1. Mike and Paul
Paul Splittorff and Mike Flanagan have both passed away this summer. Both were left-handed pitchers from the 1970s and 1980s (Splittorff 1970-1984, Flanagan 1975-1992). Both pitched their entire careers in the American League, Splittorff for one team, Flanagan for two teams although he was clearly associated with one. Both remained with their teams after their playing careers, and were broadcasters with those teams at the time of their deaths. Both teams were powerhouse franchises at the time Splittorff and Flanagan were active, but have long since fallen onto hard times.
Splittorff had a career won-lost record of 166-143; Flanagan, of 167-143. Each pitcher won 20 games once and 19 games once. Each pitcher won 14 or more games five times.
Flanagan started 404 games in his career with a career ERA of 3.90. Splittorff started 392 games with a career ERA of 3.81. Flanagan pitched 101 complete games in 404 starts (25%); Splittorff pitched 88 complete games (23%).
Both had below-average strikeout rates but better than average rates of walks, home runs, wild pitches, balks and hit batsmen. Each issued 41 intentional walks in his career, when the league average given the number of batters faced would have been 76 for each one. Both had above-average numbers of shutouts. Flanagan was 3-2 in post-season play; Splittorff was 2-0.
Splittorff was three years older than I am; Flanagan was two years younger. Their passing thus serves powerfully to remind me that my own time here could end at any moment. They faced each other only once in their careers (April 22, 1978), which seems very odd in that Splittorff faced Catfish Hunter, for example, eleven times, Jim Palmer six times, Tommy John eight times and Scott McGregor six, while Flanagan faced Tommy John and Dennis Leonard seven times each, and Jack Morris eight. Splittorff won their one confrontation, 5-3. Let us hope that they have met again beyond the veil.
2. Mark and Carlos
Suppose that Mark Reynolds is batting against Carlos Marmol in the National League in 2010. What is the chance that the outcome of that matchup will be a strikeout?
I believe. . ..if my math is correct. . .that it is about 66%. In the National League in 2010 the overall strikeout rate (strikeouts per plate appearances) was 19.5%. Reynolds, however, struck out at more than twice the league strikeout rate (39.6%), while Marmol struck out 42% of the hitters that he faced. When they match up, then, both are putting enormous upward pressure on the probability of a strikeout.
Same league. . .suppose that Luis Atilano is facing Jeff Keppinger. Atilano struck out only 40 batters while facing 385 batters, barely over one-half the league strikeout rate. Keppinger struck out only 36 times in 575 plate appearances, less than one-third the league strikeout rate. The probability of a strikeout resulting from a matchup of them (ignoring the platoon factor) is about 3%.
The math works in this way. Suppose that a .600 team is facing a .400 team. What is the likelihood that the .600 team will win?
It’s 69%. This can be figured by a method I introduced more than 30 years ago, called the Log5 method. We begin by asking "What is the logarithmic equivalent of a .600 winning percentage?" The logarithmic equivalent of a .600 winning percentage is that number which, if added to .500 and divided by the total, produces a .600 winning percentage.
That number, for .600, is .750; the Log5 of .600 is .750:
|
|
.750
|
|
|
|
|
----------------------------
|
=
|
.600
|
|
|
.750
|
.500
|
|
|
| |
|
|
|
|
The Log5 of a .500 team is .500. The Log5 of a .700 winning percentage is 1.167:
|
|
1.167
|
|
|
|
|
----------------------------
|
=
|
.700
|
|
|
1.167
|
.500
|
|
|
| |
|
|
|
|
When a .700 team plays a .500 team, the .700 team will win 70% of the time. The Log5 of a .400 team is .333:
|
|
.333
|
|
|
|
|
----------------------------
|
=
|
.400
|
|
|
.333
|
.500
|
|
|
| |
|
|
|
|
So when a .600 team meets a .400 team, the probability of the .600 team winning is 69.2%:
|
|
.750
|
|
|
|
|
----------------------------
|
=
|
.692
|
|
|
.750
|
.333
|
|
|
| |
|
|
|
|
This can also be figured by cross-multiplying the wins and losses. . .let us assume that Texas (30-20) is playing Pittsburgh (20-30). The probability that Texas will win can be figured as:
(Tex W * Pitt L) / [(Tex W * Pitt L) + (Pitt W * Tex L)]
This gets the same result:
(30 * 30) / [(30 * 30) + (20 * 20)] = .692
You can also generalize this method to figure, in essence, a "Log5" for something like "Mark Reynolds strikeout rate"; the process to do that was developed by Dallas Adams, and I don’t exactly understand how because I’m not that good at math. Suppose that we put ".600" in cell A1 of a spreadsheet and ".400" in cell B1:
In Cell C1 we put this:
=(A1*(1-B1))/((A1*(1-B1))*(B1*(1-A1))
You follow? If you put that formula in cell C1, you get .692, which means that a .600 team beats a .400 team 69.2% of the time.
