Documentation Index
Fetch the complete documentation index at: https://condor.hummingbot.org/llms.txt
Use this file to discover all available pages before exploring further.
Backtesting simulates your strategy against historical exchange data. Use it to understand parameter sensitivity and validate that your strategy behaves as expected—not to predict exact live results.
Philosophy
Backtesting market making strategies cannot perfectly simulate reality:
-
Queue Position — Order books are FIFO (First In, First Out). You never know your position in the queue because exchanges don’t reveal participant order positions.
-
Path Dependency — If one fill differs between backtest and live, all subsequent behavior diverges. The correlation breaks with a single mismatched order.
-
Order Book Opacity — Thick order books (like USDT/BRL with $600K at the first level) mean your orders may never fill even if price touches your level.
Don’t trust absolute P&L numbers from backtests. Use backtesting as a parameter research tool, not a prediction engine.
What Backtesting Is Good For
| Use Case | Example |
|---|
| Parameter sensitivity | ”What happens if I change portfolio allocation from 2% to 10%?” |
| Behavior validation | ”Is my strategy placing sell orders only after reaching 30% inventory?” |
| Relative comparison | ”Does configuration A generate more volume than B?” |
| Strategy debugging | ”Why did the bot stop trading in this scenario?” |
Supported Executors
The backtesting engine works with controllers using these executors:
| Executor | Supported | Notes |
|---|
| Position Executor | Yes | Full support including position hold |
| DCA Executor | Yes | Dollar-cost averaging strategies |
| Grid Executor | Yes | Grid trading strategies |
| Order Executor | Coming | Not yet supported |
| Arbitrage Executor | No | Requires order book data |
| XEMM Executor | No | Requires cross-exchange order books |
Candle Resolution
One-second candles are crucial for market making backtests.
With one-minute candles, the engine can only simulate one fill per minute. Real market making often has 30+ fills per minute. One-second resolution captures the granularity needed.
Currently, only Binance Spot provides one-second candles. You may need a server in a region where Binance is accessible, or use Tailscale to route requests through an allowed region.
Running Backtests
Via Hummingbot Scripts
Hummingbot 2.14+ includes backtesting scripts in the scripts/ folder:
cd ~/hummingbot
python scripts/backtest_pmm_mister.py 0.5 # Backtest 0.5 days
- Processing time (e.g., “34 seconds for 0.5 days”)
- Interactive Plotly chart in browser
Via Condor Web Dashboard
- Navigate to Bots → Backtest
- Select a controller config
- Set the time range
- Run and view results
Interpreting Results
Chart Panes
The Plotly output has three panes:
| Pane | Content |
|---|
| Top | Candles with executor markers |
| Middle | P&L lines and volume |
| Bottom | Position size over time |
P&L Lines
| Line | Meaning |
|---|
| Yellow | Total P&L (realized + unrealized) |
| Purple | Position unrealized P&L |
| Gap between yellow and purple | Executor realized P&L |
Executor Colors
| Color | Meaning |
|---|
| White | Orders placed but not filled (early stop/refresh) |
| Green | Filled orders that hit take profit |
| Blue | Filled orders moved to position hold (market went against you) |
| Purple | Executors reducing position (selling accumulated inventory) |
Example Analysis
Portfolio Allocation: 2% vs 10%
2% allocation:
- Inventory builds gradually
- Trading continues through drawdowns
- More consistent activity
10% allocation:
- Inventory builds quickly
- May hit max position during drawdowns
- Trading stops if position is underwater and at max (profit protection enabled)
# With profit_protection=true and 10% allocation:
# If market drops while at max position → bot stops trading
# Bot resumes only when position becomes profitable again
PMM Mister Parameters
Key parameters that affect backtesting behavior:
| Parameter | Description | Example |
|---|
portfolio_allocation | % of capital placed around mid price | 0.02 (2%) |
min_base_percentage | Minimum inventory before selling | 0.30 (30%) |
max_base_percentage | Maximum inventory before buying stops | 0.70 (70%) |
max_active_executors_by_level | How many times to replace a level | 20 |
cooldown_time | Seconds between level replacements | 30 |
price_distance_tolerance | Minimum price move before replacement | 0.0002 |
activization_time | Seconds before inventory is “permanent” | 1660 |
profit_protection | Only sell if position is profitable | true |
Trading Bots vs Trading Agents
| Aspect | Trading Bots | Trading Agents |
|---|
| Logic | Deterministic, coded | LLM-driven decisions |
| Tick speed | 1 second or faster | 60+ seconds |
| Backtesting | Fully supported | Not applicable |
| Use case | HFT market making | Oversight, strategy selection |
Agents can oversee and modify bot parameters, but they operate at a higher level (every minute+) rather than every tick. This separation keeps HFT performance while adding intelligent oversight.
