Publication Schedule
We expect to publish two team reviews per week, one each on
Tuesday and Friday. We'll be alternating between leagues, going
more or less in alphabetical order within each league. As new
reviews are posted, we'll activate the links in the following
list:
From November 30 through mid-March, we'll be posting a series
of 30 articles, one for each team, that take a close look at the
performances of each team and its key players in 1999.
Our purpose is to shed light on what happened in 1999 so that
we might gain a better understanding of what's likely to happen
in the future. If a team was down this past year, was it because
they suffered more than their share of injuries? If they were
up, did they get an unusually large number of career years from
their players? If so, is it reasonable to expect this to happen
again next year? Did the team add or subtract talent in their
personnel moves during the year? Or were they just lucky or unlucky?
We believe this type of analysis is most revealing if the actual
results are compared to an objective set of pre-season expectations.
The key word here is "objective". Different people have
different expectations about what will happen in the coming season.
So whose expectations should we use as our baseline?
Certainly not those of general managers and other club officials,
because they're renowned for their springtime optimism. How many
times have you heard the GM of a last place club talk about being
2-3 players away from contending next year? The reality is that
you could add Koufax, Ruth and Bench to some of these teams and
they'd still only finish in the middle of the pack. In fairness,
we can't blame the GMs for their optimism. After all, part of
their job is to create a positive feeling about the team. But
that doesn't mean we have to use their public statements as a
fair indication of what was most likely to happen.
So, rather than rely on pre-season opinions, we're using the
results of the computer simulations that we ran back in March,
1999. Why these? Because:
- our projections included expected runs scored and allowed
by each team, along with their projected totals win-loss record
and place in the standings, giving us more material to work with
than if we relied on other published predictions
- we projected a full set of statistics for over 1400 players,
and many other sources don't include as many players or as many
stats
- our projection methodology takes a lot of factors into account,
and does so quite rigorously. To summarize briefly, we start with
three years of data from both the major and minor league level,
then adjust the player stats for park effects (including minor
league parks), league (DH vs non-DH), competitive level (MLB vs
AAA vs AA), age, and expected 1999 role.
- our simulations used manager profiles for each team that
included starting rotations, relief roles, starting lineups against
LHP and RHP, platoons, and utility roles, making sure to limit
the playing time for guys who were known to be starting the season
on the DL.
- defense matters in our simulation software, and because all
of our players have ratings for range and error rates (along with
ratings for many other skills), clubs that improved their defenses
will help their pitchers and won-loss records accordingly
- our projections are a matter of public record. We published
our projected team standings and win-loss records in an article
before the season started, and we published our player projections
in the form of a Projection Disk for Diamond Mind Baseball, so
nobody can claim we're engaging in revisionist history.
- we developed them, so we don't need to ask anyone for permission
to use them, and that makes life a lot simpler.
Note: If you're interested in learning more about our
approach, you can find a detailed description in The
Diamond Mind Projection System.
We simply want to be able to say that a certain player did
better than expected, about as well as expected, or worse than
expected, and have that statement mean something. Without a set
of projections that are unbiased and reasonably based in fact,
such comparisons would be meaningless.
Content of the Team Reviews
Of course, every baseball season produces its share of team
and player performances that nobody could anticipate, and 1999
was no different. Some players raised their game to a new level,
and time will tell whether this was a fluke or an indication of
things to come. Others had their games fall off a cliff, and they
may or may not get another chance to prove that they can still
play the game.
So, whether you're evaluating the predictions of a baseball
expert or a computer simulation, it would be grossly unfair to
expect anything resembling a perfect match of forecast and actual
results. Sometimes the forecast is reasonable but the actual results
are an anomaly -- like Brady Anderson's 50 homeruns in 1996 --
that will very likely never be repeated. Sometimes the forecast
is wrong because it fails to take into account some vital information
that was available at the time -- that Sammy Sosa took up baseball
relatively late in his youth and might still be improving at an
age when most players are past their peak.
In our team reviews, we'll be presenting detailed comparisions
of three types. Each will begin with a capsule summary of the
team's performance relative to our projections. The second section
presents projected versus actual statistics for at least a dozen
position players, along with our observations about each of these
players. The third section does the same for the most important
pitchers on the team. And we'll wrap up with a few thoughts about
the team's chances for next year.
