely level of ability between .251 and .278, but in 600 at-bats this range grows to between .216 and .314. The chances of Bagwell or Snow batting outside these ranges are no greater than 5 percent.
Such vast margins of error might appear to make baseball statistics meaningless. After all, projecting a 100-point range for a player’s batting average isn’t very useful. Bear in mind, though, that every possible batting average along that 100-point range isn’t equally likely to occur. While Bagwell is 95-percent likely to bat between .255 and .352, the best bet is that he’s going to bat at the middle of that range.
STATS, Inc., in its Major League Handbook, has been making projections based on past performance for the last 13 years. STATS uses a method originally developed by Bill James, based on “a network of common-sense formulas – quasi-scientific, I suppose you could call them – which essentially assume that a player will continue to perform in a manner consistent with his past performance, modified somewhat by age and experience.”
STATS and James have refined their approach over the last decade. They adjust for park effects as well as age and experience. The results are surprisingly accurate. In a survey using nine seasons’ worth of data, STATS’ latest model yielded projections nearly identical to actual performance 62 percent of the time, extremely similar 94 percent of the time, and somewhat or very similar almost 100 percent of the time.
For purposes of the study, STATS adjusted the projections to reflect actual playing time, which makes the results more accurate than those published annually in the Major League Handbook. STATS and James note that the toughest part of projections is guessing who will play and how much. Injuries are impossible to predict, and playing time isn’t a certainty for many players. Without adjusting for playing time, STATS’ projections still result in very similar forecasts between 50 percent and 65 percent of the time.
Ultimately, James writes, “we just started doing this for the fun of it. We didn’t make any claims to accuracy to begin with, and we’re not going to make them now. Speaking for myself, I enjoy being surprised once in awhile. Establishing the predictable establishes the parameters of our future surprises.”
With all these caveats, listed below are the projections from the 2002 Major League Handbook for the Astros’ probable starting line-up:
Player Avg OBP Slg ------------------------------- Brad Ausmus .258 .331 .352 Jeff Bagwell .286 .411 .562 Craig Biggio .278 .361 .419 Morgan Ensberg .306 .389 .589 Adam Everett .248 .327 .350 Daryle Ward .294 .346 .575 Richard Hidalgo .279 .356 .527 Lance Berkman .315 .417 .579
Ensberg’s numbers are not described by STATS as projections, but rather as Major League Equivalencies, an interpretation of his 2001 minor-league performance into major-league statistics, adjusted for park effects. James has long preached that minor-league statistics have predictive value for major-league performance. Simply put, “players who hit in the minor leagues are going to tend to hit in the majors, only not quite as much.”
Ron Shandler, in his annual Baseball Forecaster, also publishes MLEs. Here is a comparison of actual performance and MLEs for Ensberg and Everett the last two seasons:
---Actual---- ----STATS---- --Shandler--- Avg OBP Slg Avg OBP Slg Avg OBP Slg Ensberg----------------------------------------------- 2000 .300 .416 .545 .282 .369 .503 .263 .360 .484 2001 .310 .397 .592 .306 .389 .589 .285 .360 .540 Everett----------------------------------------------- 2000 .245 .363 .342 .233 .331 .318 .232 .327 .329 2001 .249 .330 .365 .246 .305 .358 .234 .287 .340
STATS’ MLEs appear overly generous. They imply that Ensberg’s performance in New Orleans, adjusted for the effects of the extremely pro-pitcher Zephyr Field, is almost equivalent to the same numbers in the majors. Shandler’s more skeptical MLEs still credit Ensberg with a fine season in 2001 even by major-league standards. Both STATS and Shandler suggest that Everett needs to improve markedly to maintain a major-league job.
Here’s a comparison of STATS and Shandler’s projections and MLEs for the 2002 Astros:
----STATS---- --Shandler--- Player Avg OBP Slg Avg OBP Slg ----------------------------------------------- Brad Ausmus .258 .331 .352 .268 .331 .382 Jeff Bagwell .286 .411 .562 .298 .413 .588 Craig Biggio .278 .361 .419 .273 .355 .427 Morgan Ensberg .306 .389 .589 .285 .360 .540 Adam Everett .248 .327 .350 .234 .287 .340 Daryle Ward .294 .346 .575 .270 .322 .483 Richard Hidalgo .279 .356 .527 .278 .348 .508 Lance Berkman .315 .417 .579 .313 .404 .615
STATS and Shandler’s forecasts disagree considerably on Daryle Ward’s outlook, but they are roughly in concurrence about the rest of the Astros. Neither STATS nor Shandler project that Craig Biggio will match last year’s performance.What about MLEs for three other players in the Astros system, Keith Ginter, Chris Truby and Jason Lane?
---Actual---- ----STATS---- --Shandler--- Avg OBP Slg Avg OBP Slg Avg OBP Slg Ginter------------------------------------------------ 2000 .333 .457 .580 .314 .393 .532 .308 .400 .545 2001 .269 .380 .464 .266 .349 .457 .251 .326 .433 Truby------------------------------------------------- 2000 .284 .318 .369 .270 .307 .335 .261 .299 .337 2001 .312 .365 .539 .307 .352 .530 .281 .321 .482 Lane-------------------------------------------------- 2001 .316 .407 .608 .306 .363 .586 .284 .343 .545
Truby serves as an example of how minor-league and major-league performances can vary drastically. After playing himself out of a job in Houston in 2001, he had a nice season at New Orleans. Few fans would expect Truby to be able to replicate that play with the Astros next season, however.An aside about STATS’ projections: in the 1991 Major League Handbook, STATS predicted that a 23-year-old minor-leaguer, given the opportunity to play, would bat .318, which happened to be the highest batting average forecasted for any player in the National League. STATS didn’t overtly mean to imply that a rookie would win the batting title, but a lot of readers took the projections that way and ridiculed STATS as a consequence.
