AI Trading Platform
The AI Trading Platform learns from market data to predict which trades will be successful. It uses neural networks and technical analysis. The goal is to achieve 70% prediction accuracy with a 2:1 reward-to-risk ratio on winning trades.
What Does This Platform Do?
ABXK.AI uses artificial intelligence to analyze financial markets. The system studies price charts, calculates technical indicators, and tries to predict if a trade will make money. As time goes on, it learns from its mistakes and improves its predictions.
How AI Training Works
The learning process has four main stages:
1 Data Collection
The system runs in the background and creates thousands of practice trades. Each trade includes:
- Entry price and exit price
- Stop loss and take profit levels
- All technical indicator values when the trade starts
- The final result (win or loss)
2 Backtesting
For each simulated trade, the backtester calculates:
- CVR (Chance/Risk Ratio): This is the reward divided by the risk. A CVR of 2.0 means you could gain $2 for every $1 you risk.
- Trading costs: Real fees and other costs are taken away from profits.
- Market regime: This tells us if the market is going up, down, or sideways, and how much prices are moving.
A trade is only counted as successful if it achieves CVR of 2.0 or higher after all costs.
3 Pattern Learning
The system looks at all finished trades and tries to answer one question: Which indicator combinations lead to winning trades?
For example, it might learn that:
- RSI below 30 + MACD crossover = 75% success rate for stocks
- Ichimoku cloud breakout + high ADX = 80% success rate for crypto
These patterns are saved with weights that show how reliable each indicator is.
4 Neural Network Training
Two neural networks learn from the trade data:
- LSTM Network: This network has memory cells that help it remember patterns over time. It is good at finding trends.
- Transformer Network: This network can focus on the most important information. It is better at finding complex patterns.
Both networks receive 28-35 indicator features as input. They train on thousands of trades and learn to predict which new trades will be successful.
Complete Trade Analysis Flow
When you analyze a symbol, here is what happens:
Technical Indicators
The platform calculates these indicators every time it analyzes a market:
| Indicator | What It Measures |
|---|---|
| RSI (14) | Momentum - overbought above 70, oversold below 30 |
| MACD | Trend momentum - signal line crossovers |
| SMA 20/50/200 | Short, medium, and long-term trends |
| EMA 12/26 | Exponential moving averages (react faster) |
| Bollinger Bands | Volatility - price relative to bands |
| Ichimoku Cloud | Trend, momentum, support/resistance combined |
| ADX | Trend strength - above 25 means strong trend |
| Stochastic | Momentum oscillator - %K and %D lines |
| ATR | Volatility - used for stop loss sizing |
| Volume Ratio | Current volume vs 20-day average |
Each indicator gets a learned weight based on how well it predicted winning trades in the past.
Supported Markets
The platform can analyze these markets and symbols:
| Category | Symbols |
|---|---|
| US Stocks | AAPL, MSFT, GOOGL, AMZN, NVDA, TSLA, META |
| European | SAP, ASML, MC.PA, SHEL, NESN.SW |
| Asian | 9984.T (SoftBank), 005930.KS (Samsung), 9988.HK (Alibaba) |
| Forex | EUR/USD, GBP/USD, USD/JPY |
| Crypto | BTC/USD, ETH/USD, SOL/USD |
| Commodities | GC=F (Gold), CL=F (Oil), SI=F (Silver) |
| ETFs | SPY, QQQ, GLD, IWM |
Total: 208+ symbols across global markets.
To make sure the AI does not just memorize old data, we use walk-forward testing:
- Training period: The AI learns from 24 months of past data
- Validation period: We adjust settings using 3 months of data
- Test period: We test on 3 months of new data the AI has never seen
The test period is called out-of-sample (OOS) because the AI never saw this data during training. Only these results matter for real trading.
Current test results:
- Profit Factor: 1.13 (makes money after costs)
- Sharpe Ratio: 0.66
- Win Rate: about 9% (with CVR target of 2.0)
Cost Model
Every trade includes real-world costs that you would pay when trading:
| Asset Type | Fees | Spread | Slippage | Total |
|---|---|---|---|---|
| Stocks | 0.10% | 0.05% | 0.05% | ~0.20% |
| Crypto | 0.20% | 0.10% | 0.15% | ~0.45% |
| Forex | 0.02% | 0.08% | 0.03% | ~0.13% |
These costs are taken away from every practice trade. This gives us profit estimates that are closer to real trading.
The AI Trading Platform is an internal research project used only by ABXK.AI. Visitors cannot access or use this platform publicly. We share results on social media, but these results are not financial advice.