The A’s, Pitchers, and More Fun with Cascading by Jonah Keri January 17, 2011 Every year, baseball’s intelligentsia settle on a new sleeper team. The ballclub in question is usually young, with a bunch of players coming off strong seasons. Because of those players’ youth, we assume continued progress the next season, even a playoff run. The A’s have started to crystallize as 2011’s go-to sleeper team. They went from 75 wins to 81 in 2010. They’re armed with a universally-admired, young starting rotation, and have recently upgraded their offense and their bullpen. Niche blogs have jumped on the bandwagon. Sabermetrically-inclined analysts writing for mainstream sites have done the same. Buster Olney’s on the A’s too (Insider subscription required), profiling Gio Gonzalez as one of the key young stars who could propel Oakland to the postseason this year. Is everyone doing this wrong? One statistical concept you’ll frequently see discussed ’round these parts is regression to the mean. In baseball terms, if a player enjoys an extraordinary season, one well outside historical norms, odds are he’ll follow with a season more in line with his typical production. This is a trickier concept to nail down in the case of younger players. Since their track record is so short, we don’t yet know what a typical season might resemble. That’s where we are with much of the A’s roster, especially the starting rotation. Let’s take a look at some numbers for this fab four: Player Year ERA FIP xFIP Brett Anderson 2009 4.06 3.69 3.61 Trevor Cahill 2009 4.63 5.33 4.92 Gio Gonzalez 2009 5.75 4.47 4.02 Dallas Braden 2009 3.89 3.73 4.80 Player Year ERA FIP xFIP Brett Anderson 2010 2.80 3.21 3.75 Trevor Cahill 2010 2.97 4.19 4.11 Gio Gonzalez 2010 3.23 3.78 4.18 Dallas Braden 2010 3.50 3.80 4.41 None of the A’s top four starters had ever spent a full season in a starting rotation before 2009. So we’ve got two years of reliable data to go on. What does this relatively limited dataset tell us? Brett Anderson is very good. He’s made 49 major league starts, struck out more than 7 batters per 9 innings, walked just over 2, and yielded just 26 homers in 288 innings (0.81 HR/9 IP). We’ve now got two seasons suggesting he’ll allow about three and a half runs (or fewer) per nine innings after stripping out the effects of his defense, or a smidge more if we normalize his home run rate to account for Oakland’s cavernous dimensions. He’s the clear staff ace. After that? Who knows. Cahill came with a great pedigree as one of the best sinkerballer prospects in the game, someone who’d keep the ball in the park and induce a ton of groundouts. But he’s struck out just 5 batters per 9 IP so far in the big leagues, struggled with command in his rookie season, and benefited greatly from hitters pelting grounders right at fielders last year (.236 batting average on balls in play). Gio Gonzalez is a left-hander with a live fastball (average velocity 91.5 mph) and one of the deadliest curveballs in the game when he’s on (an incredible +21.1 run value last season). He also walked more than 4 batters per 9 IP in 2010, and that’s after slicing one full walk per game off his record from 2009 levels and nearly halving his HR/FB rate. Dallas Braden’s skills appear the most stable. He’ll strike out five and a half, walk two and a half, induce more flyballs than groundballs, and use his home park and his outfielders to limit the damage. When he’s on, this happens. But for the most part, he’s a reliable number-four starter. So we’ve got one excellent starting pitcher, one a tick better than average…and two guys who could be anything from fringe Cy Young candidates to slight liabilities depending on where their careers go from here (Gonzalez is 25, Cahill is 22). Projecting these four pitchers to continue their progress is an iffy proposition. The idea that position players peak in their mid-to-late 20s is a commonly-accepted concept in baseball circles. But pitchers are more unpredictable, with injuries and sudden, dramatic changes in performance more common. You can’t project a neat bell curve for hitters topping out at age 27; you can’t even begin to try for pitchers. There’s another variable making A’s prognostication even tougher. Last week, Matt Klaassen wrote about Cahill parlaying batted ball luck into a season in which his xFIP was more than a full run higher than his ERA. But Cahill was hardly alone in that discrepancy: Key relief pitchers like Andrew Bailey (1.47 ERA/3.80 xFIP), Brad Ziegler (3.26 ERA/4.36 xFIP) and others added to the gap. The A’s tied for 3rd in MLB (best in the AL) with a 3.56 ERA last season. But they ranked just 14th with a 4.24 xFIP. That was the biggest ERA-xFIP spread of any team in the American League, by a wide margin. Though xFIP does seek to normalize home run effects, neither it nor its FIP cousin account for the entirety of ballpark effects. Simply regressing A’s pitchers home run rates to league average may not be enough to override Oakland Coliseum’s extreme homer-suppression (0.701 HR factor) last year, and the park held down singles (0.923) and doubles (0.903) as well. So even when you adjust for the A’s stellar gloves (no infield fared better last year than Daric Barton, Mark Ellis, Cliff Pennington and Kevin Kouzmanoff, and Coco Crisp remains a very good fly-chaser), we’ve still got a lot of uncertainty on our hands, and the potential for regression – how much of their success was the defense and park, and how much was good fortune? It’s hard to tell, though the fact that the A’s have a team ERA of 3.96 against a 4.34 xFIP since 2002 (a sample of over 13,000 innings) suggests that we probably shouldn’t expect Oakland pitchers to post equal numbers. The A’s have moved to mitigate some of that uncertainty. They’ve upgraded their bullpen (Grant Balfour at 2 years/$8.1 million is a fine deal in this market, and Brian Fuentes is reportedly on his way) and revamped their lineup (welcome Hideki Matsui, Josh Willingham and David DeJesus). But even those moves are far from a sure thing. Matsui is 36, with diminishing power. Willingham is a defensive liability who’ll give back some of the runs he adds offensively. DeJesus has only topped 144 games played once in his career. The A’s made an aggressive run at Adrian Beltre for the second straight off-season, but again came up short; he would have offered a much easier shot at a multi-win upgrade than any of the players who ultimately signed. What we’re left with is…a whole lot of potential variance, same as last year. If the A’s are going to unseat the Rangers for the AL West crown, they’ll probably need further improvement – or at the very least continued strong performance – from their top four starters. Will it happen? If there’s one reason for A’s fans to be optimistic, it’s the same reason I expressed concern about the Brewers’ chances last week: cascading. One of my favorite pet theories (I have many; don’t ask me about teams’ rigid bullpen usage unless we’re WAY off the record) is that no matter how much we try to adjust for the effect of defense on run prevention, we can’t fully capture its impact. When Mark Ellis ranges into the hole to snag a sharp grounder and end an inning, he’s preserving his pitcher’s pitch count, helping him avoid pitching deeper into an inning under duress, and allowing him to go deeper into games, with less injury risk. If Anderson, Cahill, Gonzalez and Braden stay healthy and go deep into games, that’s fewer innings for weak replacement starters, fewer innings for the bullpen’s weakest arms, and a clearer path to Andrew Bailey and the end of a ballgame. Park factors may offer a similar cascading effect. If Dallas Braden knows he can get by without overpowering stuff, leave a ball up and know his ballpark will prevent an Earl Weaver Special, that can make it easier for him to get more outs earlier in the count, and reap the same health and bullpen benefits that a great defense provides. I would love to see more granular research on particular pitching staffs, their performance in relation to defense and home park, and if we’re missing something in our typical adjustments. At the risk of angering the small sample size gods, I think there’s plenty of interesting work left to be done.