keyboard_arrow_uptop

Thanks to the two of you–blood relatives included–who wrote in to remind me that I have some unfinished business to attend to. At the end of my column on “sharp” and “flat” teams, I had promised to mine the historical data and get a read on whether there’s any predictive power to how the talent on a particular team is distributed. BP business and my desire to rant about the White Sox intervened, so here we are two weeks late.

To review, “sharpness” is determined by how much talent a team has at the extremes. A team with a lot of superlatively good players and a lot of excruciatingly bad players will be sharp. So, for that matter, will a team that only has way above average (or way below average) players, which is why when we talk about sharpness, we also need to keep team quality in mind. A team where everyone is about average will be flat. Mathematically, sharpness is defined as the absolute value of the number of runs a player produces above or below average at his position, summed across a team’s entire roster. I then subtract the average sharpness score (which is in the low 300s) to set everything relative to zero.

If we calculate the sharpness score for every team between 1959 and 2006 (excluding the strike-shortened years of 1981 and 1994) and compare it against a team’s runs above average, here is what we get:

This chart looks like a bat (not the baseball kind). It’s roughly symmetric about the y-axis, but not about the x-axis. That is because there is some intrinsic relationship between a team’s sharpness and its runs above or below average. Sharpness is defined by extreme performances among individual players; generally speaking, those are going to be found more commonly when the team itself has registered an extreme performance. Thus, average teams are more likely to be flat, while very good and very poor teams are more likely to be sharp, forming the wings of the bat.

Let’s take a brief detour to run down the sharpest and flattest teams in our database.

Sharpest Offenses
Team              W-L      OPS+     Sharp
-----------------------------------------
1996 Mariners     85-76     113     174.4
2000 Red Sox      85-77      92     148.3
2001 Giants       90-72     119     147.8
2003 Cardinals    85-77     116     140.3
1959 Braves       86-70     112     140.2

None of these teams should rate as a surprise. The mid-90s Mariners cement their legacy as one of the strangest collections of talent ever. They had a very good offense–most of their sharpness comes on the “way above average” side, but there were sinkholes at left field and third base, and that’s before we start to get into the pitching staff. The 2000 Red Sox got fantastic years out of Nomar Garciaparra and Carl Everett, but ruined it with a series of below replacement-level performances elsewhere, perhaps ushering in the reign of Theo Epstein. The 2001 Giants: that’s just Barry Bonds. The recent Cardinal teams have characteristically been very sharp, and the Aaron/Mathews Milwaukee Braves were essentially a precursor to the ’90s Mariners, though they did come away with a World Series title in 1957.

Flattest Offenses
Team             W-L        OPS+     Sharp
------------------------------------------
1985 A's         77-85       104     -94.7
1968 Pirates     80-82       106     -92.4
1989 Angels      91-71        98     -91.9
1968 Cardinals   97-65       101     -91.1
2005 Blue Jays   80-82        95     -90.5

It occurred to me that you could use “dullest” rather than “flattest” as the antonym to “sharpest,” because that’s what these teams were. The 2005 Blue Jays are the flag-bearers for this type of offense, the sort of Garrison Keillor neo-sabermetric bizarro world team in which every player is exactly average (no regular had an OPS below .703 or above .818).

Sharpest Pitching Staffs
Team               W-L       ERA+      Sharp
--------------------------------------------
1996 Angels        70-91      92       144.5
1996 Tigers        53-109     79       138.7
1997 Mariners      90-72      94       116.7
1997 Blue Jays     76-86     117       114.1
2006 Twins         96-66     113       113.3
2005 Astros        89-73     118       106.8

I’ve cheated a bit to include the 2005 Astros, who are a team that you’d certainly expect to see on the list. In any event, we have three strong pitching staffs and three relatively weak ones, and there are the mid-90s Mariners again. The 1996 Angels are notable for having perhaps the biggest gap in performance between their rotation (very, very bad) and their bullpen (very, very good) of any team in the modern era.

Flattest Pitching Staffs
Team              W-L       ERA+     Sharp
------------------------------------------
1975 Red Sox      95-65      103     -95.7
1966 Yankees      70-89       98     -87.6
1980 Pirates      83-79      102     -85.7
1967 Angels       84-77       98     -83.7

The ERAs for the five primary starting pitchers for the 1975 Red Sox: 4.02, 3.95, 3.95, 4.43, and 4.42. The 1980 Padres are even more extreme: 3.51, 3.67, 3.91, 3.68, and 3.65, though they did have Rollie Fingers working out of the bullpen for them.

Sharpest Overall
Team               W-L      OPS+     ERA+     Sharp
---------------------------------------------------
2000 Red Sox       85-77      92      117     238.9
1996 Angels        70-91      95       92     216.2
1999 Mariners      79-83     100       96     202.0
2000 Expos         67-95      95       90     195.7
2005 Yankees       95-67     111       98     193.4

We’ve already introduced you to most of these teams. The 2000 Red Sox had one of the sharpest offenses of all time, but they added to that Pedro Martinez‘ 1.74 ERA… and brother Ramon’s 6.13. It’s the 1999 squad that winds up representing the Mariners, though both the 1996 and 1997 teams also rank in the top ten.

All of these teams are quite recent: the sharpest pre-1990 team is the 1966 Giants (129.4). That is mostly a reflection of the high-octane offensive environment, as it’s easier to have extreme performances when more runs are being scored, and there’s no scalar component to dampen this effect in the metric that we’re using. But it’s perhaps also a reflection of free agency, which allows unnatural monsters like the 2005 Yankees to come into being.

