Shortcomings of OPS as an Advanced Metric

Regular contributor Patriot makes his SDR debut with a column about OPS.

To many in the general population of baseball fans, the singular metric associated with sabermetrics might be On Base Plus Slugging (OPS).  OPS has gained a measure of acceptance in the mainstream as an offensive metric in both its raw form and the park- and league-adjusted variant OPS+, and given the long history of their use in sabermetrics (both having been developed by pioneering sabermetrician Pete Palmer), it is no surprise that the metrics are associated with the field and are still in common use.  However, OPS has shortcomings that can be problematic for serious application:

• OPS is not expressed in meaningful units: Ideally, a metric should be expressed in units that are fundamental to the game itself (such as runs or wins) or that can be easily explained.  On their own, the components of OPS do just fine by this standard.  On Base Average can easily be understood as the proportion of plate appearances in which the batter reaches safely, and Slugging Average is the average number of total bases per at bat (although the use of the name “Slugging Percentage” is misleading at best).  But when they are combined into OPS, it becomes impossible to articulate what the unit of measurement is or what the result is meant to represent.  While a batter with a .400 OBA reaches safely 40% of the time and a batter with a .500 SLG averages one total base every two at bats, the meaning of his corresponding .900 OPS cannot be similarly stated.  The best one can do is to state what its user intends it to represent–a measure of overall hitting productivity.
• OPS is not as accurate as competing overall measures of offensive productivity: Generally, OPS does a decent job of predicting runs scored on the team level, but it  tends to be slightly less accurate than more refined metrics.  OPS still performs credibly, but metrics based on linear weights or Base Runs perform better.

OPS could be improved by weighting OBA more heavily; studies have suggested a multiplier in the neighborhood of 1.8 to maximize correlation with team runs scored (i.e., OBA*1.8 + SLG).  Doing so, though, would take away one of the strongest selling points of the metric, which is ease of calculation.

• OPS does not have a 1:1 relationship with team runs scored: Scoring runs is the objective for the offense, so it follows that measures of offensive productivity should be relatable to team runs scored.  However, the difference or ratio between two OPS figures do not directly correspond to the same between team runs scored.  A team with an OPS 5% higher than league average will tend to score about 10% more runs than the league average–not 5%.  That is, OPS has approximately a 2:1 relationship with runs scored.

The lack of a 1:1 relationship with runs scored does not mean that OPS has no validity as a metric, but it does make it more difficult to interpret what various OPS values mean in terms of measuring offensive output.

• OPS is a statistic of convenience: Ideally, a metric should be constructed with a clear idea of both what it intends to measure and how it will be constructed in order to do so.  OPS does just fine on the first count, but leaves much to be desired from the construction standpoint.  Consider for a moment what the formula looks like if, instead of writing it simply as OBA + SLG, you write it as:

(H + BB + HBP)/(AB + BB + HBP + SF) + TB/AB

If presented with this formula and you were previously unfamiliar with OBA and SLG as separate measures, you might ask, “Why are walks and hit batters divided over one denominator (plate appearances) while total bases are divided by another (at bats)?”  The use of different denominators requires more involved analysis to understand the relative weights that OPS places on each offensive event.

OPS would have never been created without OBA and SLG already existing.  OPS is easy to calculate, but only if one presumes that OBA and SLG have already been calculated; if not, OPS is not much quicker to calculate than alternative metrics.

• OPS+ is an improvement on OPS, but still has shortcomings: OPS+ is a league- and park-adjusted version of OPS, but many people who use the statistic are not aware of how it is calculated.  OPS+ is not simply OPS/league OPS—it is actually calculated as:

100*(OBA/league OBA + SLG/league SLG – 1)

OPS+ is actually the sum of relative OBA and relative SLG rather than a relative version of the combined total.  Some have criticized the name of the statistic as misleading because other “plus” statistics (e.g., ERA+) are a simple quotient of the player’s figure and the league average.  However, there are two major benefits reaped by the way OPS+ is figured.  One is that it implicitly weighs OBA more heavily than simple OPS, as the league SLG is typically 20-30% higher than league OBA.  Additionally, OPS+ has a 1:1 relationship with runs scored, so a team with an OPS+ of 110 should score about 10% more runs than average.

While these are nice properties, many of the other shortcomings of OPS are not rectified.  The higher weight on OBA still leaves room for improvement in correlation with team runs scored, and the specific variant of OPS+ most widely disseminated utilizes a complicated and questionable park adjustment methodology.

The message here is not that OPS should never be used or that it is not an improvement on starting with a batter’s triple crown stats. But when serious analysis or a nuanced evaluation is in order, OPS leaves much to be desired.