Defensive Metrics Explained: Outs Above Average, Defensive Runs Saved and UZR
How Outs Above Average, Defensive Runs Saved and Ultimate Zone Rating each measure fielding, why they disagree about the same player, and how much data you need before trusting any of them.
Three statistics dominate the modern discussion of defense: Outs Above Average, Defensive Runs Saved and Ultimate Zone Rating. They are all attempts to answer the same question, which is how many outs or runs a fielder saved compared with an average player at his position. They often give different answers about the same player in the same season. That is not a sign that two of them are broken. It is a sign that fielding is much harder to measure than hitting, and that each metric makes different choices about how to measure it.
This article explains what each metric counts, what units it uses, how the three differ in method and data, and why a single season of any of them is a weak guide to a fielder's real ability. It is written for a fan who wants to read a defensive leaderboard without being fooled by it. The numbers in the examples are illustrations built to show the arithmetic, and they are labeled as such. The article states no player's ratings, because the historical record used elsewhere on this site does not carry any of these metrics.
Why fielding is harder to measure than hitting
A batter's job produces an event that can be counted: a hit, a walk, a home run, an out. Almost everything he does shows up in the scorebook. A fielder's job is different, because the most important thing he does is often invisible. A ball that a good fielder reaches and an average fielder does not is counted as a hit for the batter and nothing at all for the fielder. The play that did not happen leaves no mark.
The oldest fielding statistic, fielding percentage, ignores that problem entirely. It counts only the chances a fielder actually handled, and it punishes errors, so a player with poor range who never gets near a ball can post a better percentage than a player who reaches everything and bobbles a few. The problem is covered in detail in Fielding Percentage, Range Factor and Why Older Fielding Stats Mislead. The three modern metrics exist to fix it.
All three solve the problem in the same general way. They look at every ball put in play, estimate how hard it was to field, and then credit or debit the fielder depending on whether he made the play. A fielder who turns a difficult ball into an out gets a lot of credit. A fielder who misses an easy ball takes a heavy penalty. Add up the credits and debits across a season and compare the total with what an average fielder at the position would have earned.
The differences lie in how each metric decides what counts as difficult, what data it uses, and what unit it reports in.
Outs Above Average
Outs Above Average, usually shortened to OAA, is a Statcast metric. Statcast is the tracking system used in major-league ballparks, and its cameras and radar record where every batted ball goes, how hard and at what angle it was hit, and where every fielder is standing. OAA arrived in the late 2010s for outfielders and was extended to infielders afterward, so the metric has a shorter history than the two older ones.
The method starts with a catch probability. For each batted ball, the system asks how often a fielder in this situation makes the play. For an outfielder the key inputs are how far he had to go, how much time he had and in what direction. For an infielder the inputs also include how fast the runner is and how long the throw is. A routine fly ball might have a catch probability of 95 percent. A diving catch in the gap might have a probability of 20 percent.
Credit follows from the probability. If the fielder makes the play, he earns the part of the play he was not expected to make, which is one minus the probability. If he misses, he loses the probability itself. A catch with a 20 percent probability is worth 0.8 of an out. A miss on a 95 percent ball costs 0.95 of an out. Add up every play in a season and you have the fielder's outs above average.
Three features of OAA are worth keeping in mind. First, the unit is outs, not runs, so an OAA total does not tell you directly what the plays were worth to the team. Second, the metric covers the fielder's range and ability to make plays on batted balls, and the system treats throwing and the arm separately. Third, it is measured against other players at the same position, so the average fielder at a position is zero and the leaders are positive.
Catchers are not measured by OAA, because the plays a catcher makes are of a different kind. Catcher defense is a topic of its own; Catcher Framing Explained covers the largest piece of it.
Defensive Runs Saved
Defensive Runs Saved, or DRS, comes from Baseball Info Solutions, a company whose analysts watch video of every play and record the type of ball, where it was hit, how hard it was hit and who fielded it. It dates to the early 2000s, which makes it one of the two older metrics alongside UZR.
DRS converts everything into runs. A fielder who is plus 10 is credited with having saved ten runs relative to an average player at the position, and one who is minus 10 cost his team ten. A common rule of thumb is that about ten runs equal one win over a full season, so a plus-10 fielder was worth roughly a win, though the rule of thumb is approximate and depends on the run environment.
The metric is built from components, and each one handles a different part of the job. The range and play-making component, the largest one for most positions, compares the fielder with average on balls hit into his area. Infielders get a component for turning double plays. Outfielders get a component for their arms, based on how many runners they hold or throw out. Catchers have components for controlling the running game and, in newer versions of the metric, framing. First basemen have a component for scooping throws. There is also a component for bunts. The details of the components have changed over time, so check the provider's current documentation for any season.
