By Breedlove
Editor’s note – This article originally appeared on AstrosConnection.com on February 15, 2001.
The Astros pitchers and catchers report to spring training today and all I got was this lousy T-shirt. Apologies, folks, but we remain in the throes of Quiet Time. The big move since last we met was exercising the option on Brad Ausmus. While the sense of direction indicated by that move is pleasing, this space goes to a treatment on the dreaded OPS.
What The…?
Exactly what is OPS? The short answer is On-base Plus Slugging, the sum of OBP (on-base percentage) and SLG (slugging percentage.) You just add the two numbers together and the total is OPS. Any baseball statistics source you find is likely to have OBP and SLG readily available; OPS may even be added together already. But just in case their makeups are useful to you, and they will be for this discussion, here are quick explanations of where OBP and SLG come from.
OBP is a player’s on-base percentage, a numerical reflection of how often he has gotten on base as a rate. The formula for OBP is (H+BB+HBP)/(AB+BB+HBP+SF). All of those letters translate into this: Add the number of hits (H) to the number of walks (BB) to the number of times hit by pitch (HBP). That gives the total number of times a player reached base successfully. Then add the number of at-bats (AB) to the number of walks (BB) to the number of times hit by pitch (HBP) to the number of sacrifices (SF). That gives you the total number of opportunities a player had to reach base successfully. Now divide the total of successful chances from the first part into the total of opportunities from the second part to get a ratio of success. More important than knowing the formula is understanding that OBP is simply the rate at which a player has reached base.
SLG is simpler than OBP to figure, but it’s a little more difficult concept. SLG is a player’s slugging percentage, a numerical reflection of how many bases he averages per at-bat. The formula for this one is (TB/AB), or total bases divided by at-bats. Both those numbers show up on stat lines, but here’s an explanation. Total bases is the cumulative number of bases for all the hits a player has gotten, (1B*1)+(2B*2)+(3B*3)+(HR*4). So figure out singles times one, doubles times two, triples times three, and home runs times four, and then add them all together; then divide that by the number of at-bats. Like OBP and batting average, SLG is represented to the third decimal place. And similar to OBP, it’s less important to know exactly how to figure SLG than it is to understand that it’s just the average number of bases a player gets per at-bat.
Caution: Deconstruction Ahead
Suppose we figure out the average number of prison terms served by criminals and call it their on-base percentage. Then we look at the average number of years served in each prison term and call it slugging percentage. A cold-blooded killer will have a huge slugging percentage, but if he serves a prison term it will probably be his last, so his OBP will be low. The local crack-fiend will always be on base, so to speak, but he’ll be bailed out and back on the corner tomorrow. These two are nowhere near the same type of criminal, yet they would have a similar OPS. The crack-fiend’s might even be higher — then I’d mistakenly call him when I need someone iced.
Ridiculous analogy? Maybe not. Problem one with OPS is that it offers absolutely no insight about a player’s ability in any one area. Baseball teams have specific needs, yet OPS moves away from the specific to the general. If a team needs a leadoff man, should it try to pull cold-blooded killer-esque Cecil Fielder from retirement? If it’s short on power should it chase the services of metaphorical crack-fiend Rickey Henderson? OPS is an attempt to stratify players by total offensive output, but tells surprisingly little about their abilities. Much more can be learned just by looking at the elements of OPS, OBP and SLG, separately. That’s not the worst of OPS, though.
No “D” In OPS
The obvious flaw in comparing players by OPS is that it omits defense completely. Most stat-geeks don’t mind that a bit since defense doesn’t add up to much in rotisserie and fantasy leagues, but in the real world it is a huge part of the game. Many people have considered various methods of measuring the real value of a player’s defense, but no one has come up with anything satisfying yet. We have ways of measuring defense in a vacuum, but no way of measuring how well it accomplishes its goal, preventing runs from scoring. We credit all of that to the pitcher and just move on, trusting our eyes to tell us who is getting it done out there.
Theoretically we should eventually be able to remedy this. What we need are models based on game results that tell us what happens, on average, when a player boots a ball or makes a diving stab–how many runs are prevented by a great play, how many are allowed by an error. This still requires an arbitrarily defined standard, though. Someone has to decide what constitutes an exceptionally good or bad play. Still, the idea behind it is solid.
Another approach to valuing defense is to check all the innings played and see how many runs scored when there was an error versus when the play was clean. That would give you some idea of just how much damage an error causes on the scoreboard. Unfortunately, that suffers the same flaws as the error measure itself. The scorer arbitrarily assigns errors, and the greater a player’s range, the more chances for errors he will get. But again, the principle is sound.
With any luck you get the idea. Even though we can compare players defensively, there is no accounting for the real impact defense has on the score like there is for offense. Only defensive extremes get noticed, while on offense fine comparisons are attempted using metrics like OPS. Defense is largely ignored unless you’re Brooks Robinson or Dmitri Young. This is akin to having only Tim Bogar and Babe Ruth as examples of hitters.
