After a five-year slide in scoring landed on the lowest mark since 1981, everyone’s scrambling for offense. That’s a natural response, but maybe not a sensible one.
Stating the obvious, there’s no fixed scoring level that wins ballgames. You just have to score more runs than you allow. And the “pythagorean formula,” which predicts winning percentage from runs scored and allowed, has two corollaries that speak to the most efficient path to improving a team’s record:
- For most teams, a fixed gain in scoring margin is worth somewhat more if applied to run prevention.
- Prevention’s edge is larger at lower scoring levels.
I don’t mean to overstate this. The first point is strongest for teams that are already good, and the second is minor except on extreme ends of the scoring spectrum. But they’re real, measurable tendencies. And don’t they make the knee-jerk reaction to low scoring kind of funny?
I was going to run some pythagorean examples here to show the edge for run prevention over scoring, but those interested can easily work it out for themselves. Those who put little stock in the pythagorean formula can focus instead on the ratio of runs scored to runs allowed, or just jump ahead to the historical data.
Regarding ratios, here’s how a 60-run gain would affect four 2014 teams with varying run differentials:
- Athletics (+157): 729 RS, 572 RA, base ratio 1.27.
Scoring 60 more runs makes the ratio 1.38.
Stopping 60 more runs makes the ratio 1.42. - Giants (+51): 665 RS, 614 RA, base ratio 1.08.
Scoring 60 more runs makes the ratio 1.18.
Stopping 60 more runs makes the ratio 1.20. - Reds (-17): 595 RS, 612 RA, base ratio 0.97.
Scoring 60 more runs makes the ratio 1.07.
Stopping 60 more runs makes the ratio 1.08. - Yankees (-31): 633 RS, 664 RA, base ratio 0.95.
Scoring 60 more runs makes the ratio 1.04.
Stopping 60 more runs makes the ratio 1.05.
The edge for prevention isn’t always large. But wouldn’t you prefer the bigger ratio, other things being equal?
(Using a 60-run gain for the examples was somewhat arbitrary, but it’s not drastic. More than one-fourth of teams in the last 20 years gained at least 60 net runs over the year before, and one-sixth gained 100 or more. The average net change was +/- 82 runs. See method note below.)
The predicted edge for run prevention is greatest for teams that already outscore their foes by a solid margin. Per the formula, last year’s A’s would have gained 2 full wins by applying a 60-run gain to prevention instead of scoring. As the base ratio falls, the relative value of gains starts shifting from defense towards offense. But it takes a ratio below 0.90 to fully flip the edge in favor of scoring gains. Only six teams last year had a ratio that bad.
Whatever you think of the theory, it’s backed by historical data of real wins and runs. In the last 20 years:
- Teams that gained at least +5 wins from the year before improved an average of 43 runs scored, 55 runs allowed.
(198 teams, +12.2 wins) - Teams that gained at least +10 wins improved by 53 runs scored, 75 runs allowed.
(118 teams, +16.0 wins) - Teams that fell by at least -5 wins declined by 44 runs scored, 50 runs allowed.
(205 teams, -11.8 wins) - Teams that fell by at least -10 wins declined by 59 runs scored, 71 runs allowed.
(110 teams, -15.9 wins)
(Method note: Team runs gained or lost were pegged to their league’s scoring norms, so that yearly changes reflect team quality and not spikes in league context. The shortened years of 1994-95 were pro rated to a full slate, so that teams in the next years can’t show “wins gains” despite a worse won-lost percentage.)
No matter how I sort these data, run prevention shows a stronger impact. Suppose we start with changes in runs scored and allowed, and track the resulting wins. Here are the averages for teams with one-year gains of 50 to 100 runs scored or allowed, 1995 through 2014. (The groups had similar total gains in runs margin.)
- Runs scored improved by 50-100: avg. +72 RS (+75 margin) … +6.2 wins … 12.1 marginal runs per win
(111 teams) - Runs allowed improved by 50-100: avg. +70 RA (+77 margin) … +7.2 wins … 10.7 marginal runs per win
(96 teams)
The difference also shows up in teams with small gains of 20 to 49 runs on either side:
- Runs scored improved by 20-49: avg. +35 RS (+42 margin) … +3.8 wins … 11.0 marginal runs per win
(89 teams) - Runs allowed improved by 20-49: avg. +35 RA (+41 margin) … +4.2 wins … 9.8 marginal runs per win
(82 teams)
If this isn’t widely understood, it may be due to human nature. We attack problems by breaking them into components, then targeting efforts to areas of biggest apparent need. Everyone involved with baseball — management, players, writers and fans — instinctively judges teams in terms of offense and defense, gauging each component against the competition. Balance is the unstated ideal.