Because it is expressed in runs, DRS is easy to add to batting and baserunning value, and that is the reason it often appears as a building block in wins-above-replacement calculations. Its main weakness is that it depends on human judgment about where a ball was hit and how hard. The scouts who code the data are trained and their work is reviewed, but human coding is not as precise as an optical tracking system.
Ultimate Zone Rating
Ultimate Zone Rating, or UZR, was developed by Mitchel Lichtman and is published at FanGraphs. It is built on the same kind of video-coded data as DRS, which comes from Baseball Info Solutions, and it also reports in runs. Its defining idea is the zone: the field is divided into regions, and each region has an average rate of outs for balls hit there at a given speed and type. A fielder is rated on how often he converts balls hit into his zone compared with the average.
UZR is also divided into components, usually an outfield arm rating, a double-play rating for infielders, an error rating and a range rating. The range rating is the heart of it. The total is often shown as UZR per 150 games, which scales a partial season up to a full one so players with different playing time can be compared.
UZR adjusts for factors that can distort a raw count. It accounts for the handedness of the batter, the park and the base-out state, since a fielder's job changes depending on whether a runner is on. The adjustments are one reason the numbers differ from DRS even though both start from similar data. UZR is published for seasons going back to the early 2000s.
Before the video-coded era, there are other historical estimates, such as Total Zone, that use play-by-play data and positioning rather than video. Those are less precise and are mostly used to put pre-2000s fielders on the same scale as modern ones.
The three metrics side by side
The table summarizes the differences. Details such as start dates and component lists are approximate, and the providers' own pages are the authority.
| Feature | Outs Above Average | Defensive Runs Saved | Ultimate Zone Rating |
|---|---|---|---|
| Source of data | Statcast tracking | Baseball Info Solutions video review | Baseball Info Solutions video review |
| Unit | Outs | Runs | Runs |
| Core idea | Catch probability for each play | Plays made compared with average at the position | Zones of the field and the average out rate in each |
| Covers catchers | No | Yes, with separate components | No |
| Where it is published | Baseball Savant | Provider and several stat sites | FanGraphs |
| Length of record | Shortest | Early 2000s onward | Early 2000s onward |
If you remember one point from the table, remember that OAA and the other two are not measuring the same raw material. OAA uses sensors. DRS and UZR use human coding. That is why OAA is more precise about the distance a fielder had to cover and the time he had, and why DRS and UZR are more flexible about what they can include.
Why the metrics disagree
It is common for a fielder to rank near the top in one metric and in the middle of the pack in another. There are several ordinary reasons.
The first is the data. A sensor and a human scout can disagree about where a ball was hit and how hard. A small difference in the starting point of a ball can change its estimated difficulty.
The second is the model. Each metric uses a different method for deciding what counts as average. Zone-based metrics group balls into regions and treat all balls in a region the same. Probability-based metrics estimate each play individually. A ball that is in the same zone as an easy play but is more difficult will be handled differently in the two systems.
The third is the scope. OAA, as described above, focuses on range and plays, and treats the arm separately. DRS and UZR include the arm within the same total. A fielder with an excellent arm and ordinary range will look better in a metric that includes the arm than in one that does not.
The fourth is positioning. If a manager tells a fielder where to stand, the metrics have to decide whether the credit for a good position belongs to the fielder or the coach. They do not all handle it the same way, and the spread of extreme shifts, which the major leagues limited beginning in 2023, made the issue larger. The metrics generally cannot see the instruction behind a fielder's starting spot.
A hypothetical disagreement
Here is an invented example, not a real player, to show how a disagreement arises. Suppose a shortstop fields 20 slow rollers into the hole and makes the play every time. A zone-based metric sees that those balls landed in a zone where the average out rate is 60 percent, so each play earns 0.4 of an out, or 8 outs in all. A tracking-based metric sees that each ball was soft, far from the fielder and required a long throw against a fast runner, so it puts the chance of an out at 30 percent and credits 0.7 of an out per play, or 14 in all. The two metrics have scored the same 20 plays and reached different totals, and neither is wrong. They disagree about how hard a particular kind of ball is, and that is the most common source of gaps between them.
How much data is enough
The most important caution about defensive metrics concerns sample size. A hitter faces hundreds of plate appearances in a season, and the results stabilize reasonably well. A fielder faces far fewer difficult chances, and the credit or debit he earns depends heavily on a small number of plays.
