https://blogs.fangraphs.com/triple-slash-line-conundrum-voros-mccracken-edition/Been a long time, but this is a pretty interesting article on the importance of AVG, OBP, and SLG in team run scoring. Starts out with a Tangotiger poll on which is the better hitter's traditional slash line, .315 / .365 / .510 vs. .260 / .365 / .510.
TL;DR answer is there isn't a meaningful difference in terms of run scoring.
The article looks at team run scoring since 2010, regressing for AVG, OBP, and SLG to see which correlates most closely to run scoring, then does a double-regression using which two factors in combination is the best predictors for run scoring. By itself, SLG most closely correlated with run scoring (.840), and the highest correlation combo of OBP + SLG gets that up to .885. Nothing surprising there. The interesting thing is how little AVG adds if you have OBP and SLG:
As it turns out, after you predict a team’s runs scored using their OBP and SLG, higher batting average means lower runs scored. If that’s confusing, I’ll try to show it in graphical form. A positive residual means that OBP and SLG under-predicted a team’s actual runs scored. Thus, if higher batting average means more runs scored holding all else equal, you’d expect to see a line from the bottom left to the upper right on the below graph. Instead, as you can see from the superimposed fit line, the opposite is true:
As noted in the article, this makes a little bit of sense when you consider the types of hits making up each slash line. A guy who is getting on and racking up bases in SLG through a high AVG, low walk line is getting more singles. ISO on the .315 / .365 / .510 line is .195, which is good, but the ISO on the .260 / .365 / .510 guy is .250. The latter line is made up of more extra base hits that clear the bases and put a runner in scoring position (or, scores the guy who hits the homers). Of course, sequencing can impact scoring, so the more singles guy probably gets some RBIs if there are men on 3rd or 2d that walks don't, and this wouldn't show looking at team runs scored (again, noted in the article), which leads the author to say more or less the two lines are about the same, but that conclusion is kind of against most folks first instinct, which is to go with the higher average as the more productive line.
The comments are really good, too.
In other words: Batting average is so heavily correlated with a clearly better and more informative predictor (OBP) that it actually goes the opposite way than expected. This kind of multicollinearity makes regression interpretation super-weird; OBP is capturing the relevant variation that we care about with batting average in a much more comprehensive way, so batting average is only useful insofar as we find “empty batting average” types.
Another pulls some data sets for seasons of individuals with similar lines to the Tangotiger example:
Player A: Beltre 2013, Pence 2011, Pollock 2015, Segura 2016, Y. Molina 2012, B. Butler 2012, C. Seager 2016, Morneau 2014, Brantley 2019, D. Peralta 2015
Player B: C. Santana 2016, Willingham 2012, Harper 2019, Will Smith 2021, Y. Alonso 2017, Conforto 2019, J. Bautista 2013, Papi 2014, Carpenter 2018, E5 2017
Player A group combined slash: .314/.369/.505, with a .374 wOBA
Player B group combined slash: .260/.366/.507, with a .369 wOBA
The best part is the wRC for each group, both within 0.1 runs of 97 per 600 PA.
Both groups were slightly above the requirement for OBP and slightly below the requirement for SLG.
Some highlights of the differences:
Player A group had 42 more singles, six more doubles, and three more triples per 600 PA
Player B group had 11 more home runs, 34 more unintentional walks, and 36 more strikeouts per 600 PA
The B group had four more runs and 10 more RBI per 600 PA, although I suspect a lot of that has to do with lineup spot and team environment than anything.
The A group had five more stolen bases per 600 PA, for whatever that’s worth.
Finally, another suggests that an upgraded // line would have hits/PA, [BB + HBP]/PA, and bases/PA as a way to capture hitter quality. Given the comparison of the Player A /Player B pool and the relative importance of OBP and SLG in terms of predicting runs, I think the updated triple slash probably overstates the difference between the two lines.