But the problem of winning more ballgames is solved by improving the runs ratio. And for most teams, improving run prevention is the more efficient route.
__________
Two caveats to the primacy of run prevention:
(1) Pitching is more volatile than hitting.
… But maybe not as much as you would think: In the last 20 years, the average team change over the year before was +/- 56 runs scored, and +/- 60 runs allowed.
(2) Avenues of improvement depend on available talent. Teams with bad offense and good pitching (measured in context) will find it easier to add hitters who are better than what they have.
… But if everyone thinks there’s a hitter shortage, then that “shortage” seems artificial — more like a failure to accept the new normal in scoring.
The perception of a hitter shortage could make sense if the variance among current hitters was unusually small. That would make incremental gains harder to get by replacing one hitter with another. And by raw numbers alone, the variance is smaller now than a decade ago, because the numbers are down. But context is crucial: Just because it’s harder to add 10 homers to your first-base output, that doesn’t mean it’s harder to add something equal in current value to what 10 HRs did ten years ago.
Using OPS+ to gauge distance from the average, I don’t see much change in variance from ten years ago. For 2003-04 and 2013-14, I ranked the top 240 hitters by OPS+ — all the regulars, basically — and found the OPS+ averages for eight groups of 30, each representing one regular spot per team:
| OPS+ Rk. | 2003-04 OPS+ |
2013-14 OPS+ |
| #1-30 | 149 | 145 |
| #31-60 | 125 | 124 |
| #61-90 | 117 | 115 |
| #91-120 | 110 | 109 |
| #121-150 | 103 | 104 |
| #151-180 | 98 | 99 |
| #181-210 | 91 | 95 |
| #211-240 | 85 | 87 |
(Min. 600 PAs over the two years.
Group OPS+ weighted by PAs.)
The top tier in 2003-04 has a modest edge (149-145), but most of that comes from the crazy 247 OPS+ of Barry Bonds. The rest of the tiers are pretty similar for both periods.
Suppose we simply count the elite hitters: An OPS+ of at least 150 was logged by 8 guys over 2003-04, and 9 in 2013-14. At OPS+ 135 or better, the counts were 24 for 2003-04, and 23 for 2013-14. At 125 and up, each period had 44. What shortage?
The counterpart of the current hunt for hitters is a perceived glut of pitchers. All the big deals before the winter meetings were for hitters, as GMs tried to beat the rush by shopping early. But running the same exercise for starting pitchers, the variance in ERA+ seems as big or bigger now than ten years ago. I ranked by ERA+ the top 150 starters in IP for 2003-04 and 2013-14, and found the averages in tiers of 30, each representing one rotation slot:
| ERA+ Rk. | 2003-04 ERA+ |
2013-14 ERA+ |
| #1-30 | 132 | 136 |
| #31-60 | 110 | 113 |
| #61-90 | 102 | 102 |
| #91-120 | 93 | 93 |
| #121-150 | 84 | 81 |
(Min. 20 starts over the two years.
Group ERA+ weighted by IP.)
The top two slots in 2013-14 are 3-4 points above the prior period, while slots 3-4 are the same, and slot 5 is 3 points worse. If these numbers mean anything, don’t they suggest there’s more to gain now by upgrading a rotation’s back end?
The picture’s similar using Wins Above Average instead of ERA+:
| WAA Rk. | 2003-04 WAA |
2013-14 WAA |
| #1-30 | 5.9 | 5.6 |
| #31-60 | 2.2 | 2.2 |
| #61-90 | 0.6 | 0.6 |
| #91-120 | -0.7 | -0.8 |
| #121-150 | -2.3 | -2.6 |
Here, the top tier of 2003-04 fares a little better, mainly from averaging 40 more innings. But the spread from top to bottom is identical. It seems there’s still as much to gain as ever by improving a rotation.
