I constructed a lineup of eight Evan Gattises and all you got was this lousy article

Editor's note: Please welcome Ivan as an official Talking Chop writer! We've all seen his glorious, in-depth comments before, and now we'll be reading them as full posts on a regular basis. Enjoy! - Jane

As Braves fans, we're all likely aware of the various narratives about how a team focused around high-strikeout, high-potential-power hitters is frustrating, unpleasant and/or doomed to mediocrity. While picking apart those narratives can be done on a case-by-case basis, that's boring. Instead, I'd like to use those narratives as a jumping-off point to explore a few similar topics that are somewhat related to the whole rah-rah-contact-OBP-slugging-etc. rigmarole.

Generally, when we throw advanced metrics acronyms all over the place in our discussions (it's all fun and games until UZR pokes out someone's eye), we acknowledge that the metrics are useful because they evaluate players on a context-neutral basis. What does that silly hyphenated word really mean? Just that players aren't going to be awarded bonuses or penalized for things not related to their own performance, like the runners on base ahead of them, the score at the time of a given plate appearance, and so on. In general, this is done to be able to more effectively evaluate the performance and value of a given player, and to allow comparisons across players, but sometimes it can abstract away from baseball as played on the field. Even if wOBA doesn't care about whether a guy's double came in the 8th inning with the bases loaded and his team down by a run, or whether there were two outs, none on, in a tie game in the top of the 1st, we probably do at least a little (assuming we're interested in the outcome of the game, and not just individual player performances, anyway).

One particular place where the reliance on context-neutral metrics can be a bit overblown is when looking at roster construction. By considering each player separately, it is possible to lose the nuance where a team may be more than (or less than) the sum of its parts. For offensive performance, the first obvious and important question should be, and generally is, "Is this guy a good hitter?" We have tools to answer that, with a really handy one being wRC+, which simply states how much better or worse a guy was than league average. (100 is league average, a value of 125 indicates a guy was 25% better than league average, while a value of 90 indicates a guy was 10% worse than league average at creating runs... again on a context-neutral basis where each of the guy's offensive results is just bundled together and compared against league-average offensive results, and then adjusted for league and park factors so that it's truly comparable across players). But another question that we may want to ask ourselves is something that's harder to answer by looking at a single metric: "How does this guy impact the lineup's ability to put runs on the scoreboard?"

I'm going to change gears a bit, because if I don't, I'm liable to ramble on about these theoretical analytic concerns all day. I'm going to focus this article on the interplay between OBP and SLG, but really, it's on the interplay between guys that mostly reach base via singles/walks, and guys that attain a greater proportion of extra-base hits. (Note: a Fredi hat tip to TC commentariat participant Dj9401, whose question in a thread was the reason I thought of this. If you don't like it, blame him, not me. Also blame Jane. Not me.) Players can't very easily adjust their results to trade off between OBP and SLG, even if they can attempt to do so by altering their plate approaches (look at BJ Upton's stats in Tampa for an okay example of focusing on SLG over OBP as his career progressed). Given this, even if we use a benchmark value of "each point of OBP is worth 1.7 points of SLG, or thereabouts," it's hard to tell a hitter that if he's going to trade one for the other, he needs to maintain some kind of golden ratio or else the tradeoff isn't worth it. After all, they're (still) human beings, not automatons. (If you weren't told there would be no math, I also recommend this excellent investigation into this issue.) But constructing a offensive roster, and a lineup, is a little different. While, at the player level, it is hard to have a player shift his offensive production away from walks and towards doubles, or away from homers and towards singles in a reliable way, I can see how this can be somewhat easier from the front office, where there's a plethora of acquisition targets to choose from, and a number of ways of building a starting eight (yes, I said eight - I can recommend some blogs for AL baseball if you prefer).

