clock menu more-arrow no yes

Filed under:

Fred McGriff, the Hall of Fame, and Adjusting Stats

New, 8 comments

The steroid era inflated run production across baseball and cheapened Fred McGriff's career numbers and Hall of Fame case. Can we try to adjust for this?

Jonathan Daniel/Getty Images

In the past week, Tom has outlined the Hall of Fame cases for Carlos Delgado and Fred McGriff, both standout first baseman for the Jays. In both articles and the comments, the consensus was that both fall short of the standard of enshrinement for first basemen in Cooperstown, but also both seem unduly punished for having played at a time of very high offensive output driven by steroid/PED use despite neither being connected to them.

A good example is comparing Fred McGriff's 1990 and 1999 seasons. In 1990, McGriff hit .300/.400/.530, which represented a level of run production 57% above league average (wRC+ of 157) and he was crediting with 43 batting runs above average. In 1999, McGriff put up an even better raw line of .310/.405/.552, but this was only 40% above league average so he was credited with just 33 batting runs above average (38 fewer PA). These 10 fewer runs are about 1 WAR difference due to the higher run scoring.

At the same time, I've recently been thinking a lot about the nature of adjusting statistics. One of the most important insights of sabermetrics has been applying adjustments to "raw" stats to adjust for contextual factors outside the player's control, with park factors and the overall level of run scoring being two of the most significant. At a high level, even applying the most basic or imperfect of adjustments will deliver far more understanding then is obscured by the imperfect of the adjustment. However, the finer the analysis, the more important is to consider how good the adjustments are, or they can obscure as much as they reveal.

Consider a college professor who gives an exam and based on prior experience expects an average of 75%,but when the results come in the average is 85%. He/she figures that the exam was too easy, or perhaps marked too easily, and decides to adjust the grades by removing 10% from every exam to get to a 75% average. However, unbeknownst to the prof, a copy of the exam got out and was widely distributed to 2/3 the examtakers. These cheaters had an average of 90% whereas non-cheaters averaged the expected 75%. The adjustment restores the overall average, but does nothing for the inequity between cheaters and non-cheaters.

In essence, this is what the WAR framework does, by standardizing run production to the overall level of run production, and attributing value accordingly. McGriff totaled 52 bWAR and 57 fWAR, Delgado totalled 43-44 by both (but played his entire career in the "steroid era" versus only half for McGriff). Both totals short of the 65-70 level generally required for serious Hall of Fame consideration. But if they're being unfairly punished, there should at least be an effort to seriously account for it, which is what I'm going to try and do below.

To attempt such a thing, the first step has to be trying to estimate the impact of steroids. When I try to break down offensive production, there's four stats or components with which I start: K%, BB%, BABIP and ISO. These four simple stats together give so much information: how often a player puts the ball in play, the value he produces when he does and doesn't, and a broad view of his skill set (patience, ability to make contact and quality contact, power, etc). It doesn't capture everything, but it explains a lot: a model with just the four components explains 93% of the variation in league runs/game on a seasonal level from 1920-2014 (more on that below).

To start, examining each of these components in detail can give some idea of how steroids could impact production in each, thereby determining where adjustments should and should not made:

  • Isolated power is the most obvious area where steroids would increase performance, since in large measure, it's about home runs. More strength should translate into better distance on fly balls, batted ball distance is highly correlated to home-run rate; the transmission mechanism is intuitive. There's natural year-to-year variation, but the unprecedented increase is obvious:
  • ISO2
  • BABIP is largely a function of the hitter's batted ball profile and how hard he hits the ball, which basically comes back to being able to barrel up the ball and make quality contact (and is also a prerequisite to tapping into power). Things like bat speed, hands, wrist, swing length and loft come into play, but for the most part it's hard to think steroids have a huge impact on hitter ability here. There's also some element of speed (beating out ground balls) and opposition defensive quality, in particular range to get to balls. If steroids helped older players stay on the field longer (at the expense of younger players with superior defensive abilities), this is probably where steroids have the biggest impact on BABIP. But, all hitters would also benefit from that, not just the users.
  • K% is largely a function of a hitter's ability to recognize pitches; differentiate balls from strikes; general aggressiveness/patience; ability to make contact; two strike approach; and the ratio of pitches thrown to him in and out of the zone. The first five abilities are entirely about the hitter, and are not things that steroids are known to enhance. The last factor could introduce some steroid effect, but in large part K% would be minimally impacted by steroids
  • BB% is largely a function of the same factors as K%, with less emphasis on contact rate and more emphasis on aggressiveness, and the ratio of balls thrown in and out of the zone. Again, it's hard to see steroids having much impact, but the effect of greater power resulting in fewer balls in the zone would be larger than for K%

