# On Framing, Platoons, and Regression to the Mean

As you might have heard, the Toronto Blue Jays and Kansas City Royals completed a seemingly inconsequential trade. Third string catcher and Erik Kratz and starter Liam Hendriks, buried so far on the depth chart you can’t even see him, were told to pack their bags and head for Kansas City. In return, the Blue Jays received noted lefty masher Danny Valencia. On the surface, this was a trade that filled needs for both teams and was even if not unimportant. But let’s go a little deeper.

Starting off with likely the least important piece in the trade, Liam Hendriks has managed to acquire the journeyman tag despite being just 25. As he heads to his third organization, he brings with him unremarkable stats that you would really only deem acceptable from someone as far down the depth chart as he was with Toronto. Your typical Quad-A player, Hendriks started the International League All-Star Game this year, but in the majors, he posted a 5.37 xFIP, translating to a 138 xFIP- in 13.1 IP. That’s more of the same for Hendriks, who has pitched to the tune of a 115 xFIP- in his career. Despite being a former top prospect in the Twins organization, he is now more than likely to be inconsequential in this trade.

Next up is the player being brought back to the Blue Jays. Danny Valencia is a third baseman (albeit, not a particularly good one) who has honed his craft with the Twins, Orioles, Red Sox and Royals before arriving in Toronto. You won’t mistake him for Brett Lawrie at third. In exactly 2584 innings at the hot corner, he owns a -5.2 UZR/150 mark. Also, 6 random games at second base. Ryan Goins replacement?

Valencia owns a 93 wRC+ at the plate on his career, but his lefty-mashing is really why he was acquired. If you look at his platoon splits, he has a MASSIVE preference to hitting lefties. In 882 career PA against righties, he’s hit for just a 67 wRC+. However, in 498 career PA against lefties, he’s absolutely destroyed lefties for a 140 wRC+. That’s a pretty solid sample, isn’t it? It would appear that Valencia could be inserted into the lineup against lefties, which the team has struggled with, and absolutely destroy them. Possibly a very useful acquisition.

Hold on a minute. Our good friend Nik wrote about true talent platoon splits before the season.

One of the most common sabermetric faux pas is the misinterpretation of a player’s platoon splits. It is very common for fans to simply look at a player’s splits from last season, or the last couple of seasons, and form an opinion on that player’s platoon skills based on that information.

So what is Nik saying? That we can’t take one season’s worth of PA (say … 498 PA?) against a certain handed pitcher at face value? Well, that’s exactly what Nik is saying.

The Book: Playing the Percentages in Baseball is a phenomenal book written by Tom Tango, Andrew Dolphin, and Mitchel Lichtman that more or less revolutionized advanced stats in baseball by touching on concepts such as clutch hitting and pitching, weighted on base average, and the very relevant to the topic at hand platoon splits, among other topics. So what does The Book have to say on platoon splits? Well, luckily for us, at the end of each section is a quick snippet summarizing the section. The following can be found on page 163 of The Book:

The Book Says:

A right-handed hitter needs around 2,000 appearances against left-handed pitchers before his measured platoon split can be considered reliable(in other words, using the measured platoon split is more accurate than assuming the player has an average split). From a practical standpoint, right-handers are best assumed to have average platoon skills, unless one is willing to make the calculations needed to accurately estimate a player’s platoon skill. For lefties, the number is 1,000, which means that only veteran starters have reliable platoon splits.

The actual number is 2200. 2200 appearances?! Well, Valencia’s 498 appearances aren’t looking so good now. That’s a 1702 PA difference. He’s just 22.6% of the way there. So what do we need to do to accurately be able to estimate how Valencia will hit lefties?

**Regression towards the mean **is the statistical phenomenon of “averaging out,” so to speak. The technical definition is the tendency of data to gravitate towards the center of a distribution if given enough sample. Pretend the league batting average is .250. If Player A starts off a season hot, going 5-10 in his first 3 games, is he likely to continue batting .500? Likewise, if Player B starts off 2-10, is he likely to continue hitting .200? The short answer is no, probably not. Over a large enough sample size, extreme data points in a set tend to move toward the average. So we could very well see Player A could go 0-10 in his next few games, and Player B go 3-10 in his next few games. Both of these players will end with a batting average of .250, having regressed towards the mean after their respective hot and cold starts. Regression isn’t quite an exact science; I just used exact numbers to illustrate the example. However, the general concept is there: extreme data points like .400 hitters and .200 hitters generally get closer to the league average over time. This is why players don’t hit .450 over an entire season.

