This is the second installment of the primer series. To read about why and how we're doing this, you can read this. For last week's post on WAR in general, head here.
Everyday Statistics and Their Flaws
Stats We're Talking About: BA, HR, RBI ... and sometimes SB
Batting average gets a bad rap these days. There's really nothing intrinsically wrong with it. It just gets used inappropriately. On its face, batting average is essentially how good of a hitter the player is when he doesn't walk. That's actually a pretty nice thing to know as you can't give him credit for "hitting" when he walks, and it gives us a walk-independent view of a player's hitting ability.
The issue, of course, is that it's frequently used to talk about a player's overall contributions at the plate, which is a different question. That question needs to include walks. No one will argue that walks are as good as hits overall, but they really aren't that far away from being a single. The hardest thing to do at the plate is get to first base. Walks matter. A lot. And batting average simply acts as though they don't matter, which isn't the case.
The reason batting average exists this way is because Henry Chadwick, the man who created the box score, considered walks to be unmanly and more of an indication of a pitcher's failure than a hitter's success. At that point in baseball history, it was. Pitchers mostly focused on throwing strikes, and hitters wailed away. When a hitter walked, it was basically because the pitcher didn't throw the ball near the plate. Hitters didn't focus on drawing walks. But they do now because we know walks are very important, which is just something they didn't - and really couldn't - know then.
Batting average also has the issue of not differentiating between hits. Doubles and home runs are obviously better than singles, but batting average doesn't try to differentiate. This is often why home runs are brought in - it adds the element of power - but the problem with that is that it completely ignores doubles and triples, which it really shouldn't. Those matter. A lot.
RBIs also get a pretty bad rap. They aren't pointless. It's just that they are more of a one-game stat than a season stat. RBIs will likely give you a pretty good indication of how a guy performed or contributed that night, but there's too much context - who hits in front of them, when they hit HR, where they hit in the order, playing time - that has to be taken into consideration over a full season. RBIs are more of a storytelling stat for a particular game than an indicative performance stat. The criticism should be levied more that RBI is a stat far down on the totem pole for offensive analysis than it simply being useless. Yes, a lot of RBIs generally means you did pretty well, but there are so many better tools to use.
Finally, I'll make a quick note about SB. It occasionally gets used as a baserunning stand-in. But A) it's usually only when the guy in question has a lot of them and/or B) it doesn't ask about the rest of the time a player is on base. Yes, stealing bases is generally good - as long as you're successful more than 70% of the time - but it matters what you do the rest of the time as well.
Essentially, the main issue with these stats is that they're straight counting stats - you just tally the events that happen - but they don't count everything that "counts". The biggest reason for that is that the math/technology wasn't available for most of baseball history. It is now, though.
Nuanced Statistics and Why
Stats We're Talking About: wRAA, BsR, and oWAR
As we discussed last week, offensive WAR is about batting skill over the number of PA and baserunning. wRAA (Weighted Runs Above Average) is how we figure out the hitting or at-the-plate production of a player. The components of wRAA basically amount to wOBA (Weighted On-Base Average) and PA. We'll talk more about wOBA and other offensive stats (like TrueAverage and wRC+) in a few weeks, but for now, you need to know that wOBA uses linear weights to give us run values for walks, singles, doubles, triples, and home runs - even reached on errors (ROE). They scale it to look like OBP so that the numbers are familiar, but it uses all plate appearances and differentiates between the various events.
What happens next is relatively simple. You plug wOBA into this formula - (wOBA - league-average wOBA) / wOBA coefficient x PA. For example, let's use Freddie Freeman. Freeman's wOBA was .387 (very good), so let's plug it in - (.387 - .314) / 1.26 x 629 = 36.9 wRAA (Author's Note: the wOBA coefficient changes every year depending on the run environment, and 2013 was probably lower than the coefficient I used, which is why 36.9 doesn't exactly match the actual 35.3). What this means is that Freeman produced about 35 more runs than the average MLB hitter in 2013. What does that mean? Let's take a look at the ranges:
- 40+ - Elite
- 21-39 - Excellent
- 11-20 - Above Average
- (-10)-10 - Average
- -11-(-20) - Below-Average
- -21+ - Woof
This puts Freeman pretty much in the "Elite" category, and he's certainly one of the better hitters in the game (or at least was in 2013). Of course, that 35.3 doesn't match the 34.0 offensive runs Freeman produced. So how do we get there?
