Is the AL Really Superior? (Part 3)
Editor’s Note: In Part 1 and Part 2 of this series, MGL showed how players switching leagues generally performed and then adjusted those numbers to take various factors into account. In this final installment, he’ll look at a more direct and accurate method for identifying and quantifying any differences in talent between the two leagues.
Let’s see if we can come up with another, more direct and accurate, method for identifying and quantifying any differences in talent between the two leagues. We will look at how each league’s hitters perform on the road versus the other league’s pitchers in inter-league games and then compare that to how the other league’s hitters do on the road against those same pitchers.
Since we are controlling (more or less) for the pitcher and the park, we should be able to fairly and directly compare each league’s offense to one another. We can use the same, direct method for comparing pitching talent across the two leagues, by looking at how each league’s pitchers perform on the road against opposite-league batters, and comparing that to how the opposite league’s pitchers perform, also on the road, against their same-league batters.
In other words, we will compare: AL batters on the road versus NL pitchers at home with NL batters on the road versus NL pitchers at home.
Since both groups of batters are facing the same pool of pitchers in the same parks (more or less), they should have the same results if they are equally talented. Any difference in results should represent a difference in talent.
Just to make sure there isn’t something “weird” going on when an NL batter travels to an AL park and faces AL pitchers (or an AL batters travels to an NL park), we will also compare: NL batters on road versus AL pitchers at home with AL batters on the road versus AL pitchers at home.
This should also tell us the difference in talent between NL and AL batters, in reverse. We will then do the same thing with pitchers, comparing AL pitchers on the road versus NL batters with NL pitchers on the road versus NL batters (and NL pitchers versus AL batters compared to AL pitchers versus AL batters).
Here is the data for 2005 only:
In inter-league play in the AL parks, the NL batters (on the road of course) had a composite linear weights of -12.7. That is not relative to any league per se. It is the NL batters’ raw road stats versus the AL pitchers, in the AL parks of course, turned into a lwts-like number, using a constant and arbitrary value for the out (the value of the out is usually set so that the league lwts is zero). Note that we are eliminating the DH and pitcher-hitting in the data as well as all games involving the Rockies (at home and on the road).
Now, before we compare this -12.7 to what all AL batters did against their own pitchers on the road (in the same year of course), we have to establish two things in order that we are truly controlling for the pitchers faced and the pool of batters whose stats we are compiling. One, the overall strength of the NL batters (relative to their own league), weighted by the number of PA in inter-league games (we actually expect these batters to be pretty much league average batters, since almost all batters participate in inter-league play), and two, the overall strength of the AL pitchers, also relative to their own league (again, we expect these pitchers to also be about league average). Once again, if that is confusing, don’t worry about it.
As it turns out, the pool of NL batters were -1.9 runs per 500 PA (a tad worse than league average, probably because we are not including a DH in the data), and the pool of AL pitchers they faced are -.23 runs per 500 TBF (slightly better than league average).
So how did AL road batters fare against their own pitchers, roughly the same pitchers in the same parks? Remember this is last year, 2005. The AL road batters were -.5 runs per 500 PA in lwts versus their own pitchers, again, the same pitchers in the same parks that the NL road batters faced in inter-league play. The pool of AL batters were -.4 runs per 500, slightly worse than the average batter, again, because we are not including the DH. The average AL pitcher was -.41 runs per 500 TBF, also slightly better than an average pitcher. If all of these numbers are confusing, and we’re guessing they are, here is the above analysis in chart form:
Year | NL Batters on the Road v. AL Pitchers (Adjusted for the Pool of Pitchers and Batters) | AL Batters on the Road v. AL Pitchers (Adjusted for the Pool of Pitchers and Batters) | Which League is Better? | By How Many Runs per 500 PA and Per Game? |
---|---|---|---|---|
2005 | -10.6 | .3 | AL | 10.9/.81 |
So once again, we find a huge difference between the performance of NL batters and AL batters versus the same AL pitchers in the same parks (more or less), both on the road, in favor of the AL. However, we still don’t know how much of this approximately 11 run (after adjusting for the aforementioned pools of hitters and pitchers) difference between the leagues is due to an actual difference in hitter quality and how much may be due to the NL batters not being familiar with the AL pitchers or the AL parks (or how much is a “fluke” due to sample error). We’ll call this the “familiarity factor,” assuming that it even exists at all.
In order to get an idea as to how great (if it exists) the familiarity factor is, and in order to reduce our sample error, we need to look at the same data for AL pitchers facing NL pitchers in NL parks. If there is no familiarity factor (and no sample error), we would expect to see the same 11-run difference in hitting quality in reverse. If there is a familiarity factor, we would expect to see the AL batters hit around 11 runs better, minus two times the familiarity factor (don’t worry if we lost you with the math).
