Criteria for baseball rankings
This is a discussion of the criteria used in Don Davis's
rankings of the top baseball players of the past 62 years.
Refer to Baseball Rankings for the
introduction,
and to Batters and Pitchers
for the rankings based on these criteria.
The seventeen criteria for batters are as follows. In all except the
Hall of Fame and Golden Glove, a player's number of points for that
criterion
is, on a 0 to 10 point basis, his number of points as a fraction from
the lowest scorer to the highest.
- MVP, weight 4. Each time a player wins the MVP, he
receives
12, 11,
or 10 points depending on whether the ratio of his number of votes to that
of the second place finisher is greater than 1.6 (12 points), between 7/6
and 1.6 (11 points), or less than 7/6. Each time a player comes in second
in the MVP voting, he receives 6, 7, or 8 points, according to the ratios
just described. Each time a player comes in third in the MVP voting, he
gets 5 points if he got at least half the maximum possible number of votes,
and 4 otherwise. Each time a player comes in 4th or 5th, he gets 3 points.
Each time a player comes in 6th through 10th, he gets 2 points.
Each time a player gets at least one vote for the MVP but does not placein the top
ten, he gets 0.5 points. These points are totalled for each player.
- WAR, life, weight 4. This "Wins against replacement" is
a composite statistic taken from baseball-reference.com.
This is the additional number of wins they contribute to their team
compared to a marginal replacement player.
I total this for each year of a player's career.
- Best 5 WAR, weight 2. Again taken from baseball-reference.com,
this is the total of the player's best 5 seasons of WAR values.
- Top 5's in WAR, weight 3. Each time that a batter finishes in
the top
5 in their league in WAR, they receive points, 5 for 1st, down to 1 for
5th. These are conveniently tabulated on the player's page on
baseball-reference.com.
- Top 5's in Batting Runs and Runs Created, weight 1.5.
Similar to the previous Top 5's description, except this one uses
their top 5 finishes in their league in both Adjusted Batting Runs
and Runs Created, again taken from baseball-reference.com.
If a player is highest in his league in both categories during the
season, he will get 10 points here.
- Major All Star, weight 1.5. At the end of each season Sporting
News and/or
the Associated Press used to name the best player at each position in the major leagues.
I think this (i.e. number of times named to this team) is a good statistic,
and I would give it a higher weight if I had a consistent measure of it.
I have the AP team for each year since 1982 (except 1997 and 2003-9) and
the Sporting News
team for years prior to that. Unfortunately, during quite a few of those years, TSN
named All Star teams for each league, and not, as far as I know, one for the
majors. For those years, I selected the league all star who did better in
his league's MVP voting.
If a player is named best Designated Hitter,
I give 0.7 points, since he is only competing against one league. I give more
than the 0.5 that logic might dictate, to reward the player as The Best.
For 1997, I guessed the Major League team based on MVP votes, and
for 2003-13 I used various composite statistics to guess it.
- All Star, weight 1. This is the number of times named to the
league All Star team.
This does not distinguish whether the player was named to the first team by the fans
or selected for the squad by the manager. I don't think it is a great statistic,
but it measures something about the player's name recognition. It certainly
rewards longevity. Hank Aaron is first with 21 and Robin Yount last with 3, of my
rated players. Yount's poor performance in this criterion is surprising,
considering
his two MVP selections and first-time Hall of Fame selection.
- Golden Glove, weight 2. Here I just count the number of times
a player is
given his league's Golden Glove award at his position. This criterion is a
poll of players,
based on fielding prowess. This one is a slight deviation from my policy of
rating players linearly between the lowest and highest point-getters. I think
it is significantly better to win one such award than zero, since the one award
suggests that the player is probably close to winning on other occasions.
So I give 2 points for 1 selection and rate players with at least one selection linearly
between 2 points for 1 and 10 points for 16, which is the number of times Brooks
Robinson received this award. Despite its fairly low weight of 2, this criterion
is a huge differentiator. The reason is that most other criteria are highly
correlated, so that the differences between the scores of comparable people will
usually be small, but Golden Glove is not correlated with hitting at all.
The GG was just instituted in 1957. For prior years, I guess
the winner based on reputation and Fielding Runs according to Palmer and Thorn.
