Skip to main content

​
Prerequisites

Before starting, ensure you have:
1

Python 3.11+

python --version
# Should output: Python 3.11.x or higher
2

Kalshi API Credentials

Sign up at kalshi.com and get your API key.Create a .env file:
.env
KALSHI_EMAIL=[email protected]
KALSHI_PASSWORD=your-password
# or
KALSHI_API_KEY=your-api-key
3

Clone Repository

git clone https://github.com/yourusername/simple-kalshi-bot.git
cd simple-kalshi-bot

​
Installation

pip install -r requirements.txt

​
Running Evolution

​
Start from Scratch

Launch the genetic algorithm with default settings: 
python -m genetic
What happens:
1

Initialization

  • Connects to Kalshi API
  • Creates 100 random genomes (Generation 0)
  • Starts market data feed (background thread)
2

Trading Period (24 hours)

  • Each bot evaluates markets every 30 seconds
  • Bots execute paper trades based on their genes
  • Positions settle as markets close
3

Settlement Wait (up to 4 hours)

  • System waits for all positions to settle
  • Queries Kalshi API for market outcomes
4

Evolution

  • Bots ranked by ROI%
  • Top 5 survive (elitism)
  • Next 90 created via selection/crossover/mutation
  • 5 random immigrants added
  • Generation 1 begins

​
Expected Output

============================================================
  GENETIC ALGORITHM TRADING BOT - KALSHI
============================================================
Population: 100 | Bankroll: $100.0
Generation: 24h | Tick: 30s
Connected to Kalshi API: https://api.elections.kalshi.com/trade-api/v2
Market data feed started, waiting for initial data...
Feed ready: 347 markets (closing within window), 8 categories: ['climate', 'economics', 'finance', 'politics', 'science', 'sports', 'technology', 'world']

============================================================
GENERATION 0
============================================================
Created 100 bots, starting trading period...
[00:30:00] [Gen 0 | 0.5h] Active: 23/100 | Open: 12 | Trades: 23 | Settled: 0 | Est ROI: +2.1%/-0.5% | Realized: +0.0%/+0.0%
[01:00:00] [Gen 0 | 1.0h] Active: 45/100 | Open: 38 | Trades: 87 | Settled: 12 | Est ROI: +5.3%/+0.8% | Realized: +3.2%/+0.4%
...

​
Configuration

Customize evolution parameters by editing genetic/config.py:

​
Population Settings

genetic/config.py
POPULATION_SIZE = 100  # Total bots per generation
INITIAL_BANKROLL = 100.0  # USD per bot

​
Timing Settings

GENERATION_DURATION_SECONDS = 24 * 3600  # 24 hours
TICK_INTERVAL_SECONDS = 30  # Bot decision frequency
For faster testing, reduce GENERATION_DURATION_SECONDS to 3600 (1 hour). This will reduce the number of settled trades and may affect fitness quality.

​
Evolution Settings

ELITE_COUNT = 5           # Top performers that survive
TOURNAMENT_SIZE = 7       # Selection pressure
CROSSOVER_RATE = 0.7      # Crossover probability
MUTATION_RATE = 0.15      # Per-gene mutation rate
MUTATION_SIGMA = 0.10     # Mutation strength
IMMIGRATION_COUNT = 5     # Random genomes per generation

​
Resuming Evolution

Evolution automatically saves state. If interrupted: 
python -m genetic
Output: 
Resumed from generation 5, starting gen 6
The system loads the last completed generation from data/evolution/latest.json.

​
Monitoring Progress

Evolution logs are written to: 
tail -f data/evolution/evolution.log | grep INFO

​
Real-time Statistics

Every 30 minutes, you’ll see progress logs: 
[Gen 3 | 12.0h] Active: 87/100 | Open: 45 | Trades: 892 | Settled: 847 | 
Est ROI: +23.5%/+5.2% | Realized: +18.7%/+4.1%
Metrics:
  • Active: Bots that have made at least 1 trade
  • Open: Total open positions across all bots
  • Trades: Total trades executed
  • Settled: Positions that have settled
  • Est ROI: Best/Median estimated ROI (including open positions)
  • Realized ROI: Best/Median realized ROI (settled only)

​
Generation Summary

After each generation completes: 
======================================================================
GENERATION 3 RESULTS
======================================================================
Rank  Bot ID         ROI%   Trades  Settled  WinRate  Signal
----------------------------------------------------------------------
1     bot_a3f8c1    +34.2%     18       18    72.2%   mean_reversion
2     bot_9d2e45    +28.9%     24       23    65.2%   value
3     bot_7b1a92    +22.1%     15       15    60.0%   momentum
4     bot_4c6e8d    +18.5%     22       21    57.1%   contrarian
5     bot_2b9f1a    +15.3%     19       19    52.6%   mean_reversion
...