Donut Truck?
Just how fine a metric is OPS anyway? What does it omit besides the 27-plus outs a player spends in the field? It must be a lot, because the list of the 100 all-time OPS leaders includes Ryan Klesko and Ellis Burks. If we break down OPS to its simplest level, we find that it is an attempt to describe, as a rate, how many bases a player accumulated. OBP tells you how often a player got at least one base and SLG tells you how many bases a player got per at-bat. It’s a fine idea in theory, since bases are what lead to runs. Yet if bases are the metric, isn’t something missing here? How do we account for the extra bases accumulated by a player but not caused by contact with a bat? To list a few:
– Going from first to third on a single instead of first to second
– Going from second to home on a single instead of second to third
– Scoring from first on a double
– Scoring from third on a sac-fly
– Taking second or third on a sacrifice of any kind
– Avoiding the double play by beating the throw
– Avoiding the double play by forcing a fielder’s choice
– Stealing bases
If you really want to count the bases that don’t show up in OPS, preventing bases from being gained by the opposition while on defense should be included here. If a player’s defense can stop the opposition from taking twenty bases over the course of a season that another player would have allowed, that must be close to as valuable as having contributed those twenty bases on offense. If he can prevent a hit another player would have allowed it must be near in value to getting a hit of his own. I’ll approach it from an outfield perspective:
– Preventing a liner to the gap from reaching the wall
– Hitting the cutoff man
– Throwing ahead of the runner and preventing an advance
– Throwing a runner out
– Recording an out instead of letting a ball drop (probably one-third of these are singles in front of the outfielder, two-thirds go beyond the outfielder’s range for extra-bases)
The big point here is the game is teeming with plays that become lurking variables. They never show up in stat lines, but their effect on scoring is tangible. Because we can rarely say with any certainty what would have happened had a play been made differently, we can’t really pin runs on fielders unless an overt error is made, and we tend to credit a player’s runs scored to the hitters behind him. Yet it’s fair to say that how often these plays do become runs–I’ll call them Secondary Runs–is highly dependent on the abilities of the players involved after the ball is put in play, whether on offense or defense. These abilities are not measured by OPS, yet every time Mike Cameron leaps over the centerfield wall in Safeco and pulls a homerun back in the yard it’s as good as if he hit one himself.
A speedy or heady player clearly has significant advantage in this area. If he can score from first on a two-out double, he has some advantage over the guy who can only go from first to third on the same hit. If he can stop a hit in the gap from reaching the wall with a man on base, at least some percentage of the time it will save a run. The fact that measuring these occurrences can only be done subjectively does not make Secondary Runs any less real. They don’t show up in OPS at all, but I’m convinced that as much as 100 points of OPS in extreme cases can be made up via the opportunities laid out above. So what makes 100 points of OPS?
First off, the 100 points is probably distributed something like this: .035 OBP/.065 SLG. Some portion of both OBP and SLG occurs simultaneously; when one player gets a hit more than another player, he has relatively improved both his OBP and his SLG at the same time, even though it is just one event. Over 500 at-bats, the OBP and SLG differences above translate to 17 to 20 or so extra hits and walks and add about 33 total bases. Those numbers are not insignificant, but with good baserunning and defense the gap can be made up. It’s not difficult to imagine a player gaining enough bases on offense and saving enough on defense to make up that 33-base gap. True, the player with a lower slugging percentage will not drive in runs at as great a rate, and he also requires a greater degree of success from the hitters behind him to score as often, but perhaps he can make that up, as we’ll see next.
Home Sweet Home
Runs scored are all that counts in baseball, and though that number is broadly controlled by how often a player gets on base, it is not determined by it. Good baserunning will allow some players to score more often after they reach base, as a rate. A player with a greater slugging percentage will start from a better average base, but not by much. It is significant that a homerun requires no success whatsoever from the hitters who follow to score at least one run. You can figure a player’s average start base by adding total bases to walks and dividing by hits plus walks, but for the following exercise we will assume equivalent average start base.
Suppose Player A reaches base 200 times and Player B reaches base 220 times, both over 500 at-bats. We’ll pretend neither takes a walk so their OBP’s are .400 and .440 respectively, a monstrous gap of 40 points. If both players score 40% of the time they reach base, Player B will outscore Player A by 8 runs.
However, if Player A can implement some of the methods for gaining bases above and outscore Player B as a rate, he will quickly equal or surpass B even though he is getting on base less often. Say Player A is faster and makes better decisions; he actually scores 43% of the time he reaches base while Player B only scores 38% of the time. Player A will score more often than B despite reaching base 20 fewer times. OPS has no means of accounting for this dynamic.
OPS is a decent approximation that allows simplistic stratification of offensive ability. It’s not useless, but there are too many factors hidden from and within OPS for it to work by its lonesome. Bill James, guru of the statistical world, has yet to convert to OPSism after thorough analysis, and even the biggest supporters of OPS acknowledge that its elements, OBP and SLG, are not equivalent and are not the same type of calculation. They should not be lumped together simply because they take the same final form, .XYZ. But they do it anyway, for the sake of convenience and because they think it’s still better than other measures.
