Orioles Using Bloomberg’s Data Presentation Tool

Yesterday the Orioles held an event at Camden Yards with Bloomberg Sports to show off the Bloomberg Baseball Data Tool they’ve started using (you know, the one that makes those graphics we see on the broadcast showing that Brian Matusz throws 55% fastballs, 15% change-ups, etc.). I unfortunately forgot my notepad, so the recap is all from memory.

Dan Duquette* spoke briefly at the beginning, talking about the O’s using all the data available to improve the club (for example, with Buck Showalter employing the infield shift this year) and mentioning that part of the reason the O’s pulled their advanced scouts (who traditionally go to scout the clubs the team is about to play) is that the Bloomberg Tool provides them all kinds of data on those clubs (much of it things the scout wouldn’t even pick up).

* It was the first time I’ve seen Duquette in person and, though this sounds stupid, it seemed like he was an actor playing ‘Orioles General Manager Dan Duquette’. Maybe it’s the voice or the hair?

Next up was the head of Bloomberg Sports (who has also the guy who started SportsVision), whose name I don’t remember. He showed us the publically available Bloomberg Fantasy Baseball tool (which does stuff like suggest trades for you to make based on the strengths and weaknesses of your team as well as the other team) and, much more interestingly, the Pro tool used by Major League clubs (25 of 30, I believe).

Now that was neat. The interface seemed well built and relatively easy to use, though we mostly just looked at PitchFX data; what pitches Adam Jones has seen this year, the fastballs Jones has seen, the pitches he’s seen with 2 outs, the pitches he’s seen from Justin Verlander, etc. Pretty standard PitchFX stuff, but easy to chop up and it looked nice – clicking on a pitch that Jones hit for a home run brought up a video clip of it. There was also some cool looking pitch sequencing graphics. None of it was something that I couldn’t put together myself, but drop-down menus are just a bit easier to navigate than writing code to run against a PitchFX database and their graphics looked a touch better than what Excel would spit out. Those were only a couple tabs out of somewhere between 5 and 100 though, so there are surely more aspects to it.

Rick Peterson then gave his own presentation focusing mainly on – basically – the importance of not falling behind in counts as a pitcher. Peterson came off as a good guy, but the presentation was not the best*. One of the first slides in the PowerPoint was of batting average and slugging percentage when a ball is put into play in each count, and he said those numbers haven’t changed in 10 years. That immediately set of alarm bells, since offensive levels overall are way down from where they had been (slugging has dropped almost 30 points). Maybe he meant that the numbers were the same relatively speaking (between the counts), or maybe he hadn’t updated his table recently. Either way, at least some people took it seriously (one person started a question by referencing the point as true).

* I’m being somewhat nit-picky. Sue me.

Peterson also talked about batting average on balls in play on pitches at the bottom of the strike-zone (the “220 line”*, since the batting average there was .220). Not terribly surprising as the batting average on groundballs was .237 last year (via Baseball-Reference). He didn’t mention any sort of line for up in the strike-zone though – and the batting average on flyballs was actually even lower at .218 (obviously line-drives muck around with these distinctions). The audience probably wasn’t the right one for it, but this is where using wOBA would have been much better (or even OPS, since the gap for 2011 was almost 300 points). Liked the line about a “hot hitter” being one who’s been getting a lot of balls thigh-high in the middle of the plate recently, and a “cold hitter” is one who hasn’t.

* Apparently one of the reasons Jim Johnson has been one of the best closers in baseball this year is because his batting average on balls down in the zone is really low. Well yeah, a crazy low BABIP tends to lead to fewer runs allowed (JJ’s is .140 overall and .131 on groundballs). The thing that makes Johnson good isn’t so much that he has a low BABIP on pitches down as that he keeps pitches down (and limits free passes). Thus the 66% groundball allowing him to post a pretty good xFIP (~3.40) despite a low strike-out rate. That xFIP ranks Johnson just 55th among qualified relievers** and 13th among pitchers with at least 5 saves. though.

** A few spot behind Alfredo Simon, who’s posting a 1.91 ERA, 2.56 FIP, 3.32 xFIP line for the Reds – sure glad the Orioles cut him so they could hold on to Kevin Gregg (amongst others).

This led to a graphic saying the most important things for a pitcher are groundballs and swings and misses though, and I can certainly get behind that. I was thinking about asking him about the tension between the two – since fastballs down are more likely to get groundballs, but fastballs up are more likely to get whiffs* – but didn’t think it was worth holding things up and assumed I’d just get an answer about how you want both. Peterson said one of the reasons the Orioles staff has been so good this year is that they get a lot of groundballs and swings and misses**.

