AI Trading Bot - Configuration Guide
Understanding the Key Parameters
The "best" configuration depends entirely on your trading style. Before choosing a profile, it's
important to understand the four most impactful settings in your [Link] file.
1. timeframe_minutes: This sets the chart the bot looks at.
○ Low values (e.g., 15, 30): More trading signals, but also more market "noise."
Good for scalping strategies that aim for small, frequent profits.
○ High values (e.g., 240 for H4, 1440 for D1): Smoother trends and fewer, but often
higher-quality, signals. Good for swing trading strategies that aim to hold trades for
days.
2. label_threshold: This is the minimum profit target the AI tries to predict.
○ Low value (e.g., 0.001 for 0.1%): The AI will look for very small price moves. This
leads to more trades, but the profit per trade may be smaller.
○ High value (e.g., 0.005 for 0.5%): The AI is trained to only identify opportunities for
larger price moves. This will result in fewer, more selective trades.
3. buy_threshold / sell_threshold: This controls the AI's confidence level. The AI gives a
score from 0 (sell) to 1 (buy).
○ Wide range (e.g., 0.7 / 0.3): The bot will be very picky. It will only enter a trade if
the AI is extremely confident. This results in very few trades.
○ Narrow range (e.g., 0.55 / 0.45): The bot will trade much more frequently, acting
on even low-confidence signals.
4. lot: This is your trade size. This is the most important risk parameter. Always start with
the smallest possible lot size (e.g., 0.01) when testing a new configuration.
Configuration Profiles
Here are three sample configurations. To use one, copy the entire model and execution
sections and paste them into your [Link] file.
Profile 1: Conservative Swing Trader (Recommended Starting Point)
This profile is designed for patience and safety. It trades on a higher timeframe, waits for high-
confidence signals, and targets significant price moves.
● Style: Low frequency, high quality.
● Best For: Beginners, or traders who prefer a "set it and forget it" approach.
<!-- end list -->
"model": {
"seq_len": 48,
"horizon": 12,
"label_threshold": 0.004,
"test_split": 0.2,
"epochs": 50,
"batch": 128,
"buy_threshold": 0.65,
"sell_threshold": 0.35,
"feature_columns": [ "close", "ret", "ma5", "ma10", "ma20",
"rsi14", "macd", "atr14" ]
},
"execution": {
"lot": 0.01,
"deviation": 100,
"sleep_seconds": 900
},
Profile 2: Balanced Day Trader
This profile trades more frequently on a lower timeframe. It's a good middle-ground for those
who want to see more action without being overly aggressive.
● Style: Medium frequency, balanced risk.
● Best For: Users who want to actively monitor the bot throughout the day.
<!-- end list -->
"model": {
"seq_len": 48,
"horizon": 12,
"label_threshold": 0.002,
"test_split": 0.2,
"epochs": 50,
"batch": 128,
"buy_threshold": 0.60,
"sell_threshold": 0.40,
"feature_columns": [ "close", "ret", "ma5", "ma10", "ma20",
"rsi14", "macd", "atr14" ]
},
"execution": {
"lot": 0.01,
"deviation": 100,
"sleep_seconds": 300
},
Profile 3: Aggressive Scalper
This profile is for advanced users only. It attempts to trade very frequently on a low timeframe
with tight confidence levels, aiming to scalp small profits. This strategy carries a higher risk.
● Style: High frequency, high risk.
● Best For: Experienced users on a demo account or with a very small lot size.
<!-- end list -->
"model": {
"seq_len": 48,
"horizon": 12,
"label_threshold": 0.001,
"test_split": 0.2,
"epochs": 50,
"batch": 128,
"buy_threshold": 0.55,
"sell_threshold": 0.45,
"feature_columns": [ "close", "ret", "ma5", "ma10", "ma20",
"rsi14", "macd", "atr14" ]
},
"execution": {
"lot": 0.01,
"deviation": 100,
"sleep_seconds": 60
},
IMPORTANT: After changing your configuration, you must stop all scripts and run the
run_pipline.py again to train a new model based on your new settings. Then, restart the
live_trader.py with the new model. Always monitor the Win Rate and Total PnL on your
dashboard to judge a configuration's performance.