Sweeney tracker

Where each metric stands

False alarms: the share of teams for which a rule fired when 2024–25 games were checked against their own 2023–24 limits. Teams that changed coaches were not removed, so the rate is an upper bound.

Beyond the series: how much has Orlando changed?

Ten measures of how Orlando plays, beyond the ones from the series. Each change is Orlando's season-to-date value minus its 2025–26 value. The rank compares it with every other team's change from its own 2025–26 season, by size. Unusual means one of the 3 largest changes among the 30 teams. It says how unusual a change is, not what caused it.

Rate changes are in percentage points. A row compared with each opponent's usual measures every game against what that opponent normally allows or produces, so its values are above or below that norm; it is used only where it makes the measure more stable. Returning players repeats the change for rim attempt share and 3PA rate using only players who shot for the same team in both seasons, ranked against the same measure for the other teams. All ten together is the sum of Orlando's squared changes, each divided by this season's spread of that change across the 30 teams; its change column shows that sum. These rules were locked on Nov 1, 2026. The first 5 games were played before then.

Context

Does better defense still create easier offense?

“The better we play on offense, the better we defend. The better we defend, the better we play on offense.”

The post identified three useful metrics for seeing how good defense leads to good offense. Half-court scoring after a made basket or a dead ball is the one to watch.

Orlando points per possession after an opponent made basket or a dead ball, when the defense is likely set. Higher is better.

Share of Orlando possessions that start after an opponent live-ball turnover. Higher is better.

Opponent points per possession right after an Orlando made basket. Lower is better.

C3 rests on about 24-35 opponent possessions a game after an Orlando make, and its split-half reliability is low (r = 0.35), so it needs about 77 games to reach reliability 0.5. It will read 'too early' for most of the season.

More shots, better shots, and who takes them

“...getting more shots, getting better shots than our opponent”

Location values use this season’s league results through the latest update, with Orlando left out. As more league games come in, those values are restated, so past points for opportunity volume and location move slightly. If a restatement ever withdraws a signal that already fired, the chart keeps the record of when it fired and when it was withdrawn.

Scoring opportunities (FGA + 0.44 × FTA) per 100 possessions, Orlando minus opponents. Higher is better.

Points per 100 added by taking more scoring opportunities than the opponent. Higher is better.

Points per 100 added by where the shots come from, valued at league rates for five broad locations. Higher is better.

Shooter identity

This number moves with the roster. In the 2024–25 test it fired for 97% of teams, so start with the split. Mix is how returning players have changed where they shoot, holding their volume at last season’s level. That is the part closest to a coaching decision. Composition is everything else, including who is playing and Paolo Banchero’s absence.

Offensive shooter identity, points per 100 possessions. Largest mix contributions so far: Desmond Bane +0.66, Goga Bitadze −0.16, Anthony Black −0.12. 13 returning players.

Points per 100 added or lost by who takes the shots, using each shooter’s frozen preseason expectation. Higher is better.

Is the team playing to its roster?

“There’s obviously talent here - number of talented guys that I’m excited to coach.”

This metric provides context but doesn't give us direct signal. Its reliability estimate rests on 2025–26 alone and it has no backtest, because no earlier preseason DARKO ratings were saved. It also takes the rotation as given: it asks how Orlando played with the players who played, not whether the minutes went to the right players.

Net Rating minus the rating predicted from the roster’s DARKO ratings and actual minutes. Higher is better.

Who played

Minutes from PBPStats. Chart tooltips also list who did not play.

The posts behind the metrics

1. 1

   How Much "Talent" Does Sean Sweeney Inherit?

   Sep 2, 2026 · 6 min read

   The first post in a series that breaks down some comments from new Magic head coach Sean Sweeney in his introductory press conference. I'll be looking to quantify the statements, as well as set a framework for evaluating how the team is performing under his leadership according to his stated goals.

2. 2

   Shot Economy: More and Better

   Sep 9, 2026 · 14 min read

   In this post we look into the frequency and broad-location value of Orlando's shot attempts relative to their opponents in order to set a baseline against which we can evaluate Sweeney's stated desire for more and better shot opportunities.

3. 3

   How Much Offense Can Orlando's Defense Create?

   Oct 10, 2026 · 5 min read

   No advanced stats are needed to tell us that Orlando's offense needs to show a leap this season if a postseason run is the goal.

4. 4

   What is a floor raiser, and does Orlando have one?

   Oct 25, 2026 · 6 min read

   Exploring the concept of a floor raiser and its relevance to the Orlando Magic's roster decisions.

Methods and limitations

The rules were locked on October 27, 2026, before any 2026–27 value was computed, and amended once the same day. Definitions match the series posts. One possession count is used everywhere: PBPStats counted possessions, one shared count per team-game.

Detection

For T1 and the shot-economy metrics, the center line is the mean of Orlando’s 2025–26 game values. For the complementary-game metrics it is Orlando’s 2025–26 PBPStats season value, with limit widths from 2024–25 games, because PBPStats does not serve per-game possession starts for 2025–26. Limits sit 2.66 average moving ranges from the center and never change during the season.

When a season-to-date value means much

Split-half reliability across all 30 teams (2023–24 through 2025–26; T1 from 2025–26 only), stepped up with the Spearman-Brown formula, gives the games each metric needs before its season-to-date value reaches a reliability of 0.5.

What these numbers cannot see

• DARKO ratings describe expected performance, not latent talent, and rookies start near a common baseline.
• Broad shot location is not shot quality. It does not see openness, defender distance, shot clock or difficulty within a zone.
• The shot-economy terms are accounting over the same attempts, not causes. The residuals are unexplained and are not assigned to coaching.
• Possession-start type does not see cross-matches, floor balance, shot clock or the first action.

Sources

PBPStats for possessions, points, possession starts and minutes; NBA play-by-play through nba_api for shots; DARKO for player ratings (July 8, 2025 and August 7, 2026 snapshots, frozen). Raw data stays private. This page shows derived values only.