Expected Statistics: xBA, xSLG and xwOBA Explained
Expected batting average, expected slugging and expected wOBA estimate what a hitter should have produced from the quality of his contact. Learn how they are built, how to read the gap and where they fall short.
A hitter lines a ball into the gap at 105 miles per hour and the left fielder dives and catches it. Another dribbles a ball off the end of the bat toward the shortstop's right, and it rolls under a glove for a single. The scorebook records an out in the first case and a hit in the second. Anyone who watched both plays knows the scorebook has told the truth about the result and misled about the swing. Expected statistics exist to fix that. They ignore what happened to a batted ball and ask what usually happens to balls like it.
The three most common versions are expected batting average (xBA), expected slugging percentage (xSLG) and expected weighted on-base average (xwOBA). They are built from tracking data on each ball put in play, mostly the speed it leaves the bat and the angle it leaves at. They are not predictions of the future and not a replacement for the real numbers. They are a second opinion on how much of a hitter's results came from his contact and how much came from luck, defense and park.
This article explains what each statistic estimates, walks through the logic with a worked example, shows how to read the gap between expected and actual numbers, and lists the things the expected numbers cannot see. It is background for the companion articles on exit velocity, launch angle and barrels and on sample size and regression to the mean.
Where the data comes from
Expected statistics depend on Statcast, the tracking system installed in every major-league park. Radar and camera systems record the speed of each batted ball when it leaves the bat (exit velocity), the vertical angle at which it leaves (launch angle) and, with some versions of the model, other details such as the runner's sprint speed. This is information that did not exist in the box scores of earlier eras. The historical record, which this site uses for historical figures, has none of it, so expected statistics cannot be computed for a player from before the tracking era. A fan comparing Willie Mays with a current star cannot use xBA.
The tracking systems have changed over the years. Statcast began with radar in the major-league parks and later moved to a camera-based system. The newer system is more precise, but the idea is the same: measure the ball, not just the result.
What the data allows is a simple but powerful comparison. Take every batted ball in a season, or in several seasons, group the balls by exit velocity and launch angle, and count how often balls in each group became hits, and how many bases they produced. A ball hit hard at a low angle and a ball hit softly in the air are very different kinds of contact, and the groups separate them.
xBA: the probability of a hit
Expected batting average assigns each batted ball a probability of being a hit. The assignment comes from the history of similar balls. If, illustratively, 7 out of 10 balls hit at a given speed and angle have become hits, then each such ball is credited with 0.70 of a hit. Summing the probabilities across all of a player's at-bats and dividing by his at-bats gives his xBA.
A worked example shows how it goes. Suppose a hitter makes 500 at-bats, strikes out 100 times and puts the ball in play 400 times. Strikeouts count as at-bats with a zero probability of a hit, because no ball was put in play. If the quality of his 400 balls in play adds up to 140 expected hits, his xBA is 140 divided by 500, or .280. If he actually collected 120 hits, his batting average is .240. The difference of 40 points means that about 20 hits that his contact deserved did not fall in.
The example uses invented numbers to show the arithmetic. On the real leaderboards, gaps of that size are rare for a full season and common for a few weeks.
xSLG: the expected total bases
Expected slugging does the same for power. Where xBA gives each ball a probability of being a hit, xSLG gives each ball an expected number of bases. A ball with a good chance of being a home run, and a smaller chance of being a double or a single, would be credited with something like 2 or 3 expected bases. The total expected bases are then divided by at-bats.
Slugging percentage is total bases divided by at-bats, so a single counts as one base, a double as two, a triple as three and a home run as four. xSLG follows the same weights. The statistic is useful because power is more sensitive to quality of contact than average is. A hitter who consistently hits the ball hard in the air should have a high xSLG even in a stretch when many of his drives are caught.
xwOBA: the one that does the most work
Weighted on-base average, or wOBA, is a rate statistic that counts each way of reaching base in proportion to its run value. A walk is worth less than a single, a single less than a double, and a home run is worth the most. The formula, in outline, is the weighted sum of walks, hit by pitches, singles, doubles, triples and home runs, divided by at-bats plus walks (intentional walks excluded), sacrifice flies and hit by pitches. The weights are fixed each season to match the run values of that year and are scaled so that the league's wOBA is close to its on-base percentage. Readers who want to work with the underlying counts can use the stat calculator on this site.
xwOBA replaces the actual outcome of each batted ball with an expected value, in the same way as xBA and xSLG. The difference is that it keeps the actual result for the events that do not depend on contact quality: walks, hit by pitches and strikeouts. Those are credited as they happened. Only the balls in play are replaced by their expected values.
That is why xwOBA is often treated as the best single expected statistic. It combines a hitter's plate discipline with the quality of his contact in one number, on a scale that reads like on-base percentage. A player with an xwOBA well above the league average of about .320 is hitting the ball hard, walking, and avoiding strikeouts. The figure of .320 is the usual round number for a modern season; it moves a few points from year to year.
