File size: 4,736 Bytes
2bf53c7
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
---
language: en
license: cc-by-4.0
pretty_name: ProScout Players (Synthetic Football Scouting Dataset)
size_categories: 1K<n<10K
task_categories:
- information-retrieval
- recommendation
- text-classification
- tabular
tags:
- football
- soccer
- scouting
- synthetic-data
- education
---

# ProScout Players — Synthetic Football Scouting Dataset (1,112 rows)

A **clean, fully synthetic** dataset of realistic football player profiles built for **text → top‑3 player recommendations** (ProScout). It’s classroom‑safe (no PII, no scraping), diverse across positions/nations, and structured for retrieval demos.

> **Core use case:** a scout types a brief (e.g., “22–25 left‑footed inverted winger with pace+dribbling in European competition”), the system ranks candidates, and returns the **top 3** with bios and attributes.

---

## 1) Dataset Summary

- **Rows:** 1,112 players  
- **Source:** programmatically generated (no real players)  
- **Format:** CSV (tabular)  
- **Primary consumer:** ProScout (text query → recommendation)  

### Columns
- `name` — realistic football-style first + last name  
- `Nation` — country string (e.g., Spain, Brazil)  
- `position` — one of: `GK, RB, LB, CB, CDM, CM, CAM, RW, LW, ST`  
- `dominant_foot``Right | Left | Two-footed`  
- `age` — integer 17–39  
- `Goals and Assists` — string, e.g. `"12G 7A"`  
- `height_cm` — integer 165–200  
- `weight_kg` — integer 60–95  
- `play_style` — e.g., *Pressing forward, Box‑to‑box, Inverted winger, Regista, Ball‑playing defender, Sweeper keeper…*  
- `key_strengths`**3 strengths** (comma‑separated) chosen from: pace, dribbling, finishing, vision, passing, strength, stamina, tackling, positioning, heading, first touch, off the ball, composure, work rate, aggression, long shots, crossing, set pieces, 1v1 defending, interceptions  
- `league_context``Domestic league | European competition | International`  
- `bio` — 2 short sentences auto‑generated from the row fields  

---

## 2) Intended Uses

- **IR / Retrieval:** rank players against a natural‑language brief  
- **Recommendation:** shortlist top‑3 candidates for a role  
- **Pedagogy:** demos of vector search, re‑ranking, rule‑based filters  
- **Prototyping:** LLM prompts and RAG over structured attributes  

Not intended for real‑world transfer decisions or media reporting about real players.

---

## 3) Example Usage

### Load with pandas
```python
import pandas as pd
df = pd.read_csv("proscout_players_1112.csv")
df.head()
```

### Simple rules + ranking sketch
```python
query = {
    "position": "LW",
    "foot": "Left",
    "must_strengths": {"Pace","Dribbling","Crossing"},
    "age_min": 22, "age_max": 25,
    "league": "European competition"
}

def score(row):
    s = 0
    s += 10 if row["position"] == query["position"] else 0
    s += 6 if query["foot"].lower() in row["dominant_foot"].lower() else 0
    strengths = {x.strip().lower() for x in row["key_strengths"].split(",")}
    s += 3 * len({k.lower() for k in query["must_strengths"]} & strengths)
    s += 4 if query["league"].lower() in row["league_context"].lower() else 0
    s += 2 if query["age_min"] <= row["age"] <= query["age_max"] else 0
    return s

df["score"] = df.apply(score, axis=1)
top3 = df.sort_values("score", ascending=False).head(3)
print(top3[["name","Nation","position","dominant_foot","age","key_strengths","league_context","bio","score"]])
```

---

## 4) Generation Process (Synthetic)

- Names sampled from common football‑style first/last names.  
- Positions, feet, play styles, contexts sampled from curated vocabularies.  
- Ages, heights, weights sampled within realistic pro ranges.  
- Goals/assists and key strengths randomized but **constrained** to plausible values.  
- Bios programmatically assembled from each row’s attributes.

No external scraping. No real player matching.

---

## 5) Quality, Limitations & Ethics

- **Quality:** realistic enough for demos; not statistically calibrated to any specific league.  
- **Bias:** distributions are synthetic; do **not** infer real‑world prevalence from them.  
- **Limitations:** no time dimension, team assignment, injury history, wages, or contracts.  
- **Ethics:** purely synthetic; safe for teaching and prototyping. Avoid portraying rows as real people.

---

## 6) Licensing & Citation

- **License:** CC‑BY‑4.0 (data and docs).  
- **How to cite:**
  > ProScout Players (2025). Synthetic football scouting dataset (1,112 rows). CC‑BY‑4.0.

---

## 7) Maintainers

- Course project: *ProScout* — European scouting recommender (text → top‑3).  
- Contact: add your names / emails or course channel here.