Logging - Polymarket Bot

Overview

The bot uses a dual logging system:

Application Logs (logger.js): Structured JSON logs for events, errors, and state changes
Tick Logger (tick-logger.js): High-frequency price data capture in JSONL format

Both use daily file rotation and write to the data/ directory.

Application Logger

Design

Levels: INFO, WARN, ERROR
Output: Both console and file (best-effort)
Format: [timestamp] [level] message {context}
File Pattern: data/logs/bot-YYYY-MM-DD.log

Implementation

import { appendFileSync, mkdirSync } from 'fs'
import { fileURLToPath } from 'url'
import { dirname, join } from 'path'

const __dirname = dirname(fileURLToPath(import.meta.url))
const LOG_DIR = join(__dirname, '..', '..', 'data', 'logs')

function getLogPath() {
  const date = new Date().toISOString().split('T')[0]
  return join(LOG_DIR, `bot-${date}.log`)
}

function write(level, message, ctx) {
  const ts = new Date().toISOString()
  const line = ctx
    ? `[${ts}] [${level}] ${message} ${JSON.stringify(ctx)}`
    : `[${ts}] [${level}] ${message}`

  // Console output
  if (level === 'ERROR') {
    process.stderr.write(line + '\n')
  } else if (level === 'WARN') {
    process.stderr.write(line + '\n')
  } else {
    process.stdout.write(line + '\n')
  }

  // File output (best-effort)
  try {
    mkdirSync(LOG_DIR, { recursive: true })
    appendFileSync(getLogPath(), line + '\n')
  } catch {
    // If file logging fails, at least console output works
  }
}

const logger = {
  info:  (message, ctx) => write('INFO',  message, ctx),
  warn:  (message, ctx) => write('WARN',  message, ctx),
  error: (message, ctx) => write('ERROR', message, ctx),
}

export default logger

Usage

import logger from './logger/logger.js'

// Simple message
logger.info('Bot started')
logger.warn('Low balance detected')
logger.error('WebSocket connection failed')

// With structured context
logger.info('Interval closed', {
  index: 42,
  result: 'UP',
  betSize: 125.00,
  profit: 12.50
})

logger.warn('High drawdown', {
  level: 'yellow',
  drawdownPct: 8.5,
  bankroll: 4575.00
})

logger.error('API request failed', {
  endpoint: 'https://api.vatic.ai/strike',
  status: 503,
  retries: 3
})

Example Output

Console

[2024-03-15T14:30:00.123Z] [INFO] Bot started
[2024-03-15T14:30:01.456Z] [INFO] Loaded 127 previous intervals
[2024-03-15T14:35:00.789Z] [INFO] Interval closed {"index":128,"result":"UP","betSize":125,"profit":12.5}
[2024-03-15T14:40:00.012Z] [WARN] High drawdown {"level":"yellow","drawdownPct":8.5,"bankroll":4575}
[2024-03-15T14:45:00.345Z] [ERROR] API request failed {"endpoint":"https://api.vatic.ai/strike","status":503,"retries":3}

File (`data/logs/bot-2024-03-15.log`)

Identical format to console, but all messages (INFO/WARN/ERROR) go to the same file.

Log Rotation

Files rotate automatically at midnight UTC:

data/logs/
├── bot-2024-03-13.log  (38 MB, 450k lines)
├── bot-2024-03-14.log  (42 MB, 485k lines)
└── bot-2024-03-15.log  (12 MB, 140k lines, current)

Retention Strategy: Use logrotate or a cron job to compress/delete old logs:

# Compress logs older than 7 days
find data/logs -name "bot-*.log" -mtime +7 -exec gzip {} \;

# Delete compressed logs older than 30 days
find data/logs -name "bot-*.log.gz" -mtime +30 -delete

Error Handling

Best-Effort File Writes: If file logging fails (e.g., disk full, permission error), the logger continues with console output only:

try {
  mkdirSync(LOG_DIR, { recursive: true })
  appendFileSync(getLogPath(), line + '\n')
} catch {
  // Silent failure — console output still works
}

Production Deployment: Monitor disk space on the log volume. A full disk will cause file writes to fail silently.

