January 7, 2010
Checking the Numbers
Low Risk, Any Reward?
Back in my sabermetrics infancy at the "Statistically Speaking" blog, one of the first articles I wrote shared the title above and investigated low-risk pitcher signings from 2002-07 in an attempt to determine how frequently these moves positively affected teams, as well as the rate of dollars per win doled out to the players in question. The article was published in SABR's By the Numbers, and in the most recent Baseball Research Journal. The initial finding was that the wins provided by low-risk pitchers cost approximately $2 million, against the expected expense double that (if not more) per win. Those who signed at minimal risk in terms of contract value but who came with greater risk in terms of expected performance were better buys than the supposed sure things.
Looking back on it recently, I realize the initial article suffered from a few easily correctible flaws, which make for a much more accurate depiction of the low-risk free-agency market. (There were also a bit more substantial than calling the 2004 Expos the Nationals, which someone belatedly informed me of in an odd call at 9:34 p.m. last Wednesday.)
Before diving into the pitchers, let's step back and review: What are low-risk moves? Consider a liquidating video store, like the Blockbuster near my house, where I purchased 24 DVDs for $41 this past weekend. Some of the movies will be clunkers, but given my habit of watching nearly a movie per night, the potential for entertainment upside surely outweighed the risk of plopping down $1.19 for Flash of Genius with Greg Kinnear, or $0.75 for August with Josh Hartnett, both of which could be wastes of time. Movies of this ilk don't ooze Academy Award credentials, but they are certainly viable alternatives to the Comcast On Demand Free Movies library; the low cost involved earned the DVDs an invitation to my drawer, even if they never leave their boxes.
That thinking works in baseball as well. When the Rockies bring aboard a guy like Glendon Rusch, the hope is that he becomes a useful part of the staff, but the underlying expectation is that signing him barely dents your bank account; should he live up to his best-case potential, the team earns plenty on its investment. In baseball, low-risk purchases are those one-year deals with little guaranteed money involved, or a cheap option for a second year, minor-league deals laden with incentives, nabbed on a waiver claim, or some combination of these. The players themselves are risky; their contracts are not. Players ineligible for free agency are not included, though there are situations where non-tendered players merited inclusion, like Scott Olsen when I revisit this piece around this time next year.
Admittedly, my definition is fairly loose and subjective, but I manually sifted through all of the moves made before and during the 2009 season. For instance, Scott Eyre signing a $2 million deal to be the Phillies' lefty specialist out of the bullpen does not qualify, while the Mariners inking of Russell Branyan to a $1.4 million deal with incentives would qualify. The rationale here is that Eyre will wind up involved in a very small percentage of team events compared to an everyday player like Branyan, and similar pitchers could be found for less. Unlike the original article, which dealt specifically with pitchers, I checked this time around for all low-risk moves, finding 68 in total, with examples that included the Cardinals signing Trever Miller, the Rangers signing Eddie Guardado, and Bruce Chen signing with the Royals.
One of the flaws in my first attempt involved incentives. Russell Branyan didn't sign for just $1.4 million; he was also capable of earning up to $350,000 in playing time bonuses. I'd initially ignored incentives, because the goal was to gauge intent in signing a player, which lent itself to including just the base. This was and is incorrect, as some of the risk shifted to the player though incentive-laden contracts should be accounted for. So, I implemented an expected value method, in which the rate of incentives triggered out of total possible incentives was applied to the individual maximum incentive amount for each player. Last year, 15.38 percent of the available incentives to this specific group was triggered, so for a player like Branyan, on a guaranteed major-league deal, the effective salary would be equal to the formula: Eff$ = Base + (0.1538*Max Incentive Amount) or, more specifically, $1.4 million + (0.1538*$0.35 mil) = $1,453,830. Even though Branyan triggered all of his incentives, the expected rate suggests he would have received $53,830 as opposed to $350,000, with the former shifted back to the team.