*ahem, Jim Johnson

** the O’s are 15th in groundball rate and 22nd in contact rate against (and 25th in strike-out rate).

The next bit was about the importance of the 1-1 count and either ending at bats quickly or getting ahead in the count and to 2 strikes. This is a “duh” in the same way that FIP is (“getting lots of strike-outs while not walking guys or giving up home runs is good – who knew?!”). So the batting average on 0-0, 1-0, and 0-1 counts is ~.325 – never mind that there’s a 75 point difference in slugging between 1-0 and 0-1. And while OBP after (not just on) a 1-2 count is ~.225, the OBP after a 2-1 count is ~.390. Actually, I think it was maybe 1-2, 0-2, and 2-2 versus 2-1, 2-0, and 3-1, because I recall his OBP difference being a fair bit bigger than that – closer to .220 vs. .450. In any case, the point was “Efficiency”; ending at bats quickly or getting ahead in the count. Because good pitching staffs will do those things 70+% of the time, whereas average ones will only be at ~68%. Assuming I counted correctly, the Orioles are right around average this year after being below average last year. Interestingly that matches up better with their FIP/xFIP (about average) than their ERA (better than average).

Anyway, not particularly novel or well presented from my perspective (there was some straight comparing of batting averages between, say, a certain count and pitches in a certain location, which you can’t really do, partially because some pitches will be in both buckets), but it seems like most of the people in the room got a fair bit out of it and that’s good.

Peterson also talked about the bio-mechanical analysis that team has started doing (including running all of their Major Leaguers and some prospect, including Dylan Bundy, through it during Spring Training) that’s allowed them to identify and modify pitchers’ motions. Not sure it was exactly related, but he showed an adjustment they made to John Axford when Peterson was with the Brewers to change where he stood on the rubber and the improved results*, and that you could see it on the release point data from the Pitch FX. Also, that a similar change was made with Jake Arrieta recently (18 K’s to 3 walks in 20 innings since he was supposed to go to the bullpen, by the way).

* ERA from the first half of the year to the second half fell from just over 3 to just under 2! What an amazing improvement for a reliever throwing ~30 innings in each sample! Especially since his strike-out to walk ratio actually got worse and he walked a higher proportion of batters he faced later (even though the purported reason for the change was to improve his control)!

Overall though, I liked Peterson’s emphasis on process and using all available data. When someone asked him about players’ reactions to the bio-mechanical analysis and data, he had a nice point; players have blood taken for their physicals, and if tests come back saying they have high cholesterol they don’t wave it away. Same thing – it’s science!

I asked the Bloomberg guy about if anything is done with the PitchFX data before it’s integrated into the system, but he didn’t really get where I was going so I talked to one of the behind the scenes people. Basically, it’s just raw data and the Bloomberg system is handy for doing the hard work of data mining with a nice user interface. It does allow for adjustments though, so the Orioles can, for example, select which source they want the pitch classifications to come from, hard code them to correct mistakes*, or even put together a team-created algorithm to overwrite things. That’s pretty neat.

* His example was Clayton Kershaw’s slider being labeled a curveball last year. Also brought up the neural networks used for pitch identifying with the PitchFX system, so he knew what was up – how about a presentation from that guy?

I’m not sure just how much value added there is for the club with this Bloomberg system, but I’d certainly like one to play around with. Seems like maybe it could be more useful to others in the front office than the baseball analytics department, who I’d assume (hope) already had something not tremendously different in place (that is, maybe they’re doing things a little faster or neater now, but hopefully they’re not really doing anything they couldn’t have done before). There is something to be said for not only having the data but actually using it though, and that seems like more of the direction the Orioles are going in (thankfully).

The game after the presentation was… less good. The two home runs Brian Matusz allowed were bad, but he also gave up a bunch of hits in dis-pleased BABIP God fashion (.459 on the day, with at least one ball that should have been caught in the outfield and a few grounders that got through the infield). Brian Robers laying down a sacrifice bunt with two on and no outs and the team down 3-1 in the bottom of the 5th was awful since (a) he’s not a particularly good bunter, (b) it wasn’t a surprise at all to the Angels, especially after he didn’t get it down the first couple pitches but kept trying, (c) he was ahead in the count at one point, and we know how much better hitters are when they’re ahead in the count, right?, and (d) the score was not conducive to playing for one run anyway. The Orioles rightfully didn’t score again in that inning, or in the following inning when they loaded the bases. In true 2012 style, they scored all 3 of their runs on home runs (by Steven Pearce and Wilson Betemit). The Angels were worse, picking up 17 total hits but still scoring 6 of their 7 runs via the longball (and in the case of Mark Trumbo’s shot off of Tommy Hunter, it was really long). 7 > 3 though, so the O’s fell to 41-32, 4 games back of the Yankees in the AL East.