BP Comment Quick Links



July 13, 2009 Prospectus Idol EntryA Brave New World of Pitcher UsageTo read Ken Funck's Unfiltered post following up on one of the audience's suggested topics, surf here. Very few things in baseball spark as many discussions, or cause as many disagreements, as pitcher usage patterns. Closer usage, 8man bullpens, 5man rotations, pitch counts, the injury nexus—all these topics receive frequent (and deserved) attention both in the mainstream media and in sabermetric circles. However, there’s one aspect of pitcher usage which seems to rarely be questioned: the use of a single starting pitcher to work as deep into a game as possible, resting 35 days, then pitching again. Yet it may well be that this approach, the bedrock upon which all other usage is constructed, is less than ideal if your goal is to win more baseball games. In The Book: Playing The Percentages in Baseball, one of the many statistical points authors Tom Tango, Mitchel Lichtman and Andrew Dolphin make clear is the decreased effectiveness starting pitchers exhibit during each successive spin through an opposing lineup. Their careful excavation of pitching numbers from the 19992002 seasons unearthed a number of interesting finds, including a fairly steep dropoff in effectiveness after the 11th batter faced, and a surprising improvement during the 4th time through a batting order – likely explained by the fact that only starters with good stuff that day will still be pitching at that point. The chart below shows similar data, but for the 20062008 seasons (many thanks to Bil Burke for harvesting the ingredients): Starting Pitchers* 20062008 Seasons Time Through Order Games PA AVG/ OBP/ SLG OPS wOBA** 1 14,575 130,882 .262/.326/.415 .741 .328 2 14,459 127,740 .273/.335/.436 .771 .339 3 13,544 96,999 .285/.347/.460 .807 .352 4 4,325 11,355 .289/.355/.448 .803 .347 5 20 28 .500/.500/.708 1.208 .498 Total 14,575 367,004 .273/.336/.435 .771 .339 *Includes only data for games that the pitcher started. **Calculated using average linear weights for the 200608 seasons combined, including runners that reached on error, but not stolen base or caught stealing totals. As you can see, effectiveness tends to erode significantly with each successive trip through an opposing lineup. All three slash stats take a hit, as does pitcher effectiveness measured by the opposing batters’ cumulative wOBA (Tango’s measure of batter productivity, similar to EqA but scaled to look more like OBP—a higher wOBA score means less effective pitching). For comparison, during these three seasons the league average wOBA was in the neighborhood of .330—so in aggregate, once starters begin their second pass through the batting order they have become belowaverage major league pitchers. So why do we let starters pitch that long? Well, not all starters are made the same—since this chart contains all starts during those three seasons, it is sure to include spotstarters and other pitchers who quickly demonstrate they’re not capable of staying in a major league rotation. Perhaps that group is a significant drag on these numbers, and more "normal" starters wouldn’t experience this same loss of effectiveness. To test that theory, the chart below shows the same metrics, but only for the 143 starters who pitched 40+ starts in total during those three seasons: Starting Pitchers With 40+ Starts 20062008 Seasons Time Through Order Games PA AVG/ OBP/ SLG OPS wOBA 1 10,287 92,423 .258/.319/.406 .725 .322 2 10,222 90,893 .266/.327/.421 .748 .330 3 9,777 73,265 .279/.339/.449 .788 .344 4 3,615 9,674 .288/.352/.447 .800 .344 5 18 26 .478/.462/.565 1.027 .432 Total 10,287 266,281 .268/.328/.425 .753 .331 This group accounted for a little over 70% of starts, and put up better numbers—yet while we now see roughly leagueaverage pitching during their second trip through the lineup, there’s still a huge decline after the batting order turns over a second time. This isn’t true for every single starter—23 pitchers (16%) actually saw their wOBA decrease by 10 or more points during their third time through the order compared to the aggregate of their first two trips. But fully twothirds of the sample (96 starters) saw their wOBA increase by 5 or more points, and nearly half (71 starters) experienced a whopping 20+ point increase. Why is this? Fatigue is one possibility. Or perhaps batters, trying to develop a good approach against a single pitcher, retain more information between plate appearances than the pitcher and catcher, who have to develop an effective pitch sequence against 9 different batters. Regardless of the reason, this loss of effectiveness appears to be both pervasive and significant. Given all this, it seems to be a bad idea to let hitters see a starter a third time—meaning starters shouldn’t face many more than 18 batters. The average inning is made up of approximately 4.33 plate appearances, so perhaps on average a starter shouldn’t go more than 4 