So with that in mind, let's turn back to Fred McGriff, who basically had two careers, through 1994 and post-1994. Below are summaries of the two halves, along with league averages for non-pitchers weighted by McGriff's PA in each season:

McGriff PA BB% K% ISO BABIP wRC+ fWAR/650
1986-94 4714 14.4% 19.5% 0.256 0.308 151 5.3
1995-04 5460 11.5% 17.6% 0.199 0.310 119 2.3

Average BB% K% ISO BABIP R/G
1986-94 8.7% 14.5% 0.135 0.290 8.6
1995-04 9.2% 16.2% 0.162 0.300 9.6

Through 1994, McGriff was a premier hitter and player in MLB, walking a ton (+5.7% vs. league average) while striking out a fair bit (+5.0%), with huge power (+121 points of ISO) while getting hits as well (+18 points). He produced runs at a rate 50% above average, piling up 5.3 WAR per 650 PA.

After 1995 however, he was more of a league average type of player, with just 2.3 WAR per 650 PA. The major culprit was a power outage: after a career high .303 ISO in 1994, he never topped .242 again. A fascinating contrast: as power took off (league average ISO 27 points higher), McGriff's output fell substantially (57 points), for a devastating 84 point relative fall.

His BABIP increased two points, though the league rose 10 points, a relative decline of 8 points which is in line with normal aging. His walk rate fell by almost 3% but was still well above average, and he compensated by cutting his strikeout rate almost 2% despite the league rate increasing nearly the same amount. This combination was slightly less valuable, but still very productive.

Circling back to how much playing in the steroid era hurt McGriff, I referred earlier to a model to predict the change in runs/game using the four components. For simplicity, I ignored interaction between variables, which isn't a huge issue given how powerful the results were (p less than 0.00001 for each coefficient). We get runs/game = -9.07 + 23.7*BB% - 8.6*K% + 25.1*ISO + 45.9*BABIP.

Runs per game rose by a full run from 8.6 in the first half of McGriff's career (in line with the historical average of 8.7 for 1920-2014 and 8.5 for 1920-1990) to 9.6 in the second half. Plugging in the component numbers into the formula, we get an expected increase of 1.1 runs per game, which is in line. By component, that +1.1 breaks into +0.12 from BB (11%), -0.15 from K (-13%), +0.68 from ISO (61%), +0.46 from BABIP (41%).

If we assume that the increase in production from ISO and BB% was entirely entirely steroid related, that none of the increase in K% or BABIP increase was steroid related (or at least that McGriff benefited equally from the BABIP increase), and that McGriff was clean, then we can say about 70% of the increase in run scoring during the second half of McGriff's career was undue. Therefore, by adjusting those second half numbers to a run scoring environment of 8.9 runs/game, rather than 9.6 runs/game, we can properly credit McGriff's production.

Running that through, that would increase his wRC+ in the second half of his career from 119 to 127, and increase his batting runs above average by 46 from 133 to 179. That only works out to about an extra 4-5 WAR, which still wouldn't vault McGriff into serious Hall of Fame discussion.

We can go one step further. Let's assume McGriff had the 1995-04 production he did, in the same context as the first half of his career. That is, the entire increase in run production was steroid related and he got zero benefit. That would give him another 20 batting runs, or about 2 WAR. Still not enough.

For fun, we can do the same with Delgado, whose career was almostly perfectly centred on the steroid era and whose career runs/game played in of 9.51 surely ranks along the highest all-time. If Delgado had put up the same production in the 8.55 run/game average of the six years before and after his career, he would have about 100 more batting runs. That's about 10 WAR, but would still leave him in the mid-50s and short of serious consideration.

Fred McGriff and Carlos Delgado were both historically great MLB players, and great Blue Jays who had wonderful careers. But even resolving all ambiguities about the era in which they played in their favour, they still come up short of the Cooperstown bar.