Between 2010 and 2014, 338 players accumulated at least 1,000 PAs. These players had an average OBP of .328, an average SLG of .412, and an average SLG-OBP gap of .084. Chone Figgins and Jamey Carroll managed to have a SLG of .024 points lower than their OBP, while Carlos Gonzalez, Miguel Cabrera and Nelson Cruz each posted slugging marks of .180 over their OBPs. It is important to note that, in and of itself, the OBP-SLG gap is not a definitive predictor of a good hitter or a bad hitter. While higher slugging is generally a sign of a decent hitter (because a bad hitter would have low OBP and low slugging since both are bounded at zero), it's possible to be a decent hitter even if you're a little weird. Willin Rosario, Tyler Colvin, JP Arencibia, Juan Francisco, Vernon Wells and Scott Hairston are below-average bats in the 90th percentile and above of SLG outpacing OBP, while Daric Barton and Gregor Blanco are average-ish bats despite having an SLG-OBP gap in the bottom 10 percent. There are a couple of charts below that illustrate this: the first chart shows the distribution of SLG-OBP gaps among these 338 players; the second chart plots a player's SLG-OBP gap against his wRC+ to show the imperfect relationship therein.

Just a few notes about that second chart: since it's using 2010-2014 aggregates, no, BJ Upton isn't that sad gray dot with a 40ish wRC+, and also, it's weird how remarkably similar Martin Prado and Dan Uggla have been over this period (again, in aggregate). There's definitely a relationship with an R-squared of about 0.35, but as noted above, there are lots of counterexamples both ways: good hitters with small gaps, and bad hitters with high gaps. I called out some Braves and former Braves largely because I was curious, but as far as this chart goes, it wasn't really that the prioritization of SLG over OBP really did the Braves in here: Uggla and BJ have had average gaps and average-ish offensive outcomes (again, over the whole period, guys!), and Justin Upton was a good offensive performer despite a larger-than-average gap. This isn't any kind of new revelation, but it's not like the Wren-led front office targeted all-power guys with no on-base skills by overvaluing their offensive contributions...

...unless those guys were the wrong fit for the lineup. Or, in other words, is there a critical mass of hitters in terms of their OBP and SLG skillsets after which you'd be better off diversifying? That's a hard question to answer without doing a lot of rigorous testing, though I'd guess that the answer is something like high-OBP guys (even with low SLG) and high-SLG guys (even with low OBP) are complements and you want a mix of both. Before I actually test that, I want to be sure that I'm doing it on an even playing field. A lineup of Nelson Cruzes (.332 OBP, .512 SLG, 125 wRC+) would score more runs than a lineup of Denard Spans (.337 OBP, .382 SLG, 100 wRC+) not just because Cruz slugs better, but because on balance, Cruz is a better hitter. But what about guys that are roughly equivalent hitters, on a context-neutral basis, but achieve it in different ways?

To do this, I'm going to use Evan Gattis. Specifically, 2014 Evan Gattis (no more of that five-year average stuff). Why Evan Gattis? Because I thought of this title marginally before the rest of the post, and I'm not changing it now. Evan Gattis had a 125 wRC+ last year, which he achieved via a .317 OBP (far from great, actually below average among 2014 hitters with at least 200 PAs) and a .493 SLG (93rd percentile among all hitters with at least 200 PAs last season). I'm also going to use wOBA, and not wRC+ for the next paragraph or so, because wRC+ is park-adjusted while wOBA is directly derived from at-the-plate results, and since I want to compare the aggregate effectiveness of plate results without adjusting for run environment and the like, wRC+ will only confuse the issue (if it's not clear why, give it a minute, you should hopefully get it by the end of the post).

Gattis's .352 wOBA was 57th among hitters with 200 PAs last year. What I've done below is construct a pseudo-indifference curve (repressed nightmares of everyone that's ever taken a microeconomics class come flooding back into conscious minds right about now, sorry about that) of wOBA across OBP and SLG. Check it out. Note that it's not really a curve (not enough data points), nor is it perfect (though the fit is actually really good, and just-about-perfect considering I'm using a wOBA spread of .005 instead of .000), and it shows that in order to get a Gattis-like wOBA, you can use some mixture in a spread of about .050 points of OBP and about .070 points of SLG. (Math note: the slope of the best-fit line indicates that in order to stay at this particular level of wOBA, you have to trade SLG for OBP at about a 1.4:1 ratio (this ratio will necessarily differ for other wOBA levels).

