The Sacramento Kings won nine straight games in November 2022. The vibes, around game six, became unbearable. The light-the-beam ceremony at Golden 1 Center had passed from inside joke into national broadcast feature, and a city that had endured fifteen years of professional misery was suddenly tilting toward a question that felt unfamiliar: are these guys actually good?
The honest, deeply unsatisfying answer at the time was: maybe. Probably. We need three more weeks to be sure. The Kings went on to finish third in the West that season and lose a competitive first-round series. The streak was real. The trajectory it implied was, at minimum, partially real. Distinguishing between those two things is most of what reading team trends well looks like.
This piece is the thinking-out-loud guide to that distinction. It is not a model. It is a checklist a writer or fan can run, in their head, before they declare a team “back” or “cooked” on the basis of nine games.
Sample size, but with feeling
You have heard the slogan. Sample size matters. The way it actually matters, in practice, is less mathematical and more procedural.
A nine-game NBA winning streak is, statistically, not enough to declare a team improved at a meaningful level above its preseason expectation. The standard error around a team’s true net rating, given nine games, is roughly four to five points. That means a team posting a +6 net rating across those games could easily be a true +2 team having a good month, or a true +9 team being slightly held back by schedule. Either is consistent with what you saw.
What changes the math is when the trend persists across different sub-samples. If the Kings had been +6 in their first nine, +5 in their next twelve, and +5 across a road trip out west, the trend would stabilize. If they had been +9 in their first nine and then dropped to +1 in the next twelve, the streak would be revealed as the unusual run.
So the working rule, when you see a team on a tear, is: wait for the second sample. The first hot stretch is data. The second one is evidence. The third one is an argument.
Schedule strength, and why it lies
Strength of schedule is the most weaponized stat in early-season coverage. Every fan base whose team is winning insists the wins were earned against a brutal calendar. Every fan base whose team is losing insists the losses are inflated by an unfair early run. The truth is messier and more boring.
Opponent-adjusted net rating attempts to fix this by subtracting the average quality of the teams faced. The version Basketball Reference publishes is a serviceable starting point. But schedule strength is not just about opponent quality. It is also about:
- Back-to-backs. Teams playing the second night of a back-to-back lose at a meaningfully elevated rate, especially on the road.
- Travel. A team playing four games on a five-day road trip across three time zones is functionally a different team than the rested version. Coaching staffs adjust. Fans frequently do not.
- Rest disparity. When you played your opponent’s rested team versus their tired team is invisible in the season-long schedule strength number.
- Opponent injury status. A win against a Lakers team without LeBron and AD is not a win against the Lakers.
The version of schedule strength that actually works requires watching games and reading injury reports, not just glancing at season-long opponent winning percentage. It is also why “the Suns beat three teams above .500 last week” is, by itself, a useless sentence.
Injuries: the most underweighted variable
Most failed team-trend takes are, on rewatch, injury stories. A team gets a starter back from a six-week absence. Their net rating climbs five points. Pundits call it a “turn.” It is, in fact, returning to a baseline. A team loses its starting center for ten games. Their defense craters. Pundits write the season’s eulogy. The center comes back. The defense is fine.
The discipline here is to track what a team looked like with their actual rotation, separated from the small-sample chaos of a banged-up week. NBA stat sites publish minutes-restricted on/off splits that, for the price of a few minutes of digging, will tell you whether a team’s “improvement” was a real change or a return to health.
The same logic, with more brutality, applies in soccer. A Premier League side losing its first-choice central midfielder for six weeks rarely shows a clean trend line. The team that took the field is not the team the manager intended. Eight weeks of data with that absence and eight weeks without it can produce opposite conclusions about the same squad.
Role changes and rotation experiments
Coaches change. Lineups change. Sometimes a team’s “trend” is just the period in which the head coach finally figured out that his second-year wing is, in fact, a better starter than his veteran. Identifying these moments requires looking past the box score at which lineups are actually playing the most minutes.
