A reproducible, Python-based decision-support framework for football recruitment and succession planning.
This project demonstrates how player-performance data can be transformed into structured recruitment decisions through statistical profiling, role-specific analysis, risk adjustment, scouting validation, and human decision governance.
The project models succession planning for three Newcastle United players:
- Anthony Gordon
- Bruno Guimarães
- Sandro Tonali
Rather than treating statistical similarity as a transfer recommendation, the framework progressively narrows candidates through role profiling, evidence assessment, development potential, recruitment risk and structured scouting validation.
Decision-support principle: Analytics should narrow the search, quantify uncertainty, and improve decisions — not remove the need for football judgement.
Priority target: Kristian Arnstad
Strong shortlist: Tobias Bech
Priority targets: Thomas Jørgensen, Rihito Yamamoto
Strong shortlist: Darío Osorio
Priority targets: Victor Froholdt, Kodai Sano, Irakli Yegoian
Strong shortlist: Max Ejdum
These are analytical recruitment priorities rather than final transfer recommendations. Final approval remains subject to human scouting and football-domain assessment.
Read the Anthony Gordon case study
Read the Bruno Guimarães case study
Read the Sandro Tonali case study
- Multi-league player profiling and recruitment analysis
- Role-specific candidate identification
- Risk-adjusted recruitment scoring
- Structured scouting-validation workflows
- Human-in-the-loop decision governance
- Succession planning for priority squad roles
- Reproducible QA and frozen analytical releases
Player & Performance Data
│
▼
Recruitment Metrics
│
▼
Role / Player Profiling
│
▼
Risk-Adjusted Decision Model
│
▼
Executive Shortlist
│
▼
Scouting Validation
│
▼
Human Decision Gate
│
├──────────────► Recruitment Decision
│
└──────────────► Succession Planning
The project addresses three Newcastle United succession problems:
- Anthony Gordon
- Bruno Guimarães
- Sandro Tonali
The analytical workflow is designed to answer five progressively narrower questions:
- Which players are statistically plausible successors?
- Which candidates provide the strongest overall succession fit?
- How confident should Newcastle be in the available evidence?
- Which candidates warrant further scouting and validation?
- Has sufficient human evidence been collected to support a final recruitment recommendation?
The result is not simply a player ranking.
It is a staged recruitment decision-support system.
PLAYER DATA
|
v
PHASE 1
Succession Candidate Identification
|
v
PHASE 2
Recruitment Decision Model
|
+--> Succession Score
+--> Evidence Band
+--> Development Profile
+--> Recruitment Risk
+--> Decision Score
+--> Risk-Adjusted Score
+--> Decision Tier
|
v
EXECUTIVE SHORTLIST
|
v
PHASE 3
Scouting & Decision Validation
|
+--> Validation Flags
+--> Validation Questions
+--> Suggested Scouting Action
|
v
PHASE 4
Human Scouting Evidence Gate
|
+--> Role Fit
+--> Tactical Fit
+--> Evidence Review
+--> Development Review
+--> Video / Live Scouting
|
v
FINAL DECISION GATE
|
v
PHASE 5
Executive Reporting & Succession Briefs
A central principle of the project is:
Statistical recommendation is not the same as recruitment approval.
The analytical model is allowed to:
- identify candidates;
- rank candidates;
- quantify succession fit;
- assess evidence quality;
- flag recruitment uncertainty;
- prioritise scouting activity;
- create executive shortlists.
The analytical model is not allowed to produce a final recruitment approval without completed human scouting evidence.
This safeguard is implemented explicitly in Phase 4.
A candidate with incomplete scouting evidence remains:
SCOUTING REQUIRED
and is not eligible for final approval.
Phase 1 establishes the underlying candidate-identification framework.
The purpose of this stage is to move from a broad player population toward players whose statistical profiles suggest potential suitability as successors to the target Newcastle players.
The output of this stage provides the analytical foundation for later decision modelling.
Phase 1 is concerned primarily with candidate discovery rather than final recruitment judgement.
Phase 2 converts succession analysis into a structured recruitment decision framework.
Candidates are evaluated using several decision dimensions, including:
- Succession Score
- Evidence Band
- Development Profile
- Recruitment Risk
- Decision Score
- Risk-Adjusted Score
- Decision Tier
Decision tiers include:
PRIORITY TARGET
STRONG SHORTLIST
SCOUT / VALIDATE
WATCHLIST
The executive shortlist contains only the strongest approved decision tiers:
PRIORITY TARGET
STRONG SHORTLIST
The frozen Phase 2 executive shortlist contains nine candidates across the three succession problems.
Phase 3 converts analytical uncertainty into explicit scouting work.
Rather than treating uncertainty as an abstract model limitation, the system generates validation flags and associated scouting questions.
Examples include:
- evidence review;
- development projection;
- recruitment risk;
- young-player projection;
- decision uplift.
These flags determine the next recommended action for each candidate, including:
TACTICAL VALIDATION
VIDEO VALIDATION
LIVE / VIDEO SCOUTING
PROGRESS TO SCOUTING
Phase 3 therefore acts as the bridge between statistical recruitment analysis and football scouting.
Phase 4 introduces the human validation layer.
The system generates a structured scouting evidence template for every executive candidate.