Consider an illustration. Suppose a fielder has 300 chances in a season that are true coin flips, each with a 50 percent chance of being converted. By chance alone, the number of plays he makes will vary by roughly nine from year to year, because the standard deviation of 300 coin flips is about 8.7. If you assume an out is worth roughly three-quarters of a run, a figure often used as a rough guide, that is a swing of about six or seven runs from luck alone. Yet the gap between a very good fielder and an average one is often described as ten or fifteen runs a year. The noise in one season is a large share of the signal.
That is why analysts commonly recommend looking at multiple seasons, often three, before drawing a conclusion about a fielder's true ability. Metrics that sum to a single season's ranking are good descriptions of what happened. They are weaker predictions of what will happen. A player who posted a very high total one year can be expected to regress toward average the next, in the same way that a hitter's hot season is partly luck.
The same logic applies to players at the extremes within a season. A rookie's total after 40 games is only a sketch. Treat a partial-season number as a hint, and wait until the sample grows before using it to settle an argument. Expected Statistics: xBA, xSLG and xwOBA Explained makes a related point about hitting: a result is the combination of skill and luck, and the ratio depends on the sample.
What none of the metrics capture well
Even a large sample does not fix every problem. Some limits are built into the methods.
Positional comparisons are one. A plus-10 first baseman and a plus-10 shortstop are not equally valuable, since the shortstop's position is more demanding and the average hitter at the position is lower. That is why wins-above-replacement calculations add a positional adjustment on top of the raw defensive number. The raw metric tells you how a fielder compares with others at his own position and nothing more.
Another limit is what the models cannot see. A fielder who calls off a teammate, who positions a runner by looking him back, or who makes a clean transfer that prevents an infield single is doing work that is hard to code. The metrics also treat the pitcher's contribution as part of the background. A pitcher who generates weak contact makes the fielders' lives easier, and a pitcher who gives up hard line drives makes them harder. The batted-ball data accounts for the quality of contact in a way that helps with this, but not perfectly.
A third limit is injuries and aging. A fielder who loses a step often declines before his metrics show it clearly, because small changes in range accumulate over many plays. Sprinting and speed are covered in Sprint Speed and Baserunning Metrics, which is relevant because the same athletic ability that makes a good runner makes a good outfielder.
How to read a defensive leaderboard
A few practical habits help.
- Use more than one season. Three years is a common standard, and more is better for a player whose role changed.
- Look for agreement. If two metrics rate a fielder as above average, the conclusion is stronger than if only one does.
- Look at the components when they are available. A high total built on the arm tells a different story from one built on range.
- Compare within a position. Do not use a raw defensive number to compare a catcher and a left fielder.
- Pair the number with what you can see. The metrics are best used as a check on your eyes, and your eyes as a check on the metrics.
- Treat awards with the same caution. Awards for defense reflect reputation as well as measurement, and a player can win one without leading any metric.
Another way to apply this is to think of a defensive total as a range rather than a point. A fielder rated plus 8 in one metric might be plausibly anywhere from plus 2 to plus 14 once you account for noise. That is not a failure of the metric. It is the honest limit of the method.
Common questions
Which metric is the best? None is best in every respect. OAA has the most precise tracking data but a short history and a narrower scope. DRS and UZR have longer records and include the arm, but rely on human coding. Using two or three together is better than relying on one.
What does it mean when a fielder is plus 10? In a run-based metric such as DRS or UZR, it means he saved about ten runs more than an average fielder at his position over the period measured. In OAA, plus 10 would mean ten outs.
Can a good fielder have a negative rating? Yes, in a single season, because the noise is large. A negative total over several seasons is stronger evidence.
Do these metrics account for the shift? To some extent, but not fully, and the effect of the alignment is hard to separate from the fielder's own skill.
Why are catchers treated differently? Their main contributions, such as framing, blocking and throwing, require separate models. OAA does not cover them.
Where can I find the numbers? The providers publish them: Statcast for OAA, Baseball Info Solutions and various stat sites for DRS, and FanGraphs for UZR. This site's stat calculator handles traditional counting and rate statistics, and the leaders page covers the standard ones.
Where to go next
If you want the background on the older measures that these metrics replaced, start with Fielding Percentage, Range Factor and Why Older Fielding Stats Mislead. For another example of a metric built from tracking data, read Spin Rate, Spin Axis and Pitch Movement Explained, and for a broader view of analytics coverage, the stats and analytics topic page collects current articles.