analysis.py

import json
import pandas as pd
import re
from datetime import datetime

file_path = 'temp_projects/telegram_april_export/runtime/output/channel_-1003559202666_2026-04-01_to_2026-04-30/matches.jsonl'

data = []
with open(file_path, 'r', encoding='utf-8') as f:
    for line in f:
        data.append(json.loads(line))

df = pd.DataFrame(data)
df['text'] = df['text'].fillna('').astype(str)

# Global counts from all messages
total_losses = df['text'].str.count('LOSS').sum()
total_wins = df['text'].str.count('WIN').sum()
total_gale = df['text'].str.count('WIN WITH GALE').sum()

# Identify structured entries in reports
# Pattern looks for: Asset, Time, Type (CALL/PUT), and Result
# Note: 'date_utc' is the timestamp in the JSON
pattern = re.compile(r'([A-Z]{3}/[A-Z]{3}(?:-OTC)?)\s+(\d{1,2}:\d{2})\s+(CALL|PUT).*?(LOSS|WIN(?: WITH GALE)?)', re.IGNORECASE)

records = []
for idx, row in df.iterrows():
    found = pattern.findall(row['text'])
    for m in found:
        records.append({
            'asset': m[0],
            'time': m[1],
            'type': m[2].upper(),
            'result': m[3].upper(),
            'date_utc': row['date_utc']
        })

rdf = pd.DataFrame(records)

if not rdf.empty:
    rdf['date_dt'] = pd.to_datetime(rdf['date_utc'])
    rdf['hour'] = rdf['date_dt'].dt.hour
    
    losses = rdf[rdf['result'] == 'LOSS']
    gains = rdf[rdf['result'].str.contains('WIN')]
    
    print(f"1) Total Explicit Losses (Regex): {len(losses)}")
    print(f"2) Total Gains (Regex): {len(gains)}")
    print(f"3) Loss Rate (from reports): {len(losses)/len(rdf):.2%}" if len(rdf)>0 else "3) Loss Rate: 0%")
    
    print("\n4) Most Common Losing Assets:")
    print(losses['asset'].value_counts().head(5).to_string())
    
    print("\n5) Losing Entries by Hour-of-Day:")
    print(losses['hour'].value_counts().sort_index().to_string())
    
    # Near-failure: GALE occurrences
    gale_wins = rdf[rdf['result'] == 'WIN WITH GALE']
    print(f"\n6) Near-failure (WIN WITH GALE): {len(gale_wins)}")
    
    print("\n7) Repeated Textual Pattern around Losses:")
    print(f"Direction Bias in Losses: {losses['type'].value_counts().to_dict()}")
else:
    print("No structured signals found. Falling back to raw counts.")
    print(f"1) Total 'LOSS' mentions: {total_losses}")
    print(f"2) Total 'WIN' mentions: {total_wins}")
    print(f"6) Total 'WIN WITH GALE' mentions: {total_gale}")