Capsule Summary
Projected Actual
Runs for 719 665
Runs allowed 843 812
Run margin -124 -147
Wins 70 65
Pythagorean wins 68 65
Placement 5th 5th
In this section, we'll compare our team projections for runs
scored, runs allowed, run margin, wins, pythagorean wins, and
place in the standings. If the term pythagorean wins is foreign
to you, it's a concept developed by Bill James that connects runs
to wins using the following formula:
Runs^^2
Wins = 162 * -----------------------------
Runs^^2 + (Runs allowed)^^2
The notation ^^ means to the power of, so this formula computes
the expected winning percentage by dividing the square of runs
scored by the sum of the squares of runs scored and runs allowed,
then multiplies that percentage by 162 to get the projected win
total. In our capsule summaries, the projected wins figure is
the average number of wins in five simulated seasons, while the
pythagorean wins number is the expected number of wins given the
average runs scored and allowed in those simulated seasons. In
most cases, the pythagorean wins is a slightly better indicator
of the team's true talent level.
Position Player Example
Here's what the McGwire portion of the Cardinals review will
look like. It starts with the player's name, position, and age
(as of July 1, 1999) and is followed by a comparison of projected
and actual performance:
Mark McGwire, 1b, age 35
AB H 2B 3B HR R RBI HP W IW K SB CS AVG OBP SPC OPS RC
Projection StL 521 145 23 0 63 112 131 7 133 22 155 1 0 .278 .429 .685 1.114 161
Prorated StL 517 144 22 0 62 111 130 6 132 21 154 0 0 .279 .429 .681 1.109 159
Actual StL 521 145 21 1 65 118 147 2 133 21 141 0 0 .278 .424 .697 1.120 160
The top line is the projection we made in the spring. The second
line is the projection adjusted to the actual number of plate
appearances he had in 1999. The third line shows his 1999 stats.
You can compare the first two lines to see how much more or less
he played than we anticipated. And you can compare the second
and third rows to see how his performance compared with our expectations.
Note: OPS, which is on-base percentage plus slugging
percentage, is regarded by many analysts as one of the better
measures of overall offensive production. RC is runs created,
another highly-regarded statistic (developed by Bill James) to
measure overall offensive contributions.
Pitcher Example
Here's what the Randy Johnson portion of the DiamondBacks review
will look like. It starts with the player's name, position, and
age (as of July 1, 1999) and is followed by a comparison of projected
and actual performance:
Randy Johnson, Starter, age 35
Tm ERA G GS W L S INN H HR BB K AVG OPS
Projection Ari 3.00 32 32 16 10 0 234 190 22 77 312 .222 .636
Prorated Ari 3.00 36 36 18 11 0 263 214 25 87 351 .222 .636
Actual Ari 2.48 35 35 17 9 0 272 207 30 70 364 .208 .601
The playing time adjustment for pitchers is based on batters
faced. Some pitchers were projected as relievers but were used
as starting pitchers instead. In those cases, their innings went
up but their games played went down. In these cases, we simply
leave the G and GS entries blank in the Prorated line because
it would be meaningless to say that Ron Villone's prorated games
played total was 102.
It's not unusual for starting pitcher to exceed our projected
playing time figures if he stays healthy all year. We've been
projecting starters for 32 starts to allow for the injuries and
fatigue that affect many of them each year. For that reason, we
almost never project anyone to win more than 18 games. If a pitcher
stays healthy, pitches well, and gets decent run support from
his offense, he'll almost always win more games than we project.
OPS is defined in the McGwire example above.
About the Authors
I'm very happy to report that we've assembled a team of first-rate
baseball analysts -- Gary Gillette, Tom Ruane, Sherri Nichols,
and Jon Dunkle -- to evaluate the teams and players and to write
these reviews. I've known all of them for years and have the highest
regard for their baseball knowledge and analytical skills. I'll
be writing six of the reviews myself and serving as editor for
all of the others.
Copyright © 1999. Diamond
Mind, Inc. All rights reserved.