innings. But that introduces another problem: starters are generally considered to be the best pitchers on a major league staff, and are expected to accumulate high innings pitched totals. Even if teams moved to a fourman rotation, 4inning starts would mean only 160 innings for a team’s ace starter, with the remaining innings allocated to lesser pitchers. To enable starters to continue to have large workloads while minimizing innings pitched per game would require a complete reworking of usage patterns. One method would be to follow a tandem starter routine that, for brevity’s sake, I’m going to call SOMA: Shorter Outings, More Appearances. Under SOMA, starters would be paired up to pitch every third day, tossing 34 innings each per game. After accounting for off days, SOMA would allow a team’s best starters to appear in around 60 games and rack up 200240 innings per season—similar to their current workloads—with minimal impact on bullpen usage. Obviously SOMA would be a big change and the potential costs—physical, emotional, and economic—inherent in such a plan might be high. So for a team to consider SOMA they would need to see a tangible benefit on the other side of the ledger: Starting Pitchers 20062008 Seasons Time Through League SOMA SOMA Order Games PA wOBA wOBA wRAA* PA wRAA 1 14,575 130,882 .328 .330 184 262,350 369 2 14,459 127,740 .339 .330 960 104,654 786 3 13,544 96,999 .352 .330 1,797 0 0 4 4,325 11,355 .347 .330 161 0 0 5 20 28 .498 .330 4 0 0 Total 14,575 367,004 .339 .330 2,738 367,004 417 *wRAA is the number of runs above average allowed, based on wOBA and number of plate appearances. A wOBA below league average results in a negative wRAA (i.e., fewer runs allowed than average). The wRAA columns above take wOBA and plate appearances and convert them into runs. The last two columns calculate what might happen under the SOMA usage pattern. Under SOMA, each tandem starter would almost always go through the order at least once. This means that each tandem starter would have 9 plate appearances where the hitter is seeing them for the first time—18 per game for the tandem. Multiply those 18 PAs by 14,575 games, and you get 262,350 Pas—the first number you see in the "SOMA PA" column. Since we’re assuming starters would maintain the same seasonal workload, we can merely assign the remaining 104,654 PAs to the "2^{nd} time through" row. By substituting SOMA PA counts for the actual PAs, we can calculate SOMA wRAA values: the total number of runs above average that we might expect to give up under SOMA. The suggestion here is that normal usage patterns resulted in starters allowing 2,738 runs more than the average pitcher would have allowed. SOMA, by ensuring that starters would not go through a lineup more than twice, allowed only 417 runs above average. Thus SOMA might have prevented a total of 2,321 runs during those three seasons, for all starters for all teams (rounding used in the chart above has been removed below): Runs/Wins Benefit of SOMA Usage
Current wRAA Total: 2,737.53
SOMA wRAA Total: 417.47
Total Runs Saved by SOMA: 2,320.06
Runs Saved Per Season: 773.35
Runs Saved Per Team: 25.78
Wins Per Season: 2.58
SOMA appears to save a little over 25 runs per team per season. Since 10 runs equal 1 win, we can estimate that a team might be able to win 23 more games per year by adopting SOMA—not a huge number, but not trivial—assuming that starters could adapt to such a usage pattern. There are a number of other tertiary benefits of SOMA that are more difficult to quantify but might lead to an even greater runs/wins benefit, especially for early adopters:
Taken together, these factors could increase the advantage of SOMA to the point where it would be worthwhile to try. Of course, implementing SOMA would be a huge undertaking with significant risks, and likely would be viewed negatively by an industry that is already heavily invested in the current paradigm—although the use of tandem starters is not a completely foreign concept to major league teams. The St. Louis Cardinals are one of several franchises that implement a tandem starter system in the lower levels of their minor league system, with paired starters on a strict pitch count working every fourth game. Each player will start one game, then come on in relief of the other starter during their next tandem turn. John Vuch, the Cardinals’ Director of Minor League Operations, explains that the purpose of this is to help determine what role each pitcher is best suited for, and to avoid injuring young arms. "You spread usage out among your top pitching candidates and they get experience both working as a starter and as a reliever. They get experience pitching later in the game, and they learn different methods of warming up. It gives us a few months to evaluate guys and determine which are better suited to go into the starting role. It also helps keep the cumulative innings for the year down." While tandem starting in the minors