15 KiB
NBA_Go
A high-performance NBA statistics REST API built with Go (Fiber), PostgreSQL, and NGINX. Data is scraped from Basketball Reference and served through a load-balanced, containerized stack with built-in observability.
Architecture
┌──────────────────────────────────────────────────────────┐
│ NGINX (reverse proxy / round-robin load balancer :8080) │
├──────────────────────────────────────────────────────────┤
│ API Instance x3 (Fiber :5000 each) │
│ ┌──────────┐ ┌─────────────┐ ┌───────────────────┐ │
│ │ Routes │→ │ Controllers │→ │ Services (scraper) │ │
│ └──────────┘ └─────────────┘ └───────────────────┘ │
├──────────────────────────────────────────────────────────┤
│ PostgreSQL 15 (GORM ORM) │
├──────────────────────────────────────────────────────────┤
│ Prometheus + Grafana (metrics & dashboards) │
└──────────────────────────────────────────────────────────┘
Project Structure
.
├── main.go # Entry point (API server or import-data mode)
├── import.go # Bulk data import orchestration
├── config/ # Database initialization
├── models/ # GORM models (Game, PlayerAdvancedStat, PlayerTotalStat, etc.)
├── controllers/ # HTTP handlers, DTOs, pagination, filtering, sorting
├── routes/ # Route registration grouped by domain
├── services/ # Web scrapers (Basketball Reference via goquery)
├── utils/
│ ├── middleware/ # Rate limiter, metrics, API key auth
│ ├── metrics/ # Prometheus counter/histogram definitions
│ └── security/ # API key generation & hashing
├── nginx/ # NGINX load balancer config
├── prometheus/ # Prometheus scrape config
├── grafana/ # Pre-provisioned dashboards & datasources
├── docker-compose.yml # Production (Coolify)
├── docker-compose.local.yml # Local development (includes Postgres)
└── docker-compose.override.yml # Override for remote DB development
API Endpoints
| Method | Path | Description |
|---|---|---|
| GET | /api/games |
Game data with box scores, line scores, team/player stats |
| GET | /api/playeradvancedstats |
Advanced stats (PER, WS, VORP, BPM, etc.) |
| GET | /api/playertotals |
Season totals (points, rebounds, assists, etc.) |
| GET | /api/playershotchart |
Shot chart coordinate data |
| GET | /swagger/* |
Interactive Swagger UI documentation |
| GET | /metrics |
Prometheus metrics endpoint |
| POST | /admin/keys |
Create API key (requires X-Admin-Secret header) |
Query Parameters (all data endpoints)
| Parameter | Type | Description |
|---|---|---|
page |
int | Page number (default: 1) |
pageSize |
int | Results per page (default: 20) |
sortBy |
string | Field to sort by (varies per endpoint) |
ascending |
bool | Sort direction (default: false / descending) |
season |
int | Filter by season year (e.g., 2025) |
team |
string | Filter by team abbreviation (e.g., LAL, BOS) |
playerId |
string | Filter by player ID (e.g., jamesle01) |
isPlayoff |
bool | Filter for playoff stats |
Games-specific parameters
| Parameter | Type | Description |
|---|---|---|
date |
string | Filter by date (YYYY-MM-DD) |
gameId |
string | Filter by specific game ID |
include |
string | Comma-separated associations to preload: lineScores, playerGameBasicStats, playerGameAdvStats, teamGameBasicStats, teamGameAdvStats |
Example Requests
# Get top scorers for the 2025 season
curl "http://localhost:8080/api/playertotals?season=2025&sortBy=points&pageSize=10"
# Get a specific game with full box score
curl "http://localhost:8080/api/games?gameId=202501010LAL&include=lineScores,playerGameBasicStats,teamGameBasicStats"
# Get LeBron's advanced stats across all seasons
curl "http://localhost:8080/api/playeradvancedstats?playerId=jamesle01&sortBy=season&ascending=true"
# Get shot chart data for Curry in 2024
curl "http://localhost:8080/api/playershotchart?playerId=curryst01&season=2024"
Response Format
All endpoints return paginated JSON:
{
"data": [...],
"pagination": {
"total": 450,
"page": 1,
"pageSize": 20,
"pages": 22
}
}
Rate Limiting
The API enforces a per-IP rate limit of 20 requests per minute per instance. With 3 instances behind NGINX round-robin, the effective limit is ~60 requests/minute per client.