Signal distribution: {'mean_reversion': 42, 'value': 28, 'momentum': 15, 'contrarian': 10, 'price_level': 5}
Top-10 category prefs: {'politics': 8, 'sports': 6, 'economics': 5, 'finance': 4, 'world': 3}

Generation 3 stats:
  Best ROI:    34.2%
  Median ROI:  4.8%
  Mean trades: 16.3
  Active bots: 87/100

​
Viewing Results

All data is saved to data/evolution/:

​
Generation Files

ls data/evolution/
# gen_0000.json  gen_0001.json  gen_0002.json  ...
# checkpoint_gen0000.json  checkpoint_gen0001.json  ...
# hall_of_fame.json
# latest.json
# evolution.log

​
Hall of Fame

View top performers across all generations: 
import json

with open("data/evolution/hall_of_fame.json") as f:
    hof = json.load(f)

best = hof["best_ever"]
print(f"Best genome: {best['fitness_roi_pct']:.1f}% ROI")
print(f"Signal: {best['signal_type']}")
print(f"Generation: {best['generation']}")
print(f"Win rate: {best['win_rate']*100:.1f}%")
print(f"Trades: {best['settled_trades']}")

​
Load Specific Generation

import json
from genetic.genome import Genome

with open("data/evolution/gen_0005.json") as f:
    data = json.load(f)

# Get best genome
fitness = data["fitness_scores"]
best_idx = fitness.index(max(fitness))
best_genome = Genome.from_dict(data["genomes"][best_idx])

print(best_genome)

​
Stopping Evolution

Press Ctrl+C to gracefully stop: 
Interrupted by user. Shutting down...
Saved generation 3 state to data/evolution/gen_0003.json
Done. Resume anytime with: python -m genetic
The current generation will complete its evaluation before saving.

​
Testing Mode (Quick)

For rapid testing, use a shortened generation:
genetic/config.py
GENERATION_DURATION_SECONDS = 1800  # 30 minutes
TICK_INTERVAL_SECONDS = 10  # Every 10 seconds
SETTLEMENT_WAIT_HOURS = 1  # Faster settlement timeout
Short generations reduce trade count and settlement rate, which may produce lower-quality fitness signals.

​
Troubleshooting

​
No Markets Found

ERROR: No markets found. Check API credentials and connectivity.
Fix:
  1. Verify .env file has correct credentials
  2. Check internet connection
  3. Ensure Kalshi API is accessible: curl https://api.kalshi.com/trade-api/v2/exchange/status

​
API Rate Limits

ERROR: Rate limit exceeded (429)
Fix:
  • Increase TICK_INTERVAL_SECONDS to reduce API calls
  • The feed uses bulk queries and caching to minimize requests
  • Each tick makes ~2-3 API calls for 100 bots (shared data)

​
Out of Memory

MemoryError: Unable to allocate...
Fix:
  • Reduce POPULATION_SIZE to 50
  • Reduce MARKET_HISTORY_MAX_TICKS to 60
  • Close other applications

​
Bots Inactive (Low Trade Count)

Generation 0 stats:
  Active bots: 12/100
Possible causes:
  • Market filters too strict (most genomes filter out all markets)
  • Not enough markets available (check feed stats)
  • Generation duration too short for settlements
Fix:
  • Let evolution run - it will naturally select active bots
  • Check feed.get_stats() to see market count
  • Increase GENERATION_DURATION_SECONDS

​
Next Steps

Monitoring

Track evolution progress in real-time

Analysis

Analyze results and gene distributions

Architecture

Understand the system design

Genome Structure

Deep dive into the 22 genes

Build docs developers (and LLMs) love

Get started for freeTalk to us