- Lifetime RRBIs, weight 0.5. This is the sum of a player's runs
and RBIs for their
career. Palmer and Thorn dismiss these statistics as meaningless to a player's
contribution because it is so dependent on teammates. That is why I gave it such a
low weight. It certainly rewards longevity.
- RRBI, best 5, weight 0.5. This is the player's best 5 years of
runs and RBIs.
However, this is weighted by the average runs per game in the player's league
during those years. This compensates for the higher number of runs scored in recent
years compared, say, to the late 1960's.
- Lifetime OPS+, weight 1.5. This is a combination of the
player's
On Base
Average and
Slugging Percent,
adjusted to incorporate the league average, averaged over the
player's career.
It is taken from baseball-reference.com.
- Hall of Fame, weight 2. I would weight this one higher if I
knew
how to treat
current players fairly. For players elected during their first year of eligibility,
I consider the ratio of the number of votes the player received divided by the
number of votes needed to be elected. This would be 1.33 for a player who got
every voter's vote, which has never happened. For rated batters, the highest was
Cal Ripken with 1.31. For players elected the first time eligible, I give them
their ratio minus .31 and then multiplied by 10. Thus a person who just
barely made it on their first try
would receive 6.9. Players who made it, but not on their first try, receive from
4 to 6.6 points. Six points is for making it on the 4th try, 5 for making it on the 8th,
and 4 for being selected by the Veterans' Committee. Players who haven't made it
get roughly their maximum number of votes divided by 100. For players not yet voted upon,
I make a conservative estimate of what I think might happen. It is hard to guess,
when you consider that Dick Allen got at most 89 votes (with about 350 needed to elect),
Dale Murphy at most 116, and Keith Hernandez at most 52. These three players did
quite well in most of my categories. Because of this, many current players who
are not superstars with huge name recognition only received 1 or 2 points.
Most current or recently-retired players will probably end up doing
much better than my estimates, except for the uncertainty regarding
how voters will respond to steroid-use allegations. In recent years,
people like Ryne Sandberg and Eddie Murray have made a big jump
in my ratings when they received a strong first-time election to the
Hall of Fame.
- Win Shares, total, weight 2. The next three criteria are
based on Bill James' all-purpose statistic.
A player's win shares during a season are
a complicated combination of various batting, fielding, and baserunning
statistics, and are described in detail in James' book "Win Shares."
- Win Shares, best 4 years, weight 2. In James' 2000
rankings of players at each position, he includes best 3 years of
win shares and best 5 consecutive years. Mine is a compromise
between these two.
- Win Shares, top 5s, weight 2. For each year, I take
the top five non-pitchers in the majors in Win Shares.
I give 7 pts for each time
first, 5 for each 2nd, down to 2 for each 5th.
- Defensive wins, weight .5. This
complex fielding rating is taken from
baseball-reference.com.
It is a comparison with an average player, and so can be negative,
which it is for many of our sluggers. I added this statistic in 2007
because I felt that fielding was underrepresented in my ratings.
- SB, lifetime, weight, .5.
This is merely the lifetime total number of stolen bases.
- Post-season, weight 1.5 This criterion was added in 2013.
Although it is important, its fairly low weight is due to its unfairness
to people on bad teams. I take into account, postseason games played,
postseason runs and RBIs, postseason OPS, World Series won, and World Series MVPs. I combine them in a non-scientific way.
Now we describe the criteria for pitchers.
- Runs, total, weight 3. This is Adjusted Pitching Runs,
from baseball-reference.com.
It is essentially the difference between
the average league ERA and that of the pitcher, multiplied by the number of innings
pitched divided by 9. Thus it is the number of runs saved over the course of the
season compared to an average pitcher.
There is an adjustment for home
ballpark; i.e. a compensation is made for a pitcher whose home park
is a hitter's park.
- Runs, best 5, weight 3. The pitcher's five best seasons of
Adjusted Pitching Runs are totaled.
This is a measure of Peak Value, as opposed to Career Value.
Since this is a
comparison with an
average pitcher of that season, no compensation needs to be made for eras
of high or low ERA.