Tick Logger

Design

Purpose: Capture every BTC price tick (1 per second) for backtesting and analysis
Format: JSONL (newline-delimited JSON)
File Pattern: data/ticks-YYYY-MM-DD.jsonl
Compression: Highly compressible (repetitive structure)

Implementation

import { appendFileSync, mkdirSync } from 'fs'
import { fileURLToPath } from 'url'
import { dirname, join } from 'path'

const __dirname = dirname(fileURLToPath(import.meta.url))
const DATA_DIR = join(__dirname, '..', '..', 'data')

class TickLogger {
  constructor() {
    this._currentDate = null
    this._currentPath = null
  }

  write({ timestamp, price }) {
    const date = new Date().toISOString().split('T')[0]

    // Rotate if date changed
    if (date !== this._currentDate) {
      this._currentDate = date
      this._currentPath = join(DATA_DIR, `ticks-${date}.jsonl`)
      mkdirSync(DATA_DIR, { recursive: true })
    }

    const line = JSON.stringify({ ts: timestamp, price }) + '\n'
    appendFileSync(this._currentPath, line)
  }
}

export default new TickLogger()

Usage

import tickLogger from './logger/tick-logger.js'

// In main loop (runs every second)
setInterval(() => {
  const price = await priceSource.getPrice()
  const timestamp = Date.now()
  
  tickLogger.write({ timestamp, price })
  
  // ... rest of bot logic
}, 1000)

File Format

JSONL Structure

Each line is a standalone JSON object:

{"ts":1678901400000,"price":24150.00}
{"ts":1678901401000,"price":24150.25}
{"ts":1678901402000,"price":24149.75}
{"ts":1678901403000,"price":24150.50}
{"ts":1678901404000,"price":24151.00}

Fields:

ts (number): Unix timestamp in milliseconds
price (number): BTC/USD price

Example File (`data/ticks-2024-03-15.jsonl`)

$ wc -l data/ticks-2024-03-15.jsonl
86400 data/ticks-2024-03-15.jsonl  # 1 tick/sec × 60 sec × 60 min × 24 hr = 86,400 lines/day

$ du -h data/ticks-2024-03-15.jsonl
3.5M data/ticks-2024-03-15.jsonl  # ~40 bytes/line × 86,400 lines ≈ 3.5 MB/day

$ gzip -9 data/ticks-2024-03-15.jsonl
$ du -h data/ticks-2024-03-15.jsonl.gz
580K data/ticks-2024-03-15.jsonl.gz  # ~84% compression (highly repetitive structure)

Analysis Examples

Load Ticks in Python

import json
from datetime import datetime

ticks = []
with open('data/ticks-2024-03-15.jsonl') as f:
    for line in f:
        tick = json.loads(line)
        ticks.append({
            'timestamp': datetime.fromtimestamp(tick['ts'] / 1000),
            'price': tick['price']
        })

print(f"Loaded {len(ticks)} ticks")
print(f"Price range: ${min(t['price'] for t in ticks):.2f} - ${max(t['price'] for t in ticks):.2f}")

Compute 5-Min OHLC Bars

import pandas as pd

# Load ticks
df = pd.read_json('data/ticks-2024-03-15.jsonl', lines=True)
df['timestamp'] = pd.to_datetime(df['ts'], unit='ms')
df.set_index('timestamp', inplace=True)

# Resample to 5-min bars
ohlc = df['price'].resample('5T').ohlc()
print(ohlc.head())

# Output:
#                          open      high       low     close
# timestamp
# 2024-03-15 00:00:00  24150.00  24152.50  24148.00  24150.25
# 2024-03-15 00:05:00  24150.25  24155.00  24150.00  24153.75
# 2024-03-15 00:10:00  24153.75  24158.25  24152.50  24157.00