So anyway, you can see that Gattis is the sluggiest dude out of this set of guys with .350 to .355 wOBAs, while Napoli was the OBP-iest. As you'd expect, more guys are clustered in the middle, which allows there to be a decent curve/line shape. Given this, we can run simulations to compare the runs per game scored by a lineup consisting solely of these players. To do this, I'm using a little dinky lineup simulator tool I built, which uses a player's ratio of offensive outcomes (i.e., this player hit singles in this proportion of his PAs, walked in this proportion, doubled in this proportion, made an out in this proportion, etc.). It only accounts for station-to-station advances on hits, no advances on outs, no sac flies, wild pitches, double plays, etc. (I'm excluding stolen bases for this simulation, though it is possible to account for those in a simpler way than advances on outs, or double plays.) This undercounts the runs scored per game a bit, but since I'm using this for comparisons anyway at this point, it's not a huge deal. The results are shown in tabular format below. (In all results, I use a composite 2014 NL pitcher for the 9th spot. Because baseball.)

As you can see, there's not much of a clear relationship between... well... anything. Which makes sense, since we're dealing with small fractions of runs per game, and hitters that fit into various places on the continuum of OBP-SLG. It makes sense that the Todd Fraziers would be last, partly because he had the lowest wOBA, but also because he had both a below-average OBP and only an average SLG for this group. But what's harder to explain is Adrian Gonzalezes; I was expecting that the best results would come from guys in the middle of the OBP and SLG range, but instead Adrian Gonzalez has the second-lowest OBP and the second-highest SLG. So... there?

The charts below show the tiny-sample scatter for both runs per game and OBP or SLG. (Note that I didn't use the adjusted runs/game, which is really just a really lame attempt to try to control for some of the issues with not having runners move up on outs. I'm not sure how accurate it is as I haven't tested it much, I'm not sure how much those things matter.)

So given this, we know in lineups consisting of identical pretty good hitters (around 20% to 30% better than league average), OBP seems to help a bit in scoring more runs, whereas additional slugging doesn't necessarily do so. But what if we create another Franken-lineup, mixing guys together? I ran a bunch of different scenarios, mixing Napolis with Gattises, Gonzalezes with Santanas, and didn't find much that really changed the outcome. I had hoped that by pairing, for example, Napoli's .370 OBP ahead of Gattis's .493 SLG, I could get an outcome more than the sum of their parts, but in the end the results were more or less the same, and removing a Napoli for a Gattis generally marginally removed runs rather than adding them. So while it seems that SLG and OBP should really be complements in the same lineup, I'm not sure that holds, at least not for these types of players at this wOBA level.

With all that said, there's a lot of fun stuff that can still be tested. How do these relationships hold at other wOBA levels? What about other types of lineups (i.e., not lineups that consist of similarly-productive hitters)? If there are specific things you're interested in (assuming your eyes remain at least partially unglazed at thsi point), ask in the comments, it's fairly simple for me to run simulations. Or, feel free to do it yourself by sorting guys by wOBA, and using a handy online lineup simulator like this one. (Just be careful to include the correct values, H really does mean hits, and ABs really does mean ABs and not PAs, etc.)

All-in-all, I think we can take away the following, at least after this preliminary foray into this particular jungle: there's no hard-and-fast rule about the way in which a guy's wOBA will affect run scoring; the best guidance is actually his wOBA or wRC+ itself. There might be specific value to filling a specific roster or lineup hole with a guy with a specific OBP-SLG ratio, but such a relationship is not readily apparent. On top of that, the lineup simulation exercise did not reveal any dramatic spread in outcomes: the difference between the Fraziers and the Gonzalezes was only 0.4 runs per game, or a variation of 10% at maximum. While 10% is certainly important, it's hardly a revelation. If your team is composed of eight guys with a .350ish wOBA, you'll be just fine, even if they're all Todd Fraziers.