Could it be that front offices still judge pitchers by raw ERA, perhaps unconsciously? Far more qualified starters met basic ERA thresholds in 2014 than 2004 — 22-7 for ERA below 3.00, 39-18 for ERA under 3.50, and 66-33 for ERA under 4.00. But the average ERA has dropped from 4.46 to 3.74, so a 3.50 ERA doesn’t mean the same thing nowadays.
__________
If the emphasis on hitters is a bit misguided, more so is the focus on home runs. Ben Cherington could have been speaking for all general managers when he lamented that “It’s harder to find power.” That viewpoint is ironic because, even after years of decline, home runs remain historically high. Although last year’s HR rate was the lowest since 1992, it’s still higher than all but two years from 1973-92. It’s 10% above the average for 1946-92, and 1% above average for the whole post-War period.
So why do GMs talk as if the steroids era was normal, or desirable? Does every team need a 30-HR man in order to win?
- From 1973-92, just half the World Series champs had a 30-HR hitter, and just over half the pennant winners.
- Even during the steroids era, five straight champions lacked a 30-HR man (1995-99).
- The 30 pennant winners from 1993-2008 featured just three league HR champs.
- The last five World Series featured just two of the 50 men who hit 33 HRs or more.
- Just three of last year’s eleven 30-HR men were on the ten playoff teams.
- Last year’s pennant winners were led by 22 and 19 HRs.
So why is the scramble aimed at home runs, and not the more critical shortage, high on-base percentages?
Last year’s .251 batting average and .314 on-base percentage were both the lowest of the DH era. Only three qualified hitters had a .400 OBP, matching the fewest since 1968. Leadoff and #2 hitters had a combined .324 OBP, second-worst of the DH era. Yet I’ve heard no GM bemoan the lack of table-setters.
OBP affects scoring more than home runs do. For 2014, I took each league‘s top five in scoring, in OBP and in HRs (excluding the Rockies from all rankings*), and cross-checked their league rank in the other two stats. Summarizing these combined top tens:
- Top 10 in scoring averaged #3.3 in OBP, #4.9 in HRs
- Top 10 in OBP averaged #3.6 in scoring
- Top 10 in HRs averaged #4.9 in scoring
(* Coors Field is such an extreme park that the Rockies’ stats only blur the picture of the league as a whole. They rank high in scoring whether they hit home runs or not, because their OBP is always high — above league average all 20 years at Coors.)
For OBP and HRs, the league ranks of the last ten pennant winners (ordered by highest ranking, again excluding Rockies):
- OBP: #1, #1, #1, #2, #3, #4, #4, #6, #8, #10 — Avg. rank #4.0
- HRs: #2, #4, #5, #5, #5, #6, #10, #12, #15, #15 — Avg. rank #7.9
Whether gauged by scoring or by team success, OBP seems more vital than HRs. So if front offices must target offense over pitching, they might at least focus on the key stat.
Now, maybe team power can be boosted more quickly than on-base percentage. And maybe GMs fear that OBP is a losing battle: With last year’s 7.6% walk rate tying 1968 for the lowest since 1922, and with the grip-and-rip batting style firmly entrenched, building sequential offense seems a tall task.
But it can be done. Of the 25 teams last year that hit fewer homers than the year before, 12 actually scored more runs. Those twelve shed an average of 14 HRs (155 to 141, -9%), but rose by 30 runs (645 to 677, +5%). Their slugging average dipped 3 points, but their OBP rose 3 points. (They also stole 10 more bases per team. The three that had non-trivial dips in OBP averaged +27 in SB.)
If there is a threshold below which the correlation of OBP with scoring and success breaks down, we haven’t reached it yet. Seven of ten playoff teams last year ranked in their league’s top five in OBP, and just one playoff team ranked in the bottom five. Meanwhile, the Astros ranked 3rd in AL HRs, better than four of five playoff teams. But they were 13th in OBP (worst in strikeouts), and 14th in scoring. The Cubs were 2nd in NL HRs, better than all five playoff teams — but they, too, were 13th in OBP, worst in strikeouts, and thus 12th in scoring.
Overvaluing sluggers in this environment could exacerbate the falling OBP, making sequential offense even tougher, thus fueling a vicious cycle. I can’t help but think of another time that the major leagues shed 20 points of OBP in a five-year span, plus a big drop in homers, landing on the lowest scoring and OBP marks since the dead-ball era. That was 1963-67.
Be careful what you wish for, general managers. It can always get worse.
__________
As usual, all the raw data for these studies came from Baseball-Reference.com.