The scouting framework uses a five-point scale:
1 = Major concern
2 = Below required level
3 = Acceptable / neutral
4 = Strong fit
5 = Excellent fit
Potential scouting outcomes include:
RECOMMEND
CONTINUE SCOUTING
HOLD
REJECT
However, a recommendation is only eligible to progress when the required human scouting evidence has been completed.
Until that evidence exists, the final decision gate remains blocked.
This prevents analytical outputs from being misrepresented as completed football recruitment decisions.
Phase 5 converts the preceding analytical stages into concise decision-support outputs.
The executive master summary combines the key information from frozen Phases 2, 3, and 4.
It provides a single view of:
- player;
- succession target;
- decision tier;
- risk-adjusted score;
- evidence strength;
- recruitment risk;
- scouting action;
- final decision state;
- executive status.
At the current project state, all nine executive candidates remain:
AWAITING SCOUTING VALIDATION
This is intentional because no completed human scouting evidence has been entered.
Individual executive briefs are generated for:
- Anthony Gordon succession
- Bruno Guimarães succession
- Sandro Tonali succession
Each brief ranks the shortlisted candidates and communicates:
- decision tier;
- risk-adjusted score;
- evidence strength;
- recruitment risk;
- development profile;
- required next action;
- current validation status.
The project contains several safeguards against inappropriate interpretation of analytical results.
Players with weaker evidence bases can be penalised or flagged for further validation.
Recruitment uncertainty is represented explicitly rather than being hidden inside a single ranking score.
Candidates cannot pass the final recruitment gate without completed scouting evidence.
Completed and QA-tested phases are stored as frozen versioned artifacts.
This prevents later development work from silently changing previously validated results.
Validated phases are preserved in the frozen/ directory.
Current frozen versions include:
frozen/
├── phase2_v1_0/
├── phase3_v1_0/
└── phase4_v1_0/
Each frozen phase contains the relevant scripts and/or outputs required to reproduce or audit that stage of the decision process.
Frozen files are write-protected after validation.
Important outputs include:
data/outputs/phase2/
recruitment_decision_board_v1_0.csv
executive_shortlist_v1_0.csv
data/outputs/phase3/
scouting_validation_dossiers_v1_0.csv
scouting_action_queue_v1_0.csv
data/outputs/phase4/
scouting_evidence_template_v1_0.csv
validated_scouting_evidence_v1_0.csv
final_decision_gate_v1_0.csv
data/outputs/phase5/
executive_master_summary_v1_0.csv
data/outputs/phase5/succession_briefs/
anthony_gordon_succession_brief_v1_0.csv
bruno_guimaraes_succession_brief_v1_0.csv
sandro_tonali_succession_brief_v1_0.csv
all_succession_briefs_v1_0.csv
Each major analytical phase contains dedicated QA checks.
These checks include:
- expected row counts;
- player consistency across outputs;
- succession-problem coverage;
- required-column validation;
- missing-value checks;
- scoring-range validation;
- approved decision categories;
- scouting-status consistency;
- final-decision safeguards.
Frozen versions are created only after the relevant phase passes QA.
The project is designed as a staged and auditable pipeline.
Later phases consume validated outputs from earlier frozen phases rather than silently recreating historical decisions.
This provides:
- reproducibility;
- traceability;
- version control;
- decision auditability;
- separation between modelling and human judgement.
The model should be interpreted as a recruitment decision-support framework, not as an automated transfer-selection system.
A high-ranked candidate represents a player whose available statistical evidence and modelled characteristics justify greater recruitment attention.
It does not prove that the player:
- will succeed in the Premier League;
- will adapt tactically;
- is financially attainable;
- has appropriate medical availability;
- has suitable character or personality;
- is interested in joining Newcastle United;
- will maintain historical performance;
- should automatically be signed.
Those questions require additional football, financial, medical, contractual, and human assessment.
Player statistics do not fully represent tactical responsibilities, decision-making, off-ball behaviour, communication, or contextual role.
Performance across leagues and teams may not translate directly to the Premier League.
Young players and players with limited minutes may have less stable statistical profiles.
Transfer fees, wages, contract status, availability, injuries, registration rules, and player preference are not fully represented by the analytical model.
The project deliberately leaves final recruitment decisions unresolved until human scouting evidence is supplied.
This is a feature of the methodology rather than an incomplete analytical result.
Phase 1 — Succession analytics COMPLETE
Phase 2 — Recruitment decision model COMPLETE / FROZEN
Phase 3 — Scouting validation COMPLETE / FROZEN
Phase 4 — Human scouting decision gate COMPLETE / FROZEN
Phase 5.1 — Executive master summary COMPLETE
Phase 5.2 — Succession briefs COMPLETE
Phase 5.3 — Project documentation COMPLETE
Phase 5 integration QA is complete and the v1.0 analytical pipeline is frozen.
This repository is the public portfolio edition of the Newcastle United Recruitment & Succession Analytics project.
Original source football datasets and closely source-like processed datasets are intentionally excluded from this repository. Selected derived analytical outputs are retained to demonstrate the recruitment decision model, scouting-validation workflow, final decision gate, and succession-planning architecture.
See DATA_PROVENANCE.md for details on source-data provenance, redistribution scope, and reproduction requirements.
The analytical system is designed for decision support. Statistical outputs identify and prioritise candidates but do not independently constitute recruitment approval; final decisions require validated human scouting and football-domain assessment.