has a very different purpose, and doesn’t tell us much about the ability of starters to adapt to a SOMA pattern, Vuch does highlight one major emotional impediment to the use of tandem starters—the definition of the Win statistic. "Say their pitch count is 75 pitches for their tandem portion as a starter. They’re trying to be as efficient as possible so they can get through 5 innings and get credit for the win. You hate to see guys gear their performance around statistics, but that’s the reality." Building in this incentive for prospects to pitch efficiently might be beneficial to the Cardinals’ player development process. But at the major league level, where the Win statistic can help determine the size of your contract, it’s clear that a tandem starter approach would be met with widespread resistance. St. Louis manager Tony La Russa and pitching coach Dave Duncan are fully aware of this issue, as they actually experimented with a SOMA variant when both were in Oakland. For one long, strange week in July, the lastplace A’s tried grouping nine pitchers into three trios who would each throw 4060 pitches every third day. After winning only one game during the experiment it was scrapped, with the skipper admitting that the Win statistic helped scupper the plan. "We went through the thing two times," La Russa said at the time. "We got some information. I think it’s got some value, but I’m very uncomfortable knowing the starting pitcher is going out there without a chance to win. I don’t think that’s real healthy."
In addition to the emotional resistance to SOMA, there’s the open question of whether pitchers can be trained to pitch that many innings spread over that many games. The last man to do so was Dr. Mike Marshall, who in 1974 tossed 208 BP’s own Will Carroll is also unsure whether starters could adapt to SOMA, because we’re still learning how pitchers truly respond to current patterns. "The fact is that we really don’t know how a modern pitcher would react, because we only know what a pitcher can or can’t do in retrospect. There’s no logical development pattern that allows us to say 'he can do this' or 'he can’t do this.' How many pitches/innings/anything can a pitcher throw today? No idea, other than broad generalizations like 'Mo can’t go three days in a row.'" And since we don’t know how much fatigue a pitcher would experience when following SOMA, Carroll feels it would be impossible to determine its effect on injuries. "I’d really love to see a simulation… to try and see fatigue patterns, to see what the stats might look like, and more importantly whether there are issues we’re not thinking of.. Why not use the technologies we have to work out as many kinks as possible?" Without knowing the physical risks involved, it’s hard to say whether the benefits of implementing SOMA would outweigh the costs. But one thing is certain: convincing a franchise to take a chance and find out would be a difficult task. "I don’t think it can be done in one year or at one level," Carroll believes. "It has to be a massive, sweeping change with complete organizational buyin. I don’t know of any organization that would or could do that." Yet, significant change does occasionally occur in baseball, sometimes for better, sometimes for worse. Call me silly (you wouldn’t be the first), but I can imagine a situation where a franchise – maybe a small market NL club, where the rewards might be greatest—with an intellectually curious front office might be willing to take a chance and at least investigate whether SOMA really can increase starter effectiveness and help win ballgames. And if they’re successful, SOMA might truly help baseball enter a brave new world. Many thanks to John Vuch, Dr. Mike Marshall, and Will Carroll for their insight.
Ken Funck is an author of Baseball Prospectus. 57 comments have been left for this article.

With the obvious caveat that Ken interviewed me for the article, I have to say that I really like that we went out and spoke with experts ... err, two experts and me. He comes up with an intriguing model and one I'd be curious to see simulated out. (I bet someone has tried this in Strat or Diamond Mind at some point.) There could be more work here  maybe a team using this SOMA plan could come up with SOMA wins (change def to 3 innings instead of 5) or SOMA QS. I'd be interested to see what the final stats for a season would look like, especially saves.
Which all goes to say that Ken has come up with something here that really fired up my imagination. Yes, pitching and pitching patterns is something I've been interested in a long time, so YMMV. Even if it's not your bag, baby, there's still a lot here to like  he has a great process of thought, an original topic, a unique voice to his writing that he doesn't let go over the top as he did earlier in the contest, an ability to dial a phone and to work in outside input. It's not only great work, it's his best work of the contest.