Another issue involves players not at all like Branyan in this group: players signed to minor-league deals without guaranteed major-league contracts, and with a much higher rate at the big-league level with incentives. Players of this type won't receive their full major-league salary if they are rostered but a portion of the time, so their big-league contracts must first be prorated. Rodrigo Lopez signed one of these deals with the Phillies worth $650,000 at the major-league level, with possible incentives totaling $3.05 million. The former Orioles hurler only played for the Phillies for around 41 days, and while the team did not know with any certainty how long his stint would last, prorating his salary makes much more sense than simply plopping it into the study at the full base, as I did in the original article. Taking his actual days spent rostered and dividing by 185, we get 22 percent to be applied to the base salary, netting $143,000, before running through the aforementioned incentive calculations. All told, Lopez's effective salary for this study totaled $613,144, as he was expected to earn much more in the way of incentives than he actually did.
The final issue to tackle involves effective salary and marginal salary, the latter of which describes money used in dollar per win calculations paid out above what would be paid to a player making the league minimum. If not Lopez, the Phillies could have used any of a number of freely available players at the replacement level, to be paid $400,000, so the true way to determine the rate of dollars per win would be to prorate the league minimum based on the actual days rostered and subtract that from the effective salary. Using Lopez as an example once again, a replacement level player at the league minimum would have made $400,000 * (41/185), or $88,640 in his place, a figure subtracted from the previously stated $613,144 to bring his marginal 2009 salary for these purposes to $524,495.
Onto the results, the group produced 21.1 WARP3 in the aggregate, which broke down to 16.1 wins across 40 hitters, and just five wins from the 28 pitchers; the average low-risk hitter signing produced 0.4 wins, while the average pitcher halved that mark, coming in at 0.2 wins. We're clearly not talking about world-beaters here; how could we be with group members like Kip Wells, Mike Hampton, and Russ Ortiz? The next question focuses on how much such players were paid for their meager contributions. The group totaled $50,133,085 in marginal dollars, equating to $2.38 million/win via WARP3, which broke down to $1.94 million per win for hitters, and $3.78 million win for pitchers. So, hitters were more productive and more of a bargain than pitchers, but the pitcher's rate wasn't necessarily poor given that teams did not value them any more than the perceived can't-miss free agents.
It might seem counterintuitive to suggest that the market is still efficient if the better-bet free agents sign for $4-6 million and these low-risk pitchers are not far behind. What would make this assertion false is if their rate exceeded that of the others, but that didn't occur last year. Thus far, we've been using WARP3 as the win-based metric. What if we use rWAR (developed by Sean Smith), which uses a much higher replacement level? Keeping in mind previous research that free agents are signing for $6-7 million per win measured by rWAR, the $6.35 million/win rate using Smith's stat casts low-risk signings in a less friendly light: now they border on inefficient. The free agents falling out of the low-risk category tend to sport a higher expected level of performance, meaning the same amount of money could be better spent on players with a much more guaranteed range of possible productivity.
Productivity was something this group largely lacked: just eight of the 68 members exceeded the 1.5 WARP3 threshold, with four besting 2.5 wins: Fernando Tatis, Branyan, Juan Uribe, and Craig Counsell. (None of them pitch.) The low rate of solid performers here suggests, as I did in the original article, that signing players to low-risk deals is a great strategy as long as a team is not built around such players; bring in Glendon Rusch on a flier, not with the expectation of 30 starts and 180 innings. While certain diamonds in the rough can be found, as with Branyan and Uribe (or Watchmen and Catch A Fire in my DVD expedition), the general rule I came up with the first time around still applies: for every Chris Carpenter coming back from injury, there are five Bruce Chens.
Low-risk signings constitute a solid strategy given that teams rarely have to guarantee anything, and they can simply shuffle through various pieces until the one that fits emerges, but when the rate of dollars per win inches closer to that of the wider group, the risks of these moves can actually be more damaging than merely bringing up replacement-level players for the minimum. Suggesting that a low-risk move could be risky may sound like a play on words, but teams need to become more selective with their bargain-bin hunting.