There are many ways that we can predict regression in baseball. Useful stats such as BABIP, or batting average on balls in play, and HR/FB%, or Home Runs per Fly Ball (as a percentage) generally regress towards the mean. So if a player is hitting .380 with 17 home runs, but with a .415 BABIP (league average is around .300) and 26.0% HR/FB% (league average is around 10%), we can safely assume that he isn’t likely to continue performing as such. ERA estimators, such as xFIP, SIERA, and our own Chris Carruthers’ xxFIP and TIPS, are useful indicators that pitchers may be due for regression.

So how is any of this relevant? How does this help us identify how well Danny Valencia will perform versus left handers in the future? Well, we are going to regress his platoon splits, that’s how. We will follow Nik’s example outlined in his post.

First, we must identify his measured platoon split. Danny Valencia has posted a .381 wOBA against left handed pitching in 498 PA compared to a .273 wOBA against right handed pitching in 882 PA, coming out to a .312 wOBA. So, if we subtract .273 from .381 and divide that by .312, we find his split as a percentage: 34.6% (!). This means that in his career, Valencia has been 34.6% better against lefties than he has righties. Wowie. Valencia has just 498 career PA against lefties, so we regress his batting batting split 66.7% (roughly) towards the mean, as 1000/(1000+498) is equal to 66.7% (again, roughly).

Next, we plug in our numbers! Caution: These numbers are subject to rounding errors.

0.165 * .332 + 0.086 * .667 = .172

There’s our number! So we have found that Danny Valencia can be expected to hit 17.2% better against lefties than righties. That’s pretty solid, but how does it compare to what we were previously expecting? Well, it’s just under half as good as what we were originally expecting (34.6%). Ouch. So exactly how well can we expect Valencia to hit against lefties?

Steamer has projected Valencia for a .310 wOBA for the rest of the season, just .002 off of his career mark. Seems reasonable. So, when we apply the platoon split of 17.2%, we get… a .344 wOBA. That’s more Yan Gomes than it is Miguel Cabrera. Sure, it’s solid, but will he live up to his expectations of being a lefty *masher*? Probably not, unfortunately.

So maybe, just maybe, we need to hamper our expectations of Danny Valencia versus left handed pitching.

So we’ve established that Hendriks is pretty inconsequential and that Valencia can be a useful piece but honestly isn’t worth giving up much for. So, to find out if this was actually a good deal, we must find the value of the final piece to see if we did indeed get Valencia on the cheap. The final piece of this deal is Erik Kratz.

Erik Kratz is not a household name, and now, as he goes to Kansas City to serve as Salvy Perez’s backup, it is unlikely that he will ever become one. The 34 year old is the owner of a 78 wRC+ over 501 career PA. However, he owns just a .228 BABIP in his career. He posted 1.3 fWAR in just 157 PA in 2012, translating to about 3.7 fWAR over a full season, including a 112 wRC+. Steamer thinks he’s around an 87 wRC+ in terms of talent.

So, that’s your standard backup catcher, right? Well, no. No, it isn’t.

Let’s start off with Kratz’s defense. According to Baseball Prospectus, Kratz has provided the 12th most runs from blocking pitches. However, when you scale him to an average amount of playing time, that jumps to 4th overall in 2014. Kratz’s defense passes the eye test as well, and we can pretty safely say he’s above average.

Now let’s move to the area of pitch framing. This is a fun one. In 2013, Matt Carruth’s pitch framing numbers had Kratz ranked 17th in the league in pitch framing with 6.8 RAA, which equates to roughly .74 WAR, nearly a full win. Now, fast forward to 2014. Baseball prospectus releases their pitch framing data based on probability. Kratz ranks 24th overall in framing runs added by call. However, of the entire top 24, Kratz has the fewest framing chances of any of them! So, we must scale him to average playing time, and when we do that, we find that he jumps all the way up to 8th with 8.3 RAA! That’s right in between noted framers Russell Martin and Brian McCann, for those of you keeping track.

Now, some of you must be thinking, pitch framing? What are these new fangled numbers? How reliable are they really? Well, our good friend Chris tested this himself.

Compared ’14 framing data from @brooksbaseball @harrypav to ’13 data. R^2 through 3 games blows your mind.

Wow. WOW. After just 3 games, the data becomes reliable. This is even better than BP expected, stating that 290 chances (roughly 6 games) was the point of reliability for the data. But what about the eye test? Any excuse to use GIFs is a good one.