AL batters hit to the tune of 4.8 runs per 500 PA versus NL pitchers in NL parks (in inter-league play of course). The pool of batters was roughly league average again, around -.2 runs per 500 PA. The pool of pitchers faced (NL pitchers) was worse than league average, by .89 runs per 500 TBF. Again, let’s compare this to NL batters on the road facing NL pitchers. They had a collective lwts of 1.3 runs per 500 PA, the entire pool of these batters was .3 runs per 500 PA, and the pool of pitchers faced was 2.1 runs per 500 TBF better than league average.
If we again adjust for the pool of batters and pitchers faced in both groups (AL batters versus NL pitchers and NL batters versus NL pitchers), we get a difference of only 1.03 runs per 500 PA between the NL and AL batters versus the NL pitchers, in favor of the AL. Here it is in chart form:
Year | AL Batters on the Road v. NL Pitchers (Adjusted for the Pool of Pitchers and Batters) | NL Batters on the Road v. NL Pitchers (Adjusted for the Pool of Pitchers and Batters) | Which League is Better? | By How Many Runs per 500 PA and Per Game? |
---|---|---|---|---|
2005 | 4.1 | 3.1 | AL | 1.0/.07 |
So one way, we get a difference of 10.9 runs and the other way we get a difference of 1.0 runs. As I initially said, we would expect such an asymmetry if there were a familiarity factor which caused both the NL and AL to hit worse against opposite-league pitchers in opposite-league parks, regardless of the true difference in offensive quality between the leagues. A 10.9 and 1.0 run difference (per 500 PA) suggests that the familiarity factor is 4.9 runs per 500 PA and that the AL advantage is 6.0 runs per 500 PA, or .44 runs per game (in 2005).
In other words, when AL batters face NL pitchers on the road, they lose 4.9 runs (per 500 PA) right off the bat (due to being unfamiliar with the pitchers and the park) and then pick up another 6.0 runs because they are that much better than their NL counterparts, for a net gain of 1.1 runs.
NL batters facing AL pitchers lose the same (presumably) 4.9 runs because of the familiarity factor and then another 6.0 runs because they are that much worse than the AL batters, for a total loss of 10.9 runs. So, using this “direct” method of comparing the overall quality of NL batters to AL batters, by seeing how they perform on the road against roughly the same pitchers in the same parks, we have determined that in 2005, AL batters are 6.0 runs per 500 PA better than NL batters, on the average. That is a little more than we found when looking at players who switched leagues (which was around 5.4 runs per 500 PA), but certainly in the same ballpark.
Both methods, while accurate in and of themselves (although the second method is probably “cleaner” and more accurate), are necessarily subject to sampling error due to finite sample sizes. 5.4 runs per 500 PA is equivalent to around .40 runs per game and 6.0 runs per 500 PA, .44 runs per game. Either way, that is a large difference, and would explain the AL’s dominance in inter-league games in 2005, and again in 2006, assuming that the NL does not have a large edge in the pitching department.
Keep in mind that the familiarity factor of 4.9 runs really is a familiarity factor plus a sample error adjustment (in order to artificially make the NL and AL difference “symmetrical”). Once we look at more years, and hence, a larger sample size, we’ll be able to get a better idea as to the true value of the familiarity factor.
If we do the same analysis (using the “direct method) for the entirety of inter-league play, 1997 through 2005 (excluding 2006), we get an average offensive advantage for the AL of 1.1 runs per 500 PA, or .08 runs per game, equivalent to around a 50.8% advantage for the AL. The familiarity factor appears to be 1.4 runs, which is probably closer to what it is in reality (if we had no sample error).
Year | NL Batters on the Road v. AL Pitchers (Adjusted for the Pool of Pitchers and Batters) | AL Batters on the Road v. AL Pitchers (Adjusted for the Pool of Pitchers and Batters) | Which League is Better? | By How Many Runs per 500 PA and Per Game? |
---|---|---|---|---|
1997-2005 | .9 | 3.4 | AL | 2.5/.19 |
Year | AL Batters on the Road v. NL Pitchers (Adjusted for the Pool of Pitchers and Batters) | NL Batters on the Road v. NL Pitchers (Adjusted for the Pool of Pitchers and Batters) | Which League is Better? | By How Many Runs per 500 PA and Per Game? |
---|---|---|---|---|
1997-2005 | 2.1 | 2.5 | NL | .4/.03 |
In other words, in the above charts, NL batters when facing AL pitchers, have a lwts of .9. If we adjust for the familiarity factor, they would be at 2.3 runs. AL batters were at 3.4 runs against the same pitchers in the same parks, 1.1 runs better.