How to read the gap
The useful part of expected statistics is the difference between what happened and what the contact deserved. The table gives a rough guide.
| Gap | What it can mean | What to ask next |
|---|---|---|
| Actual well below expected | Bad luck, strong defense against him or a hard park | Is he hitting the ball to the same spots? |
| Actual close to expected | His results match his contact | Does the contact itself look sustainable? |
| Actual well above expected | Good luck, weak defense, a good park or real skill the model cannot see | Is he fast, or does he pull the ball to a short porch? |
The gap alone does not prove luck. It is a prompt for a question. A hitter who is fast will turn more grounders into hits than the average model expects, because the expected value for a given grounder largely assumes an average runner. A pull hitter in a park with a short fence turns more fly balls into home runs than a neutral park would allow. The expected number is a baseline, not a verdict.
Over a full season, the gaps tend to shrink. Over a week or a month, they tell very little, because the number of batted balls is small and each outcome carries a lot of weight. This is the same law that governs batting average on balls in play and every other noisy statistic, discussed in the sample-size article. A gap over 100 batted balls is much less informative than a gap over 400.
What expected statistics cannot see
No model is complete, and the limits of xBA, xSLG and xwOBA are worth knowing before leaning on them.
- Defense and positioning. The expected value for a ball in play is the league average for that ball. A club with better fielders turns more of those balls into outs. The statistic does not know which defense was behind the pitcher. The measurement of fielding itself is a separate field, covered in the article on defensive metrics.
- Spray direction and park. The basic model uses speed and angle, not direction, and tweaks to the models change over time. A fly ball down the line is treated the same as one to center, though they behave differently in different parks.
- Weather. Air temperature, wind and altitude change how far a ball carries. A ball that is a warning-track out in one city is a home run in another.
- Speed. Some versions give extra credit on balls that a fast runner can beat out, but the adjustment is limited. Readers who want the running side of the picture can read about sprint speed and baserunning metrics.
- Pitch type and quality of the pitcher. A hitter who hits the ball hard against fastballs but poorly against breaking balls looks the same in the average.
- Contact that never happens. Strikeouts and walks enter xwOBA as they happened, but xBA and xSLG count a strikeout as an at-bat with no value. A hitter who swings and misses often is penalized, as he should be.
One more limit is mathematical. The expected values come from past data, and they are re-estimated as the league changes. A change in the ball, in the pitching mix or in the way hitters swing can shift the relationship between exit velocity and results. Expected statistics are most reliable when compared within a single season.
Using expected statistics well
For a fan or a fantasy manager, the best use of expected statistics is as a filter for the question "is this real?" A hitter batting .320 with an xBA of .240 is probably headed for a decline; a hitter batting .230 with an xBA of .290 is a candidate for improvement. Neither conclusion is certain, but both are better guesses than the batting average alone.
A few habits help.
- Compare expected and actual numbers over at least a few hundred batted balls before drawing a conclusion.
- Look at all three. A high xBA with a low xSLG describes a singles hitter. A low xBA with a high xSLG describes someone who hits fewer, longer balls.
- Check whether the hitter is unusual in speed or park before assuming that a gap will close.
- Treat the expected number as a reason to look at the player more closely, not as a reason to stop.
For pitchers, the same method produces expected numbers against, such as expected wOBA allowed. The logic is the same: judge the pitcher on the quality of contact he allowed, not on where the ball happened to land. Some analysts also compute an expected earned run average from the same information.
A short history of the idea
Baseball statisticians have been trying to separate skill from luck for decades. Batting average on balls in play, line-drive rate and fielding-independent pitching were early attempts, each built from the box-score information available at the time. The common idea was that a hitter or pitcher controls part of what happens and not all of it. Expected statistics are the latest version of this idea and the most direct, because they measure the ball itself.
They also carry a lesson about all of baseball's statistics. Every number is a summary of a process, and the question to ask is how much of the process the number actually captured. Batting average captures the result. xBA captures the contact. Neither captures the pitcher's plan, the fielder's position or the wind.
Common questions
Is xBA a prediction? No. It describes how a hitter's contact would have played out against an average defense, not how he will hit next month. It is a better guide to the future than batting average because contact is more stable than results, but it is not a forecast.
Why is a hitter's xBA sometimes much higher than his average for a whole season? Common causes are a defense that played him well, a ball park that holds in fly balls and a slow running speed on ground balls. Persistent gaps in the same direction over several seasons can also reflect something the model does not see.
Can I use expected statistics for old players? No. The data comes from tracking systems that began in the 2010s, and it does not exist for earlier seasons.
What is a good xwOBA? A figure of about .320 is the league average in a typical modern season; a figure of .370 or more is usually star territory. These are rough benchmarks that move with the league's run environment.
Do expected statistics replace the traditional ones? No. The traditional statistics record what happened and decide games and awards. The expected ones help explain it.
Where to go next
To see how the quality of contact is measured in the first place, read about exit velocity, launch angle and barrels. For more on how analysts measure fielding, see defensive metrics explained, and for more articles of this kind browse the stats and analytics topic page. The leaders page shows traditional leaderboards for any season in the historical record.