The other step is looking at baserunning, which we'll use BsR (Baserunning Runs) for. This is basically a measurement of taking the extra base and not getting thrown out while on the bases, which is what good baserunners do. Stolen bases are certainly an element of this - taking an extra base or being thrown out while trying to - but BsR and other baserunning stats also give credit for going 1st to 3rd, 2nd to home, 1st to home, and even taking bases on hits. It calculates this against the averages for all players, and compares your player against the rest. There are runs values given to these events based on linear weights and run expectancies.
One thing to note about BsR and other similar stats is that there is a difference between being "a good baserunner" and a "valuable baserunner". Being a "good baserunner" is being smart, not getting thrown out, and knowing when you can take an extra base. Being a "valuable baserunner" means not getting thrown out much and taking more extra bases than your peers. Do you see the difference? When looking at Freeman, he scores a -1.2 because his speed (or lack thereof) prevents him from taking many extra bases, but he's not necessarily a "bad baserunner" that gets thrown out a lot. He may not get thrown out a lot, but he doesn't take many extra bases relative to his peers, which is why his baserunning value is below-average.
You'll note that the 35.3 minus the 1.2 baserunning runs get us to Freeman's 34.0 offensive runs. You'll also note that this is RELATIVE TO THE LEAGUE AVERAGE and not Replacement Level. If you go to the bottom of Freeman's FG page, you'll note that there's a replacement level adjustment. They add this on to help account for playing time, but it doesn't account for position. That comes later. But the Replacement Level adjustment is how we go from "relative to the MLB average" to "replacement level".
Going from oWAR (Offensive Wins Above Replacement) to general WAR requires simply adding those offensive runs and replacement level runs to the player's defensive runs and position adjustment runs (next week!). For Freeman, you'll see his positional runs are -11.1, but the positional adjustment for 1B is -12.5 runs. Well, Freeman played in 147 games of 162, which is 91% of the season, and 11.1/12.5 is 89% - the discrepancy is due to the actual calculation using innings played at the position instead of games, but this is close enough for government work.
So let's recap. oWAR is basically a collection of a player's contributions at the plate added to what he does on the bases, with an adjustment for replacement level. So Freeman was worth 34 offensive runs and 18 runs of replacement level to get him to 52 offensive runs, and at 10 runs a win, he's worth about 5.2 wins on offense.
One thing to keep in mind is that this is essentially a FanGraphs example. Baseball Prospectus and Baseball-Reference have their oWARs that have their own specific calculations, stats, etc., but the idea remains the same - add up a player's contributions at the plate and on the bases.
It's also important to remember that oWAR works similarly to WAR in that it's more of a starting point. You'll have to break it down and use context to get your answers.
What's Yet to Accomplish
Offense is rarely debated in the sabermetric community. Whether you use one metric or another, the leaderboards look pretty similar from an offensive perspective. The reason is pretty simple - offense is easy to look at and measure. You either got out, or you didn't. Offense, however, isn't completely figured out.
The first part is something I alluded to last week - production vs. talent. oWAR does a good job of measuring what happened, or production, but it doesn't directly measure talent level. Eventually, oWAR will get there, but we need a few seasons of data. There is still no xoWAR or Expected Offensive WAR based on peripherals, usual or expected BABIP, etc. that would regress out some of the luck and random variation. Production and talent are both good answers, but they are answers to different questions. It would be nice to have a metric for both.
oWAR also doesn't account for parks. It's context-neutral, and we also talked about this last week. Again, this answers an important question - what did Player A contribute to his team - but it would be nice to have one that adjusted for park and answered a different question - how "would" this player have performed elsewhere in a more neutral environment.
The last part is a nagging issue for me. Defense - which we'll talk about next week so don't use the comments as a personal sound off on these stats; do that next week - is still under much scrutiny, and while it remains so, offense remains under scrutiny as well, in my opinion. Batting, pitching, and defense are all inter-related, and until they are all "solved", they all remain somewhat suspect. Until we know more about how defense affects offense, there are still questions about everything. So I'm not as convinced as some of the accuracy of offensive metrics. Don't take that as "He said they're terrible!" I'm simply saying that I think there could be some hidden lessons elsewhere, and we can't assume that offensive statistics are perfect. Either way, they're way better than what we were using.
Central Lessons
- oWAR is the offensive component to a position player's overall WAR.
- It's composed of their production at the plate (wRAA) and baserunning value (BsR).
- Everyday stats like BA and RBI are better for telling a game story or answering very specific questions than overall analysis.