Strength of schedule and how to adjust for it
A win over a weak opponent is worth less than a win over a strong one. This guide explains strength of schedule, how ratings adjust for it, and where the adjustments break down.
Two teams finish a season with the same record, 24 wins and 6 losses. One played in a league full of strong opponents, with road trips against the best programs in the country. The other played a handful of weak teams again and again and dodged the toughest ones. Are they equally good? Almost nobody would say yes, yet the standings treat them identically. That gap between the record and the reality is the reason strength of schedule exists as a concept.
Strength of schedule, often shortened to SOS, is a measure of how good a team's opponents were. Its purpose is to let you adjust other numbers, such as win percentage or point margin, so that teams who have faced different opposition can be compared fairly. In the NBA, where the schedule is nearly balanced, the adjustment is modest. In college basketball, where more than three hundred teams play inside conferences of very different quality, it can be the difference between a team that looks like a title contender and one that is merely beating up on weaker competition.
This article explains how strength of schedule is calculated, the logic of adjusting ratings for it, the small details that matter (home games, rest, timing), and the cases in which the adjustments can be misleading. It connects with margin of victory and what it says about a team's true level, because adjusting margin for schedule is the most common way schedule strength gets used in practice.
The basic idea: not all wins are equal
Start with an obvious thought experiment. If every team in a league played every other team the same number of times, home and away, strength of schedule would be irrelevant: each team would face exactly the same set of opponents (apart from itself, which is a small distortion that we can set aside). Nobody gets an easier road because everyone walks the same road.
Real leagues are never so tidy. In the NBA, each team plays 82 games, with 41 at home and 41 on the road, but the opponents are not evenly divided. A team plays each of the four other teams in its division four times, most of the other teams in its conference three or four times, and each team in the other conference twice. Two teams in different divisions therefore have slightly different mixes of opponents. The variation is small, but it exists, and in a season where one conference is clearly stronger than the other, being in the weaker one is a real, if modest, advantage in the standings. The WNBA and the international leagues each have their own scheduling structures, with the same underlying principle: the opponents are not identical, so the schedule is not identical.
College basketball is the extreme case. Men's Division I teams play roughly thirty games, a mix of non-conference games chosen by each program and conference games dictated by the league. Some conferences are stocked with teams that would be strong in any setting; others are not. A team that wins 14 of 16 conference games in a weaker league has accomplished something different from a team that wins 12 of 20 in a harder one. Strength of schedule is the mechanism that tries to quantify the difference, and the NCAA selection committee uses a version of it every March, as explained in quadrant records and the NCAA selection committee's NET.
Two meanings of schedule strength
Before getting to the math, it helps to separate two different ideas that tend to be conflated.
The first is past strength of schedule: how good were the opponents a team has already played? This is a historical fact. It tells you how impressive a record is.
The second is future strength of schedule: how good are the opponents a team has yet to play? This is a forecast. It tells you how hard it will be for a team to keep up its pace or to win the games it needs to finish in a certain position.
Both use the same raw material, the quality of opponents, but they answer different questions and should not be mixed. When someone says that a team has the toughest remaining schedule, they mean the future version, and the claim is only as good as the ratings of the opponents involved. When a selection committee talks about a team's resume, it is mostly about the past version.
How strength of schedule is calculated
The simplest approach is to take the average winning percentage of a team's opponents. If a team's opponents collectively won 56 percent of their games, its schedule was tougher than average; if they won 44 percent, it was easier. This is quick and easy to understand, and it was the standard for generations of sports statistics.
It has two weaknesses. First, an opponent's winning percentage includes the games it played against your team, which biases the figure slightly. The usual fix is to remove those games. Second, an opponent's winning percentage is itself a product of its own schedule. A team that went 25-5 against a feeble schedule might be only a decent team, and treating it as a powerhouse would flatter everyone who played it. This is the so-called circularity problem: to know how good your opponents are, you must know how good their opponents are, and so on without end.