Maybe OPS is better than the other common offensive statistical metrics — if you are too lazy to be bothered to understand a player’s real value. OPS is more flawed than other offensive metrics because it doesn’t measure any single ability, and differences in the elements of OPS are masked when they are lumped together. Perhaps a speed bump like a suggestion of how a player with an OPS 100 points lower could be equally effective with good defense and baserunning will slow the OPS bandwagon down a little. It’s doubtful though, since they keep running it out anyway knowing Rickey Henderson and Cecil Fielder have an identical career OPS of .827.
Apples And Apples
OPS supporters probably tire of hearing the same examples though. While crack-fiend vs. cold-blooded killer is possibly new, Rickey Henderson vs. Cecil Fielder has made the rounds. Their OPS numbers are comprised very differently — Henderson has a huge OBP and Fielder was a huge slugger. Let’s look instead at two guys who share a birthday. They have played the same position for almost the same amount of time, their OPS elements are similarly constructed, and both are headed to the Hall of Fame if they can play five more seasons or so at their current paces. Yet any Astros fan can tell you with zero calculation that Jeff Bagwell is a better ballplayer than Frank Thomas just by watching the games.
How can that be? Frank Thomas has a career OPS of 1.019 and Jeff Bagwell’s is just .969, a difference of 50 points. And that’s over a ten-year sample, so it must be meaningful. In answering that question we’ll ignore the lineups around them — for every Craig Biggio, Moises Alou, and Ken Caminiti there’s an Albert Belle, Magglio Ordonez, or Robin Ventura, and an extra hitter in an AL lineup. We’ll also ignore the quality of the teams–both have played on plenty of good and bad ones. We’ll even ignore how much the spacious Astrodome held Bagwell down — maybe Enron vs. New Comiskey will make up for it when it’s all said and done.
Since 1991, both Bagwell and Thomas have been full-time starters. Though he rarely misses single games, Bagwell has had some seasons cut short by broken bones in his hand. Thomas is prone to pulled muscles and sore groins–Big Hurt indeed–and they have played in a remarkably similar number of games over this stretch. The numbers are 1470 for Thomas and 1476 for Bagwell.
Through this time period, just counting walks, hits, and HBP, Frank Thomas has reached safely 2880 times while Bagwell has gotten on 2719 times. Jeff Bagwell has converted that into 1073 runs, Thomas to 1044. That bears repeating: despite reaching base 161 fewer times, Bagwell has outscored Thomas by 29 runs; that’s with Thomas getting a built-in 31-run advantage through having hit more homeruns that scored himself.
Thomas has the advantage in RBI, 1152 to 1093, and we already know where 31 of that comes from. So after ten years and literally thousands of at-bats, the difference in RBI plus runs scored, their total runs produced, is thirty. That number is strikingly similar to the difference in homeruns, which many people like to subtract once from runs produced so they don’t get double counted. Thirty runs in ten years sounds like three a season to me. Fifty points of OPS is three runs a season?
Some of the similarity in total runs produced comes from Bagwell having stolen 167 bases while Thomas has stolen 29. Some comes from Bagwell being a heady baserunner who aggressively moves on contact, taking extra bases and forcing throwing errors. Some comes from Bagwell being able to stay out of the double play, or being able to take and score from third on a sac-fly.
While their extra-base hits already figure into OPS, it’s worth noting their doubles are almost identical, 351 for Bagwell to 350 for Thomas since ’91, and Bagwell has 22 triples to Thomas’s 7. The big picture here is that Bagwell and Thomas have made nearly identical total offensive contributions despite a 50-point gap in OPS.
Now let’s look at defense, which is where Bagwell really shines over Thomas. Bagwell has a career fielding percentage of .993, identical to the league fielding percentage. Thomas has a career fielding percentage of .991. Bagwell has a Range Factor of 9.24, way over the league average, Thomas 8.66. Range Factor is putouts and assists every nine innings, and since some games go extra innings, that is pretty close to one more putout or assist every single game.
Additionally, Bagwell’s greater Range Factor means he’s had chances on tough plays that Thomas could never dream of getting to, yet his fielding percentage is still better. And anyone who has watched Bagwell play knows he starts more double plays from first than everyone on the planet. None of this defensive edge is any secret, as the White Sox have made Frank Thomas mostly a DH these days.
Is there any chance Bagwell’s defense has prevented enough runs over the last ten years to make up any gap he might suffer in comparison with Thomas on offense? You bet. Easily. There should be no doubt in anyone’s mind that Bagwell’s superior fielding has prevented more than three runs a season. Yet Thomas has the gaudy OPS, and he will probably, mistakenly, be remembered as the better player. OPS is simply unreliable, and using it to compare players without a complete treatment is a joke.