AL batters facing NL pitchers were at 2.1 runs. Without the familiarity penalty, they would be at 3.5 runs. NL batters were at 2.5, also 1.1 runs worse than their AL counterparts. That is where we get the 1.1 runs in AL dominance from. Again, the familiarity factor of 1.4 is more like a “fudge factor” to make sure that the difference in the leagues is symmetrical. It is plausible, however, that batters lose 1.4 runs per 500 PA when facing relatively unknown pitchers in unfamiliar parks.
Even without using a familiarity or fudge factor, the 1.1 run difference for 97-05 and the 6 run difference found in 2005 is simply the average of the differences found when doing the comparisons from both perspectives (AL batters v. NL pitchers and NL batters v. AL pitchers).
For 00 to 05, we get an average dominance, in favor of the AL, of 2.0 runs per 500 PA, or .14 runs per game, less than, but similar to, what we came up with using our first, “indirect”, method (.21 runs per game). The familiarity/fudge factor for 00 to 05 was 2.1 runs.
Year | NL Batters on the Road v. AL Pitchers (Adjusted for the Pool of Pitchers and Batters) | AL Batters on the Road v. AL Pitchers (Adjusted for the Pool of Pitchers and Batters) | Which League is Better? | By How Many Runs per 500 PA and Per Game? |
---|---|---|---|---|
2000-2005 | -1.3 | 2.8 | AL | 4.1/.30 |
Year | AL Batters on the Road v. NL Pitchers (Adjusted for the Pool of Pitchers and Batters) | NL Batters on the Road v. NL Pitchers (Adjusted for the Pool of Pitchers and Batters) | Which League is Better? | By How Many Runs per 500 PA and Per Game? |
---|---|---|---|---|
2000-2005 | 2.2 | 2.4 | NL | .2/.01 |
Using the same, direct, methodology as above, let’s take a look at the NL and AL pitchers.
Without breaking down the numbers as I did above, we find that in 2005, the NL dominates slightly in pitching, to the tune of .25 runs per 500 TBF, or .02 runs per game. The familiarity factor (they pitch worse than expected) is 5.34 runs per 500 TBF (.40 runs per game), suggesting that pitchers really don’t like pitching in opposite-league parks (we were expecting to see a “reverse” familiarity factor – pitching better than expected – since they are facing batters who rarely see them pitch), or perhaps we just have a large sample error.
From 2000 to 2005, we see a .78 runs per 500 TBF (.06 runs per game) advantage for the AL, and only a .42 runs per 500 TBF familiarity factor (which seems more reasonable than 5.34). .06 runs per game is considerably less than what we got when we used the first, indirect method.
Using the “direct” method, over the entire span of inter-league play, again, excluding 2006, we get a pitching advantage for the AL of .90 runs per 500 TBF, or .07 runs per game (and a familiarity factor of .85 runs per 500 TBF, or .06 runs per game).
Here are the above numbers in chart form:
Year | NL Pitchers on the Road v. AL Batters | AL Pitchers on the Road v. AL Batters | AL Pitchers on the Road v. NL Batters | NL Pitchers on the Road v. NL Batters | Which League has the Advantage | By How Many Runs per 500 TBF (Runs Per Game) | Familiarity Factor in Runs per 500 TBF |
---|---|---|---|---|---|---|---|
97-05 | 9.13 runs per 500 TBF | 6.38 runs per 500 TBF | 5.66 runs per 500 TBF | 5.70 runs per 500 PA | AL | .90 (.07) | .85 |
00-05 | 7.30 runs per 500 TBF | 6.10 runs per 500 TBF | 4.94 runs per 500 TBF | 5.30 runs per 500 TBF | AL | .78 (.06) | .42 |
05 | 8.65 runs per 500 TBF | 3.56 runs per 500 TBF | 10.05 runs per 500 TBF | 4.45 runs per 500 TBF | NL | .25 (.02) | 5.34 |
The requisite PBP data is not yet available for 2006 in order to do a “direct” analysis. However, the combined pitching and batting advantage in the AL in 2005 (.42 runs per game) using the direct method is commensurate with the .46 (for 2005) gotten using the previous, indirect method (players who switch leagues).
The previous, indirect method also yielded a .57 run per game advantage for 2006. In order to estimate the 06 advantage, presumably for the AL, using the direct method, let’s look at the transfer of talent from 05 to 06 adjusted for the talent differential between the leagues as calculated from the direct method.
Remember that from 05 to 06, 38 batters went from the NL to the AL. Their collective lwts in 05 was -12.3. Using the direct method, we found that there was a 6 run offensive difference between the NL and AL in 2005. So that -12.3 was really -18.3 in AL currency. The AL transferred 34 players who were -15.2 runs. So the AL actually lost an average of 3.1 runs per player transferred, and the NL had to fill a 4 player “hole” with partial replacement players.