The standard way out is to solve the whole system at once. The idea behind most ratings that adjust for schedule is the following.
- Give every team an unknown rating.
- Say that the expected margin in a game between two teams equals the difference in their ratings, perhaps with an allowance for home court.
- Find the set of ratings that best fits all the results of the season.
This is a type of regression, and it is how many well-known systems work in practice. The result is a number for each team that reflects both how it performed and who it performed against. The chain of opponents' opponents is resolved all at once rather than step by step.
The simple rating system
A transparent version of this idea is what is often called the simple rating system, or SRS. It defines a team's rating as its average point margin plus the average rating of its opponents. In symbols, rating equals margin plus schedule strength, where schedule strength is itself the mean of the opponents' ratings. Because the ratings appear on both sides of the equation, the system has to be solved by iteration or by algebra, but the intuition is easy: a team's rating is how much it wins by, corrected for how good the teams it beat were.
Here is a small illustration with invented numbers. Suppose Team P has an average margin of plus 8 points per game, but its opponents average minus 4 in rating. Its adjusted rating is about plus 4. Team Q has an average margin of only plus 5, but its opponents average plus 3. Its adjusted rating is about plus 8. Judging by raw margin, P looks better; judging by adjusted rating, Q does, and by a wide gap. Which view is more accurate? Over many games, the adjusted one, because Q's margin was earned against stronger competition and P's against weaker.
Adjusting efficiency rather than margin
Modern college ratings usually go a step further. Rather than adjusting only the final margin, they adjust offensive and defensive efficiency, measured per possession. The approach underlies the best-known college ratings: each team gets an adjusted offensive efficiency and an adjusted defensive efficiency, which tell you roughly how many points per 100 possessions the team would score and allow against an average opponent. The difference between them is an adjusted net efficiency, a close relative of the net rating discussed elsewhere on the site. We cover the details in college efficiency metrics: KenPom, NET and the ratings used to select the field.
Using efficiency rather than margin has two advantages. It strips out the effect of tempo, so slow teams are not penalized for winning ugly. And it separates the two ends of the floor, so you can see that a team's strength is built on defense, offense or both. The adjustment mechanism is the same as before: an offense that scored 1.10 points per possession against a defense that usually allows 1.00 earns more credit than one that scored 1.10 against a defense that usually allows 1.15.
The details that matter
A rating that adjusts for schedule can be made better, or worse, by a handful of modeling choices.
Home and road
Teams win more often at home. If you ignore that fact, a team with a lot of home games will look better than it is, and a team with many road games worse. The usual fix is to include a home-court term: the expected margin is the difference in ratings plus a number of points for the home team, and the size of that number is estimated from the data. The size of the effect and its causes are discussed in home-court advantage in the numbers. When you hear that a team's schedule has been difficult because it played many road games, this is the logic behind the claim.
Rest and travel
In the NBA, a team playing the second night of a back-to-back, or the fourth game in six nights, often performs below its usual level. Some analysts include a rest adjustment in their schedule-strength calculations. The strategic side of this problem is in back-to-backs, rest days and travel, but from the viewpoint of a rating system the point is that identical opponents are not identical tests if one comes on a rested night and another on a tired one.
Blowouts and diminishing returns
If a rating uses raw margin, a 40-point win counts twice as much as a 20-point win. Yet the second 20 points rarely say as much about quality as the first, since starters sit and the game's character changes. Many systems cap the margin credited for any one game or let it taper off. Capping reduces the incentive to run up scores, which is a particular concern in college ratings that might otherwise be gamed, and it also tends to improve predictions.
Recency
A team in January is not the team it was in November. Injuries, trades, and growth within a season change what a team is. Systems that weight recent games more heavily follow the team's current form but are noisier, since they use less data. Systems that weight all games equally are stable but slow to react. Choosing between the two is a trade-off with no universal answer.