Search for Price Spikes

# Find all ticks where price jumped >$50 in 1 second
jq -s 'to_entries | .[] | select(.key > 0 and (.value.price - .[.key-1].price | fabs) > 50) | .value' data/ticks-2024-03-15.jsonl

# Output:
# {"ts":1678901423000,"price":24201.50}  # Spike: $24150 -> $24201.50 (+$51.50)
# {"ts":1678905789000,"price":24088.25}  # Drop: $24145.00 -> $24088.25 (-$56.75)

Compression Strategy

# Compress yesterday's tick file
DATE=$(date -d yesterday +%Y-%m-%d)
gzip -9 data/ticks-${DATE}.jsonl

# Automated daily compression (cron)
0 1 * * * gzip -9 /path/to/data/ticks-$(date -d yesterday +\%Y-\%m-\%d).jsonl

Storage Math:

Uncompressed: 3.5 MB/day × 365 days = 1.3 GB/year
Compressed (gzip): 580 KB/day × 365 days = 206 MB/year

High-frequency tick data is cheap to store long-term.

Centralized Logging Integration

Ship Logs to Elasticsearch

import logger from './logger/logger.js'
import { Client } from '@elastic/elasticsearch'

const esClient = new Client({ node: 'http://localhost:9200' })

function write(level, message, ctx) {
  const ts = new Date().toISOString()
  const line = ctx ? `[${ts}] [${level}] ${message} ${JSON.stringify(ctx)}` : `[${ts}] [${level}] ${message}`
  
  // Console + file (original behavior)
  // ... (console.log, appendFileSync)
  
  // Also ship to Elasticsearch
  esClient.index({
    index: `polymarket-bot-logs-${new Date().toISOString().split('T')[0]}`,
    document: {
      timestamp: ts,
      level,
      message,
      context: ctx || {},
      host: process.env.HOSTNAME
    }
  }).catch(err => console.error('Failed to ship log to ES:', err))
}

Use Winston for Production

Replace the minimal logger with Winston for advanced features:

import winston from 'winston'

const logger = winston.createLogger({
  level: 'info',
  format: winston.format.combine(
    winston.format.timestamp(),
    winston.format.json()
  ),
  transports: [
    new winston.transports.Console(),
    new winston.transports.File({ 
      filename: 'data/logs/bot.log',
      maxsize: 10 * 1024 * 1024,  // 10 MB
      maxFiles: 10                // Keep 10 files (100 MB total)
    }),
    new winston.transports.Http({
      host: 'logs.example.com',
      port: 8080,
      path: '/logs'
    })
  ]
})

export default logger

Log Analysis

Parse Structured Context

# Extract all interval close events
grep '"Interval closed"' data/logs/bot-2024-03-15.log | jq -r '.context | "\(.index) \(.result) \(.profit)"'

# Output:
# 128 UP 12.50
# 129 DOWN -100.00
# 130 UP 8.25

Count Errors by Type

# Count ERROR lines by message
grep '\[ERROR\]' data/logs/bot-*.log | awk -F'] ' '{print $3}' | cut -d' ' -f1-3 | sort | uniq -c | sort -rn

# Output:
#  42 API request failed
#  15 WebSocket connection failed
#   8 Insufficient balance
#   3 Rate limit exceeded

Daily Summary Report

#!/bin/bash
LOG_FILE="data/logs/bot-$(date +%Y-%m-%d).log"

echo "=== Daily Summary ==="
echo "Total lines: $(wc -l < "$LOG_FILE")"
echo "INFO:  $(grep -c '\[INFO\]' "$LOG_FILE")"
echo "WARN:  $(grep -c '\[WARN\]' "$LOG_FILE")"
echo "ERROR: $(grep -c '\[ERROR\]' "$LOG_FILE")"
echo
echo "=== Most Common Messages ==="
awk -F'\] \[' '{print $3}' "$LOG_FILE" | cut -d' ' -f1-4 | sort | uniq -c | sort -rn | head -10

