|
The Principle of Automatic Improvement by David Luciani Published January
17
The Principle of Automatic Improvement
by David Luciani
Published January 17, 1999
An understanding of a phenomenon that I will call "automatic
improvement" is as important in making good projections as any other factor
commonly used. I have never seen this idea discussed elsewhere but it plays a
key part in our projections and I can’t imagine making an accurate forecast
without understanding it.
Automatic improvement is not "real" improvement. Real improvement
is when a player actually becomes better at what they are doing, regardless of
how much everyone else in the league is improving. A player spends the winter at
a batting cage that throws 105 MPH golf balls. It’s likely that there will be
real improvement in the player’s ability to make contact with 90 MPH fastballs
in the subsequent major league season. The player has actually changed their
skill. The idea of "automatic" improvement is understanding that the
quality of the league itself changes and that the sum of all baseball
performance must remain relative to the rest of the league.
A Triple-A hitter is called up to the major leagues. At the moment he’s
called up, he’s facing a player pool of 100% true major leaguers relative to
the pitchers he faced at Triple-A. As quickly as the same night he is recalled,
a transformation of the major league player pool begins to occur. Pitchers that
this batter was facing at Triple-A are recalled by their respective clubs. The
player’s skill, even if not really improving (see our definition of real
improvement above), at least remains constant when one excludes the effects of
age. But the caliber of competition the player faces is being slowly reduced as
new players are called up from the minors. Relative to the competition of the
major leagues, the player’s performance increases as the major league talent
pool is replaced by players that were at or lower on the minor league rung than
this player was when he was first recalled.
Using real statistics, our hypothetical batter is recalled from Triple-A on
the final day of the 1987 season. We’ll also say that our batter was a
Triple-A standout, dominating the pitching at that level. As the 1988 season
begins, pundits everywhere say that he could handle Triple-A pitching but now
that he’s facing major league pitching he won’t be able to hit. As it
turns out, about 7% of the time, the batter is facing a pitcher who was at or
below him on the minor league ladder in 1987. The player should therefore retain
at least 7% of his full minor league skill, excluding the effects of age and the
minor league park/environment.
As we advance to 1989, 16% of the time, he is facing pitchers who were at or
below Triple-A in 1987. The player’s performance back from 1987 better
translates as the player faces competition more in line with what he has already
shown he is capable of hitting against. By 1990, he is facing pitchers 27% of
the time who in 1987 were at or below Triple-A. As we advance to 1998, presuming
our player has managed to remain in the majors, he is facing a pitcher 86% of
the time who was at or below him on the minor league ladder back in 1987.
A player’s minor league skill, therefore, should be expected to gradually
translate more to his major league performance as the talent pool in the majors
is replaced. Even excluding the effect of age, if a player is able to stick
around in the majors long enough, his skill will appear to improve even if it
stays exactly the same because baseball statistics are relational to the rest
of the league. Remove Roger Clemens from the AL last year and give his
innings to the "average Blue Jays pitcher not named Roger Clemens" and
every third American League batter ends up with one fewer strikeout. It's
not the batters are any better in a league without Roger Clemens. It's
that the quality of the pitching in the league actually drops when Roger Clemens
isn't in it!
Mixing automatic improvement with other major factors such as the effect of
age and the effects of "real" improvement means that your projections
will have a much more sound basis, especially as it comes to players in their
early to mid twenties who have been struggling so far in the majors.
Expansion has made statistics ambiguous as the number of new players coming
into the league is far higher than normal levels. The following chart should
give you an idea of what is really happening as far as talent replacement in the
big leagues is concerned. For pitchers, we use batters faced. For batters, we
use plate appearances. If we were to count the number of new players rather than
BFP or PA, the statistics would misrepresent the numbers. Many players are
recalled from the minors only to get 10 at bats or 5 innings:
% of Batters Faced by Pitchers Who Had Never Appeared in the
Major Leagues That Year
1988 7%
1989 6%
1990 7%
1991 9%
1992 7%
1993 10% (expansion year, two new teams)
1994 6% (strike-shortened, no September call-ups)
1995 16% (strike-shortened to 144 games)
1996 7%
1997 10%
1998 10% (expansion year, two new teams)
% of Plate Appearances by Batters Who Had Never Appeared in
the Major Leagues Before That Year
1988 3%
1989 4%
1990 4%
1991 5%
1992 5%
1993 6% (expansion year, two new teams)
1994 2% (strike-shortened, no September call-ups)
1995 6% (strike-shortened to 144 games)
1996 5%
1997 4%
1998 4% (expansion year, two new teams)
The chart reflects a few interesting trends. First, it is clear that new
pitchers are being broken in more quickly than new batters. If such a trend is
accurate, according to the principles of automatic improvement, it means that
hitting should gradually become more dominant as this trend continues. Because
the same batters are facing more inexperienced pitchers, they are facing
"easier" competition as the years go by. Note that this is not a trend
started by expansion. From 1988-92, the new pitchers almost doubled the new
batters.
Also of note is that 1995 saw an incredible influx of new pitchers even
though the season was slightly shortened and there was no expansion that year.
Despite expansion in 1998, the same number of new faces were given playing time
as in 1997, a shift from former expansion ideas that saw experimentation with
new, young players.
The best method to ensure that your projections are accounting for automatic
improvement is simple. Add up all of your projections and be sure that for
every hit you project for a batter, there is a pitcher you project to give up
that hit. Whatever you do, don't pro-rate your statistics so
that the sum of them adds up to last year's totals for the league in question.
The emphasis in baseball continues to shift in favor of hitting and if you use
last year's league totals, you will underestimate the average hitter and
overestimate the average pitcher. Find out what every player's ability is
and then make adjustments based on the sum of your pitching projections and the
sum of your hitting projections. Your projections will not only look good
that way but everyone will wonder what your secret ingredient is.
|