But you can also put Mark Reynolds strikeout rate in cell A1 (.396) and the league strikeout rate in B1 (.195), and get an answer in C1, which is .730:
What this means, literally, is that a .396 team will beat a .195 team 73.0% of the time. And we can do the same with Carlos Marmol’s strikeout rate, compared to the league:
Which means, literally, that a .416 team will beat a .195 team 74.6% of the time, but also means that comparing Carlos Marmol’s strikeout rate to the league norm, Carlos Marmol has a winning percentage of .746.
Look, all of this sounds horribly obscure and theoretical, but it works like a dream in practice, and you can take my word for that or you can check it out, but you will save a hell of a lot of time if you take my word for it.
When a good team plays a good team they are pushing against each other and thus pushing in the direction of .500, but when a good team plays a bad team they are pushing the percentages in the same direction, and thus pushing toward 1.000 (or toward .000). When Carlos Marmol pitches against Mark Reynolds they are both pushing the strikeout probability upward toward 1.000, so that’s not good against good; that’s good against bad. To reflect this, we have to convert Carlos Marmol’s .746 winning percentage to its complement, .254. Then we put Mark Reynolds "strikeout win percentage" (.730) in cell A1, and the league’s "strikeout win percentage" against Carlos Marmol (.254) in cell B1, and we get the outcome in C1:
Or, if you prefer, you can put Carlos Marmol’s "strikeout win percentage" (.746) in cell A1 and the league’s "strikeout win percentage" against Mark Reynolds in cell B1 (.270), and you’ll get the same thing; it makes no difference which you consider the Alpha and which you consider the Beta. Anyway, .888 is what we could call the "Mark Reynolds/Carlos Marmol compressed strikeout win percentage."
Mark Reynolds and Carlos Marmol have a very high compressed strikeout win percentage (.888), but in the league as a whole, 80.5% of plate appearances do NOT result in strikeouts. The .888, then, is battling head to head with that .805.
And the result is .658, which is the frequency with which Carlos Marmol will strike out Mark Reynolds—65.8% of the time. Let’s display that as one continuous block:
|
|
A
|
B
|
C
|
|
1
|
.396
|
.195
|
.730
|
|
2
|
.416
|
.195
|
.254
|
|
3
|
.730
|
.254
|
.888
|
|
4
|
.888
|
.805
|
.658
|
If Mark Atilano faced Jeff Keppinger, they would both be pushing the likelihood of a strikeout down. Atilano’s "strikeout win percentage" is .323:
While Keppinger’s is .216:
When we put those together, we have to convert one of them into its complement so that they will both push in the same direction, rather than pushing against one another, so the combined "strikeout win percentage", with both Keppinger and Atliano pushing it downward, is .116:
|
.104
|
.195
|
.323
|
|
.063
|
.195
|
.216
|
|
.323
|
.784
|
.116
|
Which will result in strikeouts in 3.1% of plate appearances:
|
|
A
|
B
|
C
|
|
1
|
.104
|
.195
|
.323
|
|
2
|
.063
|
.195
|
.216
|
|
3
|
.323
|
.784
|
.116
|
|
4
|
.116
|
.805
|
.031
|