The Diamondbacks and analytics

I think its safe to say the Diamondbacks have sent out "somewhat mixed" signals regarding the role of baseball analytics, and the importance they will play for the team going forward. On the one hand, they created the post of Director of Baseball Analytics and Research, and hired Dr. Ed Lewis, who had previously worked with Chief Baseball Officer Tony La Russa in St. Louis. But there appear to be sharp demarcation lines drawn, marking the boundaries beyond which sabermetrics shall not pass. Witness the following quotes from front-office staff regarding advanced baseball statistical analysus:

• "La Russa said he thinks that analytics is one aspect of the D-backs' baseball operations department that needs to be improved upon and that there is great value to numbers" -- Steve Gilbert
• "I want to know about it and I'm going to educate myself in it." -- Dave Stewart
• "It stops before the first pitch is thrown...  It's not that we devalue it. We value it when it's used appropriately. We do not value its intrusion into the game." -- Tony LaRussa
• ""I think James [Shields] is a throwback guy by the way he goes about his business and the innings he pitches. I think the fact that Tony is here and that we have more baseball people – he probably sees us as a true baseball team vs some of the other teams out here that are geared more toward analytics and those type of things." -- Dave Stewart

The last statement in particular, anointing us as "a true baseball team," resulted in a lot of sniping from the expected sources, though there was a very well-reasoned defense of it from, of all places, The Guardian [It's an English newspaper, so I'm frankly impressed they didn't call the game "rounders" or "pseudo-cricket"] But it's interesting to contrast the Diamondbacks' approach, which appears to leave in-game decisions entirely in the hands of the manager, with that of the Dodgers (under former Wall Streeter, Andrew Friedman), whose manager Don Mattingly said bluntly, ""If you’re not using analytics as part of your decision-making process, you’re a fool."

#ShotsFired, And this week, Tony La Russa returned them with a vengeance on Arizona Sports, again showing a firm belief that sabermetric stops at the dugout steps. "Once the analytics intrude into the competition -- by that I mean if your manager or your head coach somehow has his hands tied because the organization believes that strikeouts don't matter, the manufacturing game, hit-and-run, sacrifice, you don't want to lose outs, you want to handle the bullpen according to some organized printout... It's really an important preparation tool. But if you let it interfere with the decision making of your manager and coaches, you're going to be easier to beat."

In favor of La Russa's approach, you need only look at the picture accompanying the Arizona Sports piece, which show more bling than I've seen on anyone since Liberace, in the shape of his multiple World Series rings. But some of the stuff he's saying appears to be flat-out wrong, for instance his apparent belief that strikeouts "matter". We've been taking the contrary position here, going back to the days of Jerry DiPoto in 2010. That year, the D-backs set an all-time record for the NL, fanning 1,529 times, but still scored more than league average. This season, we struck out 24% less (only the Cardinals had fewer), yet finished 11th in the league for runs scored.

Unfortunately, going by La Russa's comments, it appears that he, Chip Hale and new hitting coach Mark Grace are all on the same page. Remember one of Grace's mantras from his commentating days: "Good things happen when you put the ball in play"? Unfortunately, so do bad things - they don't get much worse than the double-play, a rare occurrence with the strikeout. This largely counters the positive impact of the "productive out", which is rarer than you'd think. Last year, in the majors, just 2.86% of all PAs were productive outs as defined by the Elias Sports Bureau: 5,263 in total. But there were 3,609 double-plays - each of which cost two outs.

If you look at game situations, you can see what actually happened, in terms of runs scored, before and after each kind of out, and figure out their cost. And it turns out, strikeouts are basically indistinguishable from other outs, in terms of their impact on the game: Our friends at Beyond the Box Score tell us that in 2013, a strikeout was a couple of hundredths of a run worse than a non-K out. To quote a line attributed to physicist Niels Bohr, this works, whether you believe in it or not. Although in La Russa's defense, the difference was a lot more during his "formative years" - about five times as much, in the late seventies as now.

I'm certainly in favor of the team's apparent support for analytics in terms of roster constructions, player analysis, etc. Indeed, given the choice, it's probably better this way, than the other way around, with strong analytics support only in the dugout. I tend to the view that 95% of managerial in-game decisions are obvious ones, and the difference between the best and worst is unlikely to be more than a couple of wins a year [to me, the "intangible" aspect of management - handling the players to get the most out of them - is a much more important skill than deciding whether or not to bunt] But there's no reason it has to be an either-or thing.

For instance, last year, there were 19 successful sacrifice bunt attempts by Diamondbacks position players, 14 of which actually reduced our win expectancy - because that's generally what sacrifice bunts do. Now, there is a place for the bunt, particularly for speedy hitters, with a good chance of legging out an infield hit [And it's not implausible it may be increasing in importance, particularly for batters to whom a heavy defensive shift is applied. Drop a few squibs down towards third, stroll to first, and that shift will quickly be reconsidered] But we had Jordan Pacheco - last stolen-base, 2012 - bunt once. In the top of the fifth inning.

Did Gibson not know the numbers? Or did he just "trust his gut" in situations where, even successfully executed, a bunt made the Diamondbacks less likely to win? Either way, as Dave Cameron put it, "At the end of the day, most teams are still entrusting their in-game strategy to people who simply don’t understand the basic probabilities of the sport." [For amusement, I also looked at what analytics advocate Don Mattingly did with the Dodgers: 15 position player bunts, 13 of which reduced LA's win expectancy. Please keep on doing so.] That's why La Russa's apparent attitude doesn't make sense, because analytics can help with in-game management.

I can see his point, when he says "The game is so dynamic, it changes from inning to inning. So you have to allow your manager and coaches to adjust.." The stats may say a bunt is a poor choice, but if you see there is a defensive shift on, and a player can handle the bat well, it's definitely worth considering. However, analytics can, and I'd say should, still inform that decision. La Russa seems even to be ruling this out, saying it would "interfere with the decision making of your manager," which appears to me a needless handicap on Chip Hale. The more knowledge you have, the better-educated your decision will be, and the greater the chance of long-term success.

Adding a final twist, I note that La Russa and Stewart are both scheduled to speak at the 2015 SABR conference, here in Phoenix next month, which is about as geeky and stat-heady an event as imaginable [and I speak as a severe stat-head, of course]. So it could be that the entire topic is some kind of elaborate trolling operation, designed to confuse the opposition. It's certainly confusing me, and I wouldn't put it past the fiendish mind of La Russa - who, don't forget, has hit his pitcher eighth more often than ever other manager during the live-ball era combined, in defiance of tradition, and in line with sabermetric thinking.

We'll see what happens as the season unfolds.