Flattest Overall
Team              W-L       OPS+     ERA+      Sharp
----------------------------------------------------
1966 Yankees      70-89      103      98      -151.9
1976 Braves       70-92       88      98      -143.6
1968 Pirates      80-82      106     107      -141.1
1983 Indians      70-92       92      96      -136.1
1968 Cardinals    97-65      101     116      -131.1

If the 1977 Yankees were the Bronx is Burning team, the 1966 version might inspire The Bronx is Gradually and Entropically Disintegrating by Means of the Second Law of Thermodynamics. They did have Mickey Mantle, who was still good for a 169 OPS+, but everyone else was either well past his prime or non-descript. It’s curious that the pennant-winning 1968 Cardinals make the list, since that was the year that Bob Gibson had his 1.12 ERA, but the offense was extremely balanced, with every regular performing just about at the league average for his position. Incidentally, that was an extremely interesting team: Steve Carlton, Orlando Cepeda, and Roger Maris, three players that you wouldn’t generally associate with the Cardinals, all played important roles.

Our original article divided teams into a matrix of nine categories based on their sharpness and their run differential. We will now define those categories by placing the boundaries at 50 runs above and below average. For example, a team like the 1968 Cardinals, considerably above the 50-run threshold in run differential and considerably below it in sharpness, would be classified in group 1F. Descriptive data for the nine classifications follows.

Type    Description       n     W     L     Pyth W    Pyth L    RS/G     RA/G
-----------------------------------------------------------------------------
1F      Strong-Flat       63   89.8  71.6    90.0      71.4      4.4      3.9
1N      Strong-Natural   228   91.3  69.8    91.1      70.0      4.6      4.0
1S      Strong-Sharp     103   93.9  66.8    93.1      67.6      5.2      4.4
2F      Average-Flat     116   81.9  79.2    81.8      79.3      4.1      4.0
2N      Average-Natural  221   80.4  80.0    80.3      80.2      4.4      4.4
2S      Average-Sharp     44   80.8  81.0    81.1      80.7      4.9      4.8
3F      Poor-Flat         68   70.3  90.3    71.0      89.6      3.8      4.4
3N      Poor-Natural     225   68.3  92.8    68.9      92.2      4.0      4.7
3S      Poor-Sharp        86   66.6  94.3    66.6      94.2      4.4      5.3

As we could predict from the bat-shaped chart, these classifications are somewhat asymmetric; average teams, for example, are about three times more likely to be flat than to be sharp. The classifications differ in other respects too. There is a fairly strong correlation between sharpness and both runs allowed and runs scored, suggesting that our model is a bit askew with respect to run scoring environments. Strong, sharp teams (Class 1S) had more wins on average than strong, flat teams (Class 1F), which is a natural extension of the principle that the better a team, the sharper it is liable to be; the opposite is true for teams on the flat side of the spectrum. Finally, there is some evidence that sharp teams tend to outperform their Pythagenport record (the effect is statistically significant at the 95 percent level), perhaps because they have greater ability to leverage their assets. Teams with an extremely top-heavy bullpens (one or two relief aces but little depth), for example, have been shown to outperform their Pythagenport projections, and will also tend to be classified as sharp.

Because of these systematic differences, we require a bit of finesse in how we evaluate the performances of the teams going forward. In particular, I developed a simple regression model to predict a team’s number of wins and losses based on its runs scored, runs allowed, wins and losses in the previous season. The composite results of that analysis for each of our nine classifications follow:

Type    Description      Wins n    Year n-1   Predicted     Actual      Delta
------------------------------------------------------------------------------
1F      Strong-Flat        63        89.8        86.3        86.3        +0.0
1N      Strong-Natural    228        91.3        87.1        86.5        -0.6
1S      Strong-Sharp      103        93.9        88.8        89.3        +0.4
2F      Average-Flat      116        81.9        81.6        82.3        +0.7
2N      Average-Natural   221        80.4        80.9        81.6        +0.7
2S      Average-Sharp      44        80.8        81.1        80.4        -0.7
3F      Poor-Flat          68        70.3        75.8        74.4        -1.4
3N      Poor-Natural      225        68.3        74.2        73.6        -0.6
3S      Poor-Sharp         86        66.6        72.5        74.0        +1.5

The key column to look at is ‘Delta,’ which is the excess of actual wins to predicted wins. Among the average and strong teams, the results are disappointing in their ambiguity; teams in classification 2N show some tendency to outperform their projection, for example, while teams in 1N tend to underperform it.

On the other hand, there are some stronger associations among the poor teams. Poor, sharp teams (Class 3S) outperform their projection by 1-2 wins, while poor, flat teams (Class 3F) underperform it by about the same margin; the results are on the verge of statistical significance depending on what test is used. The idealized Class 3S team has both some extremely weak spots (i.e. near or below replacement level) which can be replaced relatively cheaply; in addition, it has a few strong spots that can serve as focal points for the rebuilding effort. This is good news for the Tampa Bay Devil Rays, for example. The Class 3F teams, on the other hand, simply have a lot of mediocre talent, and a longer clean-up process ahead.

There is also some evidence that this trend becomes more profound as we proceed outward by more than one season (see chart below). For example, if we look five years outward for teams with enough contiguous data, the Class 3S teams have nearly a two-win advantage over their flatter counterparts, even though they started from a lower baseline.

What we may be seeing here is simply the fossil effects of good management. The idealized Class 3S team–think again of the Devil Rays–is not spending extraneously on mediocre talent, but has begun to develop some pre-arbitration star talent from its farm system. The Class 3F teams, on the other hand, may be more like the Pirates and the Reds, and are using up too many resources trying to stem their losses, rather than identifying a solid long-term game plan.

### Latest Articles

2/22
11
2/22
0
• ##### Transaction Analysis: The Half-Measure Mets
2/22
1
You need to be logged in to comment. Login or Subscribe