Debugging with Logs

Trace Interval Lifecycle

# Follow logs in real-time
tail -f data/logs/bot-$(date +%Y-%m-%d).log

# Example output:
[2024-03-15T14:30:00.123Z] [INFO] Interval 42 started
[2024-03-15T14:30:15.456Z] [INFO] Prediction captured {"probability":0.6843,"direction":"UP"}
[2024-03-15T14:31:00.789Z] [INFO] Early prediction snapshot {"betSize":125,"evAtCapture":0.0843}
[2024-03-15T14:34:30.012Z] [INFO] Final prediction snapshot {"probability":0.7021,"direction":"UP"}
[2024-03-15T14:35:00.345Z] [INFO] Interval closed {"index":42,"result":"UP","profit":12.5}

Reproduce Production Issue

# Extract all context from a failed interval
grep 'index":42' data/logs/bot-2024-03-15.log | jq -r '.context'

# Output:
{
  "index": 42,
  "result": "DOWN",
  "earlyPrediction": { "probability": 0.6843, "direction": "UP" },
  "earlyPredictionCorrect": false,
  "betSize": 125.00,
  "profit": -125.00,
  "drawdownLevel": "yellow",
  "abstentionReason": null
}

Performance Monitoring

Log Volume

Component	Logs/Second	Daily Volume (Uncompressed)
Application	1-5	5-20 MB
Tick Logger	1	3.5 MB
Total	2-6	8.5-23.5 MB/day

Disk I/O Impact

Synchronous writes: appendFileSync() blocks the event loop (~1ms/write)
Acceptable for low frequency: Application logs are sparse (1-5/sec)
High frequency: Tick logger writes every second but payload is tiny (40 bytes)

Production Optimization: For extreme write frequency (>100 Hz), use a write buffer or switch to asynchronous appendFile() with a queue.

Interval Tracking

State machine generating log events

History Store

Persistent JSON storage for intervals

Metrics

Analyze performance from logs

Prediction Engine

Data & Tracking

Risk Management

API Reference

​Overview

​Application Logger

​Design

​Implementation

​Usage

​Example Output

​Console

​File (data/logs/bot-2024-03-15.log)

​Log Rotation

​Error Handling

​Tick Logger

​Design

​Implementation

​Usage

​File Format

​JSONL Structure

​Example File (data/ticks-2024-03-15.jsonl)

​Analysis Examples

​Load Ticks in Python

​Compute 5-Min OHLC Bars

​Search for Price Spikes

​Compression Strategy

​Centralized Logging Integration

​Ship Logs to Elasticsearch

​Use Winston for Production

​Log Analysis

​Parse Structured Context

​Count Errors by Type

​Daily Summary Report

​Debugging with Logs

​Trace Interval Lifecycle

​Reproduce Production Issue

​Performance Monitoring

​Log Volume

​Disk I/O Impact

​Related

Interval Tracking

History Store

Metrics

Build docs developers (and LLMs) love

Overview

Application Logger

Design

Implementation

Usage

Example Output

Console

File (`data/logs/bot-2024-03-15.log`)

Log Rotation

Error Handling

Tick Logger

Design

Implementation

Usage

File Format

JSONL Structure

Example File (`data/ticks-2024-03-15.jsonl`)

Analysis Examples

Load Ticks in Python

Compute 5-Min OHLC Bars

Search for Price Spikes

Compression Strategy

Centralized Logging Integration

Ship Logs to Elasticsearch

Use Winston for Production

Log Analysis

Parse Structured Context

Count Errors by Type

Daily Summary Report

Debugging with Logs

Trace Interval Lifecycle

Reproduce Production Issue

Performance Monitoring

Log Volume

Disk I/O Impact

Related