- Runs, best 2, weight 1. This is really a Peak Value, just the
best two years.
Pedro Martinez's years in 1999 and 2000 and Clemens in
1997 were the best by anyone in this
era. Bob Gibson's 1.12 ERA in 1968 and Sandy Koufax's three years of
1,2,2 in the MVP
and twice named AP Athlete of the Year did not fare so well because the average ERA
was very low during that era.
- Cy Young, weight 3.5. I give 6 points for each time the player
wins his league's
Cy Young, 5 for each second, down to 1 for each sixth place finish. For years prior to
1970, when the current voting methods were adopted, I use a combination of the
existing Cy Young vote, together with MVP votes for pitchers.
- Top 5's, weight 3. I give 5 points for each first, down to 1
point for each fifth,
in the pitcher's league each year in each of six categories. These are strikeouts,
adjusted ERA+, Walks and Hits per Inning Pitched, adjusted pitching runs,
Wins against Replacement for pitchers,
and Wins above Replacement for players. In the latter, pitchers are
compared with
batters, as well as other pitchers. All but the
first of these are highly correlated. This is a combination of peak and
career
statistics. You have to have a great year to get a first in these, and you have to have
lots of great years to come out tops in this category, as Clemens does by a large margin.
The ratings are taken from baseball-reference.com
- MVP, weight 2. I give this less weight for pitchers than for
batters, because
it seems that often pitchers don't get a lot of respect from MVP voters. I give
10 for each first down to 1 for each 10th, plus 0.5 for each time receiving some
votes but not in the top ten.
- Hall of Fame, weight 2. This is done almost exactly the same
as for batters,
except that for first-time namers it is ratio of votes to cutoff minus .32 times 10,
to reflect Tom Seaver and Nolan Ryan receiving 1.32 times the cutoff
number of votes.
- W-L, weight 1.5. For each year I compute the pitcher's wins
minus (his team pct)
times (his number of decisions). This is the number of extra games that he won compared
to an average pitcher on his team. This seems like a good statistic, although I think
it rewards the pitchers on a bad team a little bit too much. Tom Seaver is tops in this
with 65.8 additional wins, while John Smoltz is last with only 1.3.
Note that he played for good teams. Steve Carlton had by far the best
single season in this statistic in 1972 when he was 27-10 for a team with
a .372 winning percentage.
- K's, total, weight 0.5. This is just career strikeouts. Palmer
and Thorn do not
feel this is important to helping the team win, hence my low weight for it.
It has a lot of fan appeal, at least.
The next K statistic is adjusted for
the
average number of K's per game in the league, but this one is not. Career totals are
career totals, plain and simple.
- K's per inning, weight 0.5. Take the pitcher's number of
strikeouts per inning
over their career. Then, for each year the pitcher is active, compute the league's
total (both teams) number of strikeout per game, and average this over the pitcher's
career. Take the ratio of the first number to the second and multiply by 100.
Nolan Ryan is tops in this, too. Although Randy Johnson and Pedro Martinez have
a higher number of strikeouts per inning than did Ryan, there are many more strikeouts
per game in recent years than there were during Ryan's era. Ryan's rating
of 101 in this is approximately
1 strikeout per inning in an era when there were on average ten strikeouts per game,
while in 2000 that average was up to 13 per game.
- Career adjusted ERA, weight 1.5. Taken from
baseball-reference.com, this is the person's career ERA adjusted for
the average ERA in his era, and (I think) the stadium in which the
pitcher was pitching.
- Win Shares, total, weight 2. See comments for batters.
- Win Shares, best 4, weight 2. See comments for batters.
- Win Shares, top 4s, weight 2. Each season I find the top
four pitchers in the majors based on Win Shares, breaking ties as I
did for batters. I give 4 points for each time first, 3 for each second,
2 for each third, and 1 for each 4th.
- WAR, lifetime, weight 1. See comments for batters.
- WAR, best 5, weight 1. See comments for batters.
- Post-season, weight 1.5 This is important but is given a fairly low weight due to its unfairness to people on bad teams.
I consider post-season W-L, postseason innings and strkeouts, postseason ERA, World Series's on winnign team, and World Series MVPs. I combine these in a non-scientific way.