A working example. The 2023-24 Orlando Magic spent the first six weeks of the season starting one lineup and the next six weeks starting another, after an injury forced a rotation experiment that the staff ended up liking. The team’s net rating profile, viewed across the whole season, looked like steady improvement. The version of the team that ended the season was, in lineup terms, not the same team that started it. A “trend” piece written in February would have missed the structural change.
Watch for:
- New starter inserted in week six. Does the bench unit’s net rating change because the substitute slid into a more comfortable second-unit role?
- Coach pulls a specific lineup off the floor after a bad stretch. Was the lineup actually bad, or did its limited minutes happen against tough matchups?
- A rookie’s role expands. The team’s offense looks different because the offense literally is different now.
None of this shows up in the season-long net rating. All of it shows up in five-man lineup data and minutes distributions, which are public on Basketball Reference and the NBA’s official site.
Regression to the mean: the most misused phrase in the genre
“Regression to the mean” is a real statistical phenomenon. It is also, in the hands of sports writers, occasionally a way to say “I was wrong and have decided to make that someone else’s problem.”
The strict version of regression to the mean is that extreme values, on subsequent measurements, tend to be closer to the average than the first measurement was. A team shooting 42% from three over a fifteen-game stretch will, almost certainly, shoot closer to its true talent level over the next fifteen games. A striker scoring at twice his xG over ten matches will, almost certainly, score closer to his xG over the next ten matches.
The thing regression to the mean does not tell you is which way the regression will go. A team can regress upward. A team can regress downward. The phrase is often deployed as if it predicts the direction, which it does not. It predicts the magnitude of the deviation, not the sign.
The further trap is timing. Regression happens over a long horizon. A team that “should” regress can defer that regression for two months. By the time the math catches up, the playoffs are already underway. The 2018-19 Brooklyn Nets and the 2014-15 Atlanta Hawks are two of the more famous examples, in opposite directions, of regression arriving on its own schedule.
Where this gets weird
A short list of trend-reading traps that have humbled me personally.
Year-over-year coaching changes. A team’s trend in October is partly the trend of last year’s roster learning a new system. Five weeks of growing pains can look like a downward trend that resolves cleanly by mid-November.
Three-point variance. Most NBA team trends, if you decompose them, contain a meaningful chunk of three-point shooting luck. A team can post a +5 net rating stretch where most of the delta is the bench going 41% from three. Same lineup, same plays, different month. The bench goes 33%. The trend “reverses.”
Strength of opponent illness. A common mid-season pattern: a team beats three opponents whose star players were out, then loses two to teams at full strength. The “trend” inverts in a week. Reading the trend without reading the inactive lists in front of it is reading the wrong thing.
Vibes leaking into coverage. A team that is fun to write about — a Sacramento, an OKC, a Lyon — gets more flattering trend pieces. A team that is not fun to write about — most of the Pistons era, the late-2010s Hornets — gets fewer. Reader instincts are right to discount the first kind and pay attention to the second.
The five-question test before you call it a trend
Before declaring a team improved, declined, or “back,” ask:
- Has this held across at least two non-overlapping ten-game samples?
- Has the rotation been roughly consistent, or has the team been integrating a returning starter?
- What does opponent-adjusted net rating say, with injury context factored in?
- Is the offense or defense doing the work? Three-point luck or rim defense regression are different stories.
- Would I bet on this trend continuing for the next twenty games, at fair odds? If not, why am I writing a 1,500-word piece declaring it real?
The fifth question is the one that, in a year of writing about this, has produced the most edits to my own takes before they went live.
Sacramento, three years later
The Kings were not a one-year team. They were not, however, the team the November 2022 nine-game streak implied either. Their actual trajectory was somewhere in between — a real improvement on the previous decade, a real but ceilinged team, a real story that did not end in a championship. Most team trends end up like that. The truth is closer to the middle than the streak suggested, and farther from the floor than the bad month implied.
If you remember nothing else from this, remember the second sample. The first hot stretch is data. The second hot stretch is evidence. The third one is the start of an argument worth having. For more on reading individual players inside these team contexts, our piece on player stats that matter more than raw totals is the obvious next read.
