ghost-outfield-position-only-v1

Ghost-Outfield Position-Only v1 — League-Average Rearguard Positioning (Velocity-Less)

The velocity-less variant of Ghost-Outfield v1: predicts a league-average outfield rearguard position per tracking frame, team, and lateral slot, using a 16-feature vector with the 4 velocity features dropped. Built for StatsBomb-360 freeze-frames, which carry no per-player velocity. Point estimate (ghost_gr_x/y, goal-relative) served as the exact pickle-free boosted HistGradientBoostingRegressor mean.

Part of the silly-kicks soccer analytics library (TF-60 rest-defense; ADR-087).

Why a separate variant

StatsBomb-360 freeze-frames have no temporal history, so the 4 situational-velocity features of the default variant (ball velocity components, ball speed, opponent-forward centroid velocity) cannot be computed. This variant is trained without them so it is valid on velocity-less input. The serve seam auto-selects it from the frames' declared velocity availability; if it is not bundled, a velocity-less frame yields honest-NaN — never the default velocity model, which is invalid on velocity-less input.

Variants

Model description

• Point estimate only. ghost_gr_x/y serve the exact boosted HGBR mean, reconstructed pickle-free (baseline + Σ_trees leaf_value; no sklearn at inference). No KDE / density read-out.
• Possession-conditioned & leakage-safe: one model serves both the in-possession rest-defense rearguard and the defending line (a live team_in_possession feature); no input encodes the modeled team's own rearguard coordinates (the target).
• Parameters-only, pickle-free: npz + JSON metadata + SHA-256 checksums; no raw provider tracking data (ADR-011/044).
• Fail-closed load: SHA-256 + behavioural chirality + feature-contract; the chirality frame hash is pandas-major-invariant.
• FOV honest-NaN (StatsBomb-360): an insufficiently-observed rearguard region is returned NaN (ghost_outfield_source="fov_cropped"), never a fabricated ghost from the deepest-n visible players.
• Hyperparameters: HistGradientBoostingRegressor, 500 trees, max depth 8, 5-fold CV; trained at 1 frame/second.

Metrics

Slightly higher MAE than the default variant, as expected: dropping the velocity features removes information. Use default on velocity-bearing tracking; use this variant only on velocity-less input.

Training data

179 matches / ~4.17M per-(frame, team, slot) rows (at 1 fps) of licensed professional tracking (Sportec/DFL Bundesliga, SkillCorner, Gradient Sports FIFA World Cup 2022). Only the learned parameters are published; no raw provider tracking data is redistributed (parameters-only artifact). Trained from a clean, CI-green commit.

Usage

import silly_kicks.tracking as tracking

# Auto-selected on declared velocity-less freeze-frames when model=None; or explicitly:
model = tracking.GhostOutfieldModel.from_variant("position_only")
served = tracking.serve_ghost_outfield_positions(frames, model=model, home_team_id=1)

Input frames must be LTR-normalized (home attacks right). All coordinates are goal-relative to the modeled team's defended goal.

Limitations

• League-average positioning, not a specific player's or team's style.
• Velocity features dropped — use default where velocity is available.
• Static per-frame estimate; LTR normalization required.

References

Le et al. 2017 (Data-Driven Ghosting, MIT Sloan). Boosted-mean point-estimate variant of the ghosting concept (no density estimation). See the silly-kicks NOTICE for the full citation.

More information

• License: MIT · Library: silly-kicks (v4.109.0+) · GitHub · ADR-087.

F1b re-fit (4.128.0)

F1b float32-frame re-fit (silly-kicks 4.128.0 / ADR-106; training_commit=3ca609f). Re-fit on float32-stored frames (the storage rounding exceeds the trained-model feature-contract atol); the metrics table above reflects this re-fit, and the mirror is republished to match the wheel.