Pressing

Summary

Per-team-season aggregates of where each team wins the ball back. The team style profiling on /analysis/style already exposes a z-scored "pressing" dimension, but a z-score hides the denominator and conflates two different questions: where did you win it back vs how often. The pressing page surfaces both: the share split (att / mid / def of total recoveries) plus per-match volume.

The single-axis press_height_index = (att − def) / total lives in [−1, 1]. Positive = recoveries skew the attacking third (active high press), negative = recoveries skew the defensive third (deeper block). Most teams are negative because the league-wide split sits around 10% / 45% / 45%; the question is who is less negative.

The feature lives at /analysis/pressing and is also exposed on the team quadrant page (Pressing Height, Press vs Possession).

Inputs

Sums are over regular-season team-matches only (matches.is_playoff = 0), matching the default scope of every other team-season aggregate (form, splits, xPts, SQI, set-pieces).

All three keys are present in every season the feed covers (2019-2025+), so there is no per-season metric availability gap.

Derived rates

Computed server-side on read from the stored counters. None when the denominator is zero.

• Att 3rd share = poss_won_att_3rd / poss_won_total
• Mid 3rd share = poss_won_mid_3rd / poss_won_total
• Def 3rd share = poss_won_def_3rd / poss_won_total
• Recoveries / match = poss_won_total / matches
• Att 3rd / match = poss_won_att_3rd / matches

Per-match volume is paired with share because they answer different questions. Share answers "where on the pitch", which is mostly tactical. Per-match volume answers "how often", which is partly tactical (aggressive teams force more turnovers) and partly a function of how often the team loses the ball in the first place.

Pressing height index

A single composite axis for the team quadrant page:

press_height_index = (poss_won_att_3rd − poss_won_def_3rd) / poss_won_total

Range [−1, 1]:

• +1 = every recovery in the attacking third (max press).
• −1 = every recovery in the defensive third (deepest block).
• 0 = balanced (or all middle-third).

This is precomputed in compute_pressing.py so the same number powers both the dedicated page and advanced_metrics.py. No z-scoring involved — it's a pure, per-team-season ratio that's directly comparable across seasons.

Data caveats

• Tempo confound on per-match counts. A team that loses the ball more often will recover it more often in absolute terms. Share metrics cancel this out (they sum to 1 within a team-season); per- match counts don't. Read share as the primary signal for tactical intent and per-match as the secondary signal for activity.
• Pitch-thirds, not pitch-zones. The feed only emits the three thirds. There's no way to split a recovery in the attacking third by flank or central, or by distance to goal.
• 2020 Island Games had 7-11 matches per team in a neutral bubble without travel, rest variance, or crowds. Rows from 2020 are flagged on the page and should not be compared directly to other seasons.
• Recovery, not pressure. SDP's possWonAtt3rd counts where the ball was recovered, not where pressure was applied. A team that presses high but invites turnovers in deeper zones will look like a deep-block team on this metric. Pairing with PPDA on the quadrant page partially mitigates this — PPDA captures intensity, height captures location.

Why this isn't a model

Same reason as set-pieces: the data volume is small (~55 team-seasons), and the underlying counts are large enough that simple ratios already carry the signal. Z-scoring within season (as team_style does) hides the denominator and forces the reader through one extra layer of interpretation. This page sticks to raw counters + direct rates + one bounded composite index.