Here, we have Kratz while he was still with the Phillies. As you can see, Papelbon throws one well outside here. Well outside. FoxTrax makes that abundantly clear. Kratz, however, manages to pull this back into the zone ever so quietly and pick up the extra strike.

Again with the Phillies. This pitch is low, and unfortunately, there’s no pitch tracker. However, it’s pretty apparent that the pitch is low. If you watch closely, you can see Kratz flick his glove right back into the zone to steal the extra strike.

Unfortunately, those were the only two GIFs I could find. However, Kratz passes the eye test AND the data test. He appears to be a genuinely above average framer.

So, let’s put everything together and attempt to discern Kratz’s value. In 2013, players average 3.2 PA/G. This will be important later. His baserunning is negligible. -0.5 in his career over 163 games which is basically a full season. Next up, his fielding. Kratz has posted 6.2 Fld in his 163 games. If we’re keeping track, we’re at 5.7 runs above average. On to the bat. Steamer, ZiPS, and Oliver all place him around a .304 wOBA. If we turn that into wRAA, we get -4.8 runs per 150 games. So, adjusting everything to 150 games, we’re left at +0.4 RAA. Now, we need to calculate his framing runs above average. Over his career, Kratz has been worth a total of 28.9 runs above average. Scaled to 150 games, that’s 26.6 runs! So we’re left at exactly 27 runs per 150 games.

But we aren’t done yet. We need to calculate replacement level runs so that we’re calculating runs above **replacement. **Fangraphs places replacement level as 17.5 runs per 600 PA. To turn that into games, we simply find the runs/PA, which is a number with a lot of decimals, and multiply it by 3.2, and then by 150. We’re left with 14 runs. Added on to our previous total of 27 is 41 runs above average! On to our last component: positional adjustment. Being a catcher is worth 12.5 runs per 162 games. So, adjusted to 150 games is 11.6 runs above average. We end up with 52.6 runs above average! For Erik Kratz!

However, we need to adjust this again. 150 games is nice for normal players, but catchers are under tons of physical wear and tear and stress. The league leaders in catcher games played last year were all around 100. That’s a nice round number. So we end up with 35.0 runs. If we now adjust this based on 2014’s runs per win figure, 9.175, we come out to **3.8 wins above replacement level per 100 games. **Wow! All this for a (supposed) backup catcher!

Now, let’s check out Nik’s post on the value of replacement level.

Let’s assume that $/WAR growth this year has been the same as the mean per annum increase across our entire sample, +9.81%. This gives us a current cost per win of $7.72M. We will drop this to a nice, round figure of $7.5M. For subsequent seasons of long term contracts, we will apply $/WAR growth of +9% per annum.

These will be the value calculation figures that we will use at Breaking Blue for all transaction analysis: $7.5M $/WAR in 2014 with 9% $/WAR growth per future season.

So we can see in this excellent piece that the value of a win is about $7.5M. If we take Kratz’s 3.8 wins, that gives us a whopping $28.5M in value! According to fangraphs, he’s making $500k this year, so we can say that Kratz’s surplus value is about $28M! Now, if we take 0.5 wins off of Kratz each of the years that he’s controllable for, we get 3.8 wins for this year, 3.3 wins for the year after, 2.7, and then 2.2. When you total this, multiply by the $7.5M figure, and subtract the league minimum salary for every year, you get…

**$89.5 million.**

He is 34 though. So maybe the aging curve should be steeper. Take off 0.6 wins per year? $85M value. 0.7? $80.5M. Take off an entire win? $67M. What if we take off 1.5 wins per year? $44.5M. In fact, you have to go all the way down to taking off *2.5 wins per year to dip into negative surplus value. *And even then, it’s only $-500k in surplus value. That’s just insanity.

So … Erik Kratz, huh?

This isn’t the first time we’ve heard about Kratz’s skill, either. Chris wrote about him on March 20th. He noted that Kratz’s framing and blocking needed more recognition. But credit for identifying the backstop must go to Marc Hulet of Fangraphs.com. He identified Kratz’s defense as above average on July 16, 2009, five full years earlier.

For whatever it’s worth, Kratz’s regressed platoon split against lefties comes out to a .317 wOBA, a league average bat that would have worked in a platoon with Lind and Navarro that had Kratz at C and Navarro at DH versus lefties. When you factor in framing, this may have been every bit as valuable as Valencia.

Erik Kratz isn’t your typical catcher. Standing massively at 6’4″, 255, you’d expect to see him at first base. But Kratz is a catcher. A good one, too. Hopefully, one day, a team will realize.

## Recent Comments