Exceeding the limit returns:
HTTP 429
{"error": "Rate limit exceeded. Try again later."}
Getting Started
Prerequisites
- Docker & Docker Compose
- Go 1.23+ (for local development)
- A
.envfile with database credentials
Environment Variables
DB_HOST=postgres
DB_USER=your_user
DB_PASSWORD=your_password
DB_NAME=your_db
DB_PORT=5432
ADMIN_SECRET=your_admin_secret
Local Development
# Start everything (Postgres, 3 API instances, NGINX, Prometheus, Grafana)
docker-compose -f docker-compose.local.yml up --build -d
# Or use the Makefile shortcut
make up
Services will be available at:
| Service | URL |
|---|---|
| API (via NGINX) | http://localhost:8081 |
| Prometheus | http://localhost:9090 |
| Grafana | http://localhost:3001 (admin/testing) |
| API instance 1 (direct) | http://localhost:5001 |
| API instance 2 (direct) | http://localhost:5002 |
| API instance 3 (direct) | http://localhost:5003 |
Importing Data
The application provides a dual-mode entry point. To run data migrations and imports:
1. Local CLI Execution (Recommended for automated & targeted imports)
# Export environment variables from .env (e.g. remote or local database)
export $(grep -v '^#' .env | xargs)
# Run the import pipeline
go run . import-data
2. Local Docker Development Stack
docker-compose -f docker-compose.local.yml run --rm db-init
3. Production / Coolify Deployment
docker compose --profile init run --rm db-init
Ingestion Pipeline & Scraping Architecture
The data import engine (import.go & services/) features a resilient, multi-stage ingestion workflow designed to safely scrape and persist NBA statistics without rate limits or data loss:
1. Auto-Detect Missing Box Scores Engine (importMissingBoxScores)
- Dynamic Database Discovery: Automatically queries PostgreSQL for any games in the
gamestable that lack corresponding records inline_scores(WHERE game_id NOT IN (SELECT DISTINCT game_id FROM line_scores WHERE deleted_at IS NULL)). - Zero Hardcoding: Eliminates the need to manually configure date ranges or game IDs when fixing missing data across multiple historical seasons (e.g., 2008, 2013, 2017).
- 20-Game Batches: Groups detected missing games into safe 20-game chunks with 2 concurrent workers and staggered worker starts.
- Immediate Incremental Persistence: Scraped data (
line_scores,player_game_basic_stats,player_game_adv_stats,team_game_basic_stats,team_game_adv_stats) is immediately committed and upserted into PostgreSQL after every batch. - Inter-Batch Cool-Off: Enforces a 20-second pause (
\pm 25\%jitter) between batches to maintain compliant request rates against upstream sources. - Defensive URL Construction: Automatically constructs
/boxscores/{gameID}.htmlif a game record has an emptybox_score_url.
2. Date-Range Chunked Ingestion (importBoxScores)
- 20-Day Temporal Session Chunks: Large date spans (such as an entire 9-month season) are partitioned into 20-day sliding windows (~80–120 games per chunk).
- Off-Season Smart Skipping: Summer months (July–October) with 0 games are automatically identified and skipped in milliseconds without triggering scraping pauses.
3. Fault Tolerance & Safety Guarantees
- Graceful Interrupt Handling (
Ctrl+C): CapturesSIGINTandSIGTERMviacontext.Context. If interrupted, all data from completed batches is safely preserved in PostgreSQL. - Idempotent Resumption: Re-running
go run . import-dataautomatically discovers only the remaining pending games, skipping all previously completed games.