That minor-leaguer got his chance, and ended up winning the Rookie of the Year award in a landslide, receiving 23 first-place votes to Orlando Merced’s one. He didn’t win the batting title, but Bagwell did post a .387 OBP and .457 slugging average, not terribly different than the .400 and .436 projected by STATS. Not a bad guess about a player who had zero major-league experience.
Great Leap ForwardA lot of noise has been made about how the Cardinals upgraded their offense with Tino Martinez and the Cubs improved their line-up with Moises Alou.
Alou is a windfall for the Cubs. Consider Alou’s 2001 statistics versus those of Cubs left fielders as a group last season:
Player Avg OBP Slg ---------------------------- Moises Alou .331 .396 .554 Left Fielders .277 .343 .475
Assuming Alou stays healthy, the Cubs are gaining 132 points of OPS from this move. Before declaring that the Cubs will see their line-up’s productivity grow by a similar margin, however, consider these numbers:
Sammy Sosa Avg OBP Slg ---------------------------- 1998-2000 .305 .383 .639 2001 .328 .437 .737
Sosa’s 2001 performance was one of the great seasons in baseball history. It was also 152 points of OPS better than his established level of performance as measured by his three previous seasons, which were by no means shabby. If Sosa returns to earth a bit, it could eat up much of the improvement by Alou on the other side of the outfield.At least Alou is an improvement, however. There is no reason to believe that Tino Martinez is for the Cardinals. Consider these numbers:
Player Avg OBP Slg ---------------------------- Tino Martinez .280 .329 .501 First Basemen .232 .339 .489
Even with Mark McGwire batting .188 in a majority of the Cardinals’ at-bats at first base, Cardinals first basemen as a group still compiled almost exactly the same OPS as Martinez. Here’s the bigger reason for Cardinals fans to worry:
Tino Martinez Avg OBP Slg ---------------------------- 1998-2000 .267 .341 .461 2001 .280 .329 .501
Martinez had his best season in four years in 2001. He might match that performance again, but the odds are against it. If Martinez fails to repeat his 2001 season, then the Cardinals have paid $7 million per year for a downgrade at first base.But at least they made a big-time move, right?
Contraction Inaction
This off-season’s Sitzkrieg is over. Commissioner Bud Selig finally admitted the obvious last week by announcing that Major League Baseball would not dissolve two teams, most likely the Twins and the Expos, before the start of the 2002 season. Thanks to Selig for giving everybody involved with those two organizations and other contraction candidates plenty of advance notice of what their plans should be for the next eight months.Selig still insists that the owners are adamant about dropping two franchises, probably before the 2003 season. This appears unlikely, however. Consider the following scenario. In the near future, Jeffrey Loria, current owner of the Expos, will sell the team to Major League Baseball for $120 million. Loria will then buy the Marlins for $159 million from John Henry, who has already been approved to acquire the Red Sox for $660 million pending divestiture of the Marlins.
When next off-season rolls around, Major League Baseball, on behalf of the owners, will still have possession of the Expos, an asset for which it paid $120 million. Not only will the owners be $120 million in the hole for the Expos, but there will probably be ownership groups in a number of places, including Washington or Northern Virginia, willing to pay at least two or three times that much to buy the team.
Almost seven years ago, the ownership groups for the Diamondbacks and Devil Rays paid more than $130 million each for the privilege of building new franchises from scratch, and they had to wait three years before playing a game. A purchaser of the Expos would obtain an existing Major League roster, complete with young stars like Vladimir Guerrero, Jose Vidro and Javier Vazquez, with a functioning farm system, and would be able to start play in 2003.
And while the Expos may be worth only $120 million in Montreal, an ownership group from the nation’s capital would pay what the franchise is worth in that highly lucrative market, where discussion has already arisen of renovating RFK Stadium for temporary play and constructing a new ballpark for the long term.
The sale would mean millions of dollars to each existing owner, even after subtracting what Major League Baseball will pay for the Expos. Are the owners so gung-ho on contraction that they are willing to forego that kind of money? Orioles owner Peter Angelos, in nearby Baltimore, might extract a heftier fee in exchange for the move, but the other 29 existing owners could give Angelos an extra several millions of dollars and still make off like bandits.
One wonders whether this is part of the entire contraction ploy: allowing Loria simply to move the Expos to Washington himself would enrich only Loria and set a precedent Major League Baseball wants to avoid. Permitting Loria to swap the Expos for the Marlins in exchange for letting all of the owners have a piece of the relocation pie seems like a win-win situation.
It also avoids the imposition of a “relocation fee,” something that might make Congress examine more closely Major League Baseball’s anti-trust exemption. Instead, the owners can just claim they sold the Expos to new ownership at the franchise’s fair market value. Its fair market value in the nation’s capital rather than Montreal, that is. Throw in a possible new ballpark in Minnesota, and the economic viability of the Twins and the Expos, in the eyes of Major League Baseball, might not seem so dire next winter.