Strength of schedule and strength of record
One distinction deserves a section of its own, because it prevents a common confusion. A rating that adjusts for schedule and is built to predict future games is answering the question, how good is this team? A different kind of measure is built to reward what a team has accomplished, given the schedule it actually played. The second question is, how impressive is this team's record?
Suppose a team is somewhat better than average but has an extremely hard schedule, so it finishes 18-12. A predictive rating might call it a strong team, because it is, in terms of ability. A resume measure might also rate it highly, because 18 wins against that schedule is hard to come by. Now take a team that is the same quality but has an easy schedule and finishes 26-4. The predictive rating still sees the same team, just a bit better than average, but the resume measure sees a tremendous record against weak opposition and might not be impressed, or might be impressed by the sheer number of wins depending on its design. The two measures can disagree, and both can be correct for their own purpose.
In the context of tournament selection, the argument over which to prefer is, essentially, whether the committee wants the best teams or the most deserving teams. The quadrant system, which sorts wins and losses by the quality and location of the opponent, is a way of encoding deservingness: a win on the road against a top team is worth much more than a home win over a team near the bottom.
Where the adjustments go wrong
No schedule adjustment is flawless, and the following problems come up repeatedly.
- Thin connections. In a league with many conferences, there are few games between them. If conferences rarely play each other, the data linking their ratings is sparse, and the estimate of how one league compares with another rests on a few non-conference games. The ratings within a conference are reliable; the comparison across conferences is much less so.
- Preseason assumptions. Some systems begin each season with a prior estimate of a team's quality drawn from the previous year, then update as games are played. Early in a season, that prior can dominate and make the rating reflect last year as much as this one.
- Roster churn. Teams change from year to year and during the year. A rating built on the games of a full season describes the average version of the team, which may not match the version playing now.
- Garbage time and style. A team with a slow, controlled style compiles smaller margins and can look weaker than it is on margin-based ratings. Efficiency-based ratings help but do not eliminate the issue.
- Sample size. With only 30 games a team has not produced enough data to pin down its rating precisely. Add the noise of close-game luck, and ratings carry meaningful uncertainty. Our look at small samples in basketball explains why.
A related approach worth knowing is the Elo family of ratings, in which each game updates the ratings of both teams by an amount that depends on how surprising the result was. Beating a strong team moves the needle a lot; beating a weak team barely at all. That is schedule adjustment built into the update rule, and we describe it in Elo ratings in basketball.
A practical way to read a schedule
You do not need a statistics package to apply the idea sensibly. A few habits go a long way.
- Look at who, where and when. For any impressive or disappointing record, scan the opponents, the number of home and road games, and whether a long stretch came against strong or weak teams.
- Prefer adjusted numbers to raw numbers for comparisons across leagues. If you are comparing teams that do not play the same opponents, an adjusted rating is worth more than a raw record.
- Be wary of extremes. A team ranked first because of a beautiful record in a weak league is a candidate for an overrating; a team ranked low because of a brutal schedule might be better than it appears.
- Check the gap between record and rating. A team whose record is much better than its adjusted rating is either lucky or has been helped by the schedule; both tend to fade.
- Remember the future. The remaining games matter at least as much as the past games when you are asking whether a team will make the playoffs. The full list of upcoming opponents is on the schedule page, and current records are on the standings page.
If you would like to try the math yourself, how to build a simple team rating at home shows how to set up a basic version of the system described above using nothing more than a spreadsheet and the season's results.
What to take away
Strength of schedule is a reminder that a record is not a measurement of quality in isolation; it is a measurement of quality against a specific set of opponents. In a nearly balanced league like the NBA, schedule differences are small and a team's net rating is a good guide as it stands. In college basketball, where the gap between conferences is large and the number of games small, adjusting for schedule is essential to say anything sensible about who is good. The best adjustments solve for every team's rating at once, account for home court and, where possible, work on a per-possession basis. They remain estimates, with real uncertainty, so treat them as strong evidence rather than final answers.