Database Verification Queries
Run the following SQL queries in PostgreSQL to verify data completeness and monitor ingestion progress:
-- 1. Check count of remaining missing games (returns 0 when fully complete)
SELECT count(*) AS remaining_missing_games
FROM games g
LEFT JOIN line_scores ls ON g.game_id = ls.game_id AND ls.deleted_at IS NULL
WHERE ls.game_id IS NULL AND g.deleted_at IS NULL;
-- 2. Inspect recently imported line scores
SELECT ls.game_id, g.date, ls.team, ls.q1, ls.q2, ls.q3, ls.q4, ls.ot1, ls.total
FROM line_scores ls
JOIN games g ON ls.game_id = g.game_id
ORDER BY ls.updated_at DESC
LIMIT 20;
-- 3. Verify total games vs total line scores
SELECT
(SELECT count(*) FROM games WHERE deleted_at IS NULL) AS total_games,
(SELECT count(DISTINCT game_id) FROM line_scores WHERE deleted_at IS NULL) AS games_with_boxscores,
(SELECT count(*) FROM line_scores WHERE deleted_at IS NULL) AS total_line_scores;
Stopping
docker compose down
# or
make down
Production Deployment
The main docker-compose.yml is configured for deployment on Coolify with an external coolify network. It expects the database to be provisioned separately (no local Postgres service).
The docker-compose.override.yml disables the local Postgres container and removes depends_on constraints, allowing API services to connect to a remote database specified in .env.
Observability
Prometheus Metrics
Exposed at /metrics on each API instance. Tracked metrics:
nba_http_requests_total— counter by method, endpoint, statusnba_http_request_duration_seconds— histogram by method, endpointnba_db_operations_total— counter by operation, entity
Grafana
Pre-provisioned dashboards visualize request rates and endpoint usage. Access at port 3001 (local) or 3000 (production).
API Key Management (Optional)
API key authentication is available but currently disabled. To create keys for future use:
# Create a key
curl -XPOST http://localhost:8080/admin/keys \
-H "X-Admin-Secret: $ADMIN_SECRET" \
-d '{"label":"my-app"}'
# → {"id":1, "apiKey":"ab12cd…"}
# Revoke a key
curl -XPOST http://localhost:8080/admin/keys/1/revoke \
-H "X-Admin-Secret: $ADMIN_SECRET"
To enforce API keys, uncomment app.Use(middleware.APIKeyAuth(db)) in main.go.
Regenerating Swagger Docs
swag init -g main.go -o docs
Running Tests
go test -v .
Load Testing
cd test
go run loadtest.go -n 100 -c 10 -url "http://localhost:8080/api/playeradvancedstats?page=1&pageSize=20" -log results.log
Tech Stack
| Component | Technology |
|---|---|
| Language | Go 1.23+ |
| Framework | Fiber v2 |
| ORM | GORM |
| Database | PostgreSQL 15 |
| Scraping | goquery |
| Load Balancer | NGINX |
| Monitoring | Prometheus + Grafana |
| Docs | Swagger (swaggo) |
| Containerization | Docker + Docker Compose |
Workflows
1. Ingesting a New Season from Scratch
- Set the target season in
import.goforimportPlayerTotalsScrape,importPlayerAdvanced,importGameSchedules, andimportMarkPlayoffGames. - Enable schedule and season totals imports in
main.go. - Run
go run . import-datato ingest schedules and season totals. - Run
importMissingBoxScoresto automatically ingest all game box scores and line scores in 20-game chunks.
2. Auto-Detecting & Filling Data Gaps
- Ensure
importMissingBoxScores(db)is active inmain.go. - Run
go run . import-data. - The engine automatically finds any missing games in PostgreSQL across all seasons, splits them into 20-game batches, and ingests them with immediate DB commits.
3. Local Development & API Testing
- Start the local stack with
make upordocker-compose -f docker-compose.local.yml up -d. - Open Swagger documentation at
http://localhost:8081/swagger/index.html. - View Grafana metrics dashboards at
http://localhost:3001(login:admin/testing).