The net number of runs transferred to the NL from the AL in 2006 was 42 runs or .016 runs per game. So the NL actually got better offensively in 2006 to the tune of .016 runs per game. This is commensurate with the .01 runs per game figured earlier, using the indirect method only (rather than the direct plus indirect method this time). That puts the AL advantage in offense in 2006 at .444 runs (the 05 total) minus .016 runs (the migration to the NL in 06), or .43 runs.
Let’s look at the pitching migration again from 2005 to 2006. Remember we told you that at least one other researcher had estimated that a large shift in pitching talent went from the NL to the AL between 05 and 06. According to the direct method, the NL had the pitching advantage in 2005, by .25 runs per 500 TBF. We’ll use this number to figure the migration of pitching talent from 05 to 06. In 2005, 35 pitchers went from the NL to the AL. Their collective lwts was -2.3 runs (below-average pitchers). If we adjust that for the difference in talent between the leagues (.25 runs in favor of the NL), we get -2.05 runs.
The AL sent 39 pitchers to the NL with a collective lwts of -6.5. So, 4.45 runs per pitcher in talent did indeed go from the NL to the AL, although the AL had to replace 4 pitchers with near-replacement-level talent. The net result was 45 total runs sent to the AL from the NL, or .017 runs per game. That still leaves the NL with a small pitching advantage of around .002 runs per game.
Essentially there appears to be parity in pitching between the two leagues this year, despite an influx of talent into the AL (because the NL presumably had the advantage last year). (There is some anecdotal evidence that the AL has some excellent young pitchers that are not reflected in the above numbers and analysis, such that the AL may indeed be the better pitching league this year as well.)
Finally, let’s look at the year-by-year differences in offensive and pitching talent as suggested by the “direct” method of comparison, together with the year-by-year inter-league win/loss records. We’ll also expand the years to include 97-99. Remember when we did the same thing using the indirect method of comparison, we appeared to overrate the AL talent in almost every year and in all 7 years (00-06) combined.
Expected win/loss records versus actual win/loss records (using the “direct” method):
Year | AL Offensive Advantage | AL Pitching Advantage | AL Total Advantage | AL Wins | NL Wins | AL Win Percentage | Expected (Pythag) AL Win% |
---|---|---|---|---|---|---|---|
1997 | -.08 | .21 | .13 | 97 | 117 | .453 | .514 |
1998 | .20 | -.06 | .14 | 114 | 110 | .509 | .514 |
1999 | -.28 | .08 | -.20 | 116 | 135 | .462 | .478 |
2000 | .18 | .11 | .29 | 136 | 115 | .542 | .530 |
2001 | -.11 | .09 | -.02 | 132 | 120 | .524 | .498 |
2002 | -.34 | .35 | .01 | 123 | 129 | .488 | .501 |
2003 | .35 | -.25 | .10 | 115 | 137 | .456 | .510 |
2004 | .31 | -.02 | .29 | 126 | 125 | .502 | .529 |
2005 | .44 | -.02 | .42 | 136 | 116 | .540 | .542 |
2006 | .43 | 0 | .43 | 128 | 75 | .630 | .543 |
Totals | .11 | .05 | .16 | 1223 | 1179 | .509 | .517 |
Note: Comparing the direct method to inter-league win/loss records is in some sense redundant, as those win/loss records are to some extent predicated on the inter-league performances we are measuring with our direct method.
Note: One reason why we may be overestimating the AL pitching advantage using the indirect method (and why it is larger than the direct method), and even the direct method, is that it is possible that NL pitchers, at least initially, have a harder time adjusting to the AL and the DH than do AL pitchers when switching to the more “lineup friendly” NL. More research would be necessary to address this issue.
While the above analyses are somewhat problematic for various reasons, taken in their entirety, it is clear that the conventional wisdom with regard to the current AL dominance in inter-league play is correct – namely that the overall or average talent in the AL is likely much better than that in the NL. However, it appears that most or all of that advantage is in the offense and that the quality of pitching is roughly equivalent in both leagues.
The difference in offensive talent is somewhere around .4 to .5 runs per game, which would give the average AL team around a 55% advantage over the average NL team, assuming that the pitching were equal. If it is true that the AL has the better pitching overall as well (and there is some suggestion in the data that that is true), then the AL advantage may be as high as 56 or 57%.
So if you happen to be in Las Vegas soon, walk into the nearest sportsbook and start loading up on the AL to the win the All-Star game and the World Series, assuming that the cat has not been let our of the bag. Considering the AL’s utter and complete dominance in inter-league play so far this year, the cat may not only be out of the bag, he may be on his way to Vegas himself!