LEYTON ORIENT · INTELLIGENCE WORKSPACE

From matches to identity.

Find persistent patterns. Test what might come next.

ORIENT
INTELLIGENCE

Aggregating fixture evidence…

PLAYING STYLE

How this team uses the ball

Pass share by temporal phase. Counts are pooled across the selected fixtures.

Phase shares describe passing behaviour. They do not measure time in possession or establish a league-wide style ranking.

RECORDED TEAM SHAPE

Starting formations

Source lineup formations, separate from average passing positions.

Match rows below expose recorded formation changes. Starting-shape filters select whole fixtures; they do not isolate minutes played in that shape.

CROSS-MATCH NETWORK

The connections that keep returning

Network viewMinimum distinct fixtures

Up to 12 directed links, ordered by distinct-fixture recurrence then pass count. Node locations use supplied coordinates; zone bands assume increasing x is forward. Player averages may combine different lineups and roles.

ROUTE PRIORITIES · EXPERIMENTAL FORECAST

Which repeated routes deserve the next review?

Target: a shot within 10 seconds after completing a three-pass route, before control ends. Candidates need six eligible occurrences across three fixtures. Ranking uses the lower observed confidence bound, not raw conversion alone.

Route / phase	Fixtures / eligible	Shots	Observed / 95% interval	Smoothed estimate	Recorded xG	Evidence

Select a route to inspect its occurrences.

CONSISTENCY AND OPPOSITION

What changes from match to match?

Full-fixture style metrics. The phase filter above affects networks and route forecasts only.

Fixture	Starting shape / changes	Pass completion	Forward ≥10 grid units	Regain → loss median	Shot within 10s of regain

Methods and forecast limits

Style rates pool numerators and denominators from the same fixture sample. Turnover timings pool completed regain-to-loss intervals. Forward-pass share uses all coordinate-complete pass attempts; unknown coordinates are excluded. Starting formations and changes come from event types 34/40 and qualifier 130 using the repository's formation dictionary. Unknown codes stay explicit.

Network receivers follow the immediately next teammate on-ball item in original feed order. Routes reuse the non-overlapping three-pass zone motifs and control-spell definitions in Match centre. Routes whose completion cannot be located inside their control spell are excluded. Short incomplete outcomes are censored, unless a shot was already observed. Missing xG suppresses the route total.

The smoothed estimate combines route outcomes with ten observations at the selected phase's baseline rate. It is a provisional frequency forecast conditional on completing that route, not the probability of successfully attempting it or proof it is the best tactical choice. Opponent pressure, player availability and defensive organisation are not modelled.

The forward test trains only on strictly earlier dates, then scores eligible later occurrences of supported routes. Brier score measures squared probability error; lower is better. The comparator uses the same training phase baseline and the same test occurrences. Routes with insufficient history are not scored. Confidence intervals are descriptive Wilson intervals and do not account for within-match dependence. A small, selected local sample does not validate a forecasting model.