Trading with AI: a complete and repeatable strategy

Also about Tutorials:

Analyze an Index From Scratch

Discover how to analyze the DAX using fundamentals, seasonality, AI projections, and technical signals to spot market weaknesses, plan trades, and turn data into clear, actionable strategies.

Artificial Intelligence has fundamentally changed how we process information. When interacting with Large Language Models (LLMs) like ChatGPT, the responses feel incredibly human and precise. However, few realize that the underlying logic driving these text generators can be effectively applied to financial analysis to anticipate asset movements.

This tutorial explores how the Forecaster software adapts the operating principle of LLMs to technical analysis. It details a complete trading strategy, covering everything from selecting the right asset to defining precise stop loss and take profit levels based on historical data.

From Words to Numbers: The Token Principle

To understand this trading strategy, one must first understand how ChatGPT actually works. The AI does not possess a conscious brain; instead, it operates as a sophisticated statistical engine. Programmers fed the algorithm vast amounts of text from the internet, teaching it to break down sentences into tokens—small fragments of words, similar to puzzle pieces.

Consider the following sentence structure:

“Mom went to the grocery store and bought…”

If asked to complete this sentence, a human might instinctively say “milk”.

Why? Not because it is the only correct answer, but because statistically, in billions of sentences analyzed online, the word “milk” appears most frequently in that specific context. The AI predicts the most probable next token based on patterns.

Forecaster applies this exact logic to financial charts. Instead of analyzing sentence structures, it analyzes price sequences. It treats candles, trends, and retracements as “tokens” to calculate the most statistically probable future movement.

In the case of the DAX, the aggregated net income curve is particularly significant. This line sums the net income of all companies in the index, producing a combined measure of corporate profitability. At present, this aggregated net income is declining, while index prices remain elevated near all-time highs.

The Pattern feature: Predicting the Future by Reading the Past

The core of this strategy lies in the Pattern feature. The software’s process can be broken down into three logical steps:

Pattern Recognition: The software analyzes the recent price movement of an asset (e.g., the last 3 months of Nvidia or the S&P 500).
Historical Search: It scans decades of historical data to find every instance where price action formed a pattern nearly identical to the current one.
Statistical Outcome: Once these “historical twins” are identified (e.g., 10 similar cases from the past), the software observes what happened next.

If the price increased in 8 out of those 10 historical instances, the system identifies an 80% historical probability of a bullish trend.

Decoding the Chart

The Pattern tool visualizes this data through three specific lines:

Green Line: The average of all historical bullish outcomes.
Red Line: The average of all historical bearish outcomes.
Blue Line (Most Correlated Case): The single event from the past that most closely mirrors the current market behavior (e.g., a specific movement from 2018).

Note: This is not a crystal ball. It is a statistical projection indicating that, if history repeats itself, a specific outcome is more likely than others.

Assessing Signal Quality: The Robustness Score

Not all patterns carry the same weight. To filter out noise, Forecaster uses a Robustness score (rated from 1 to 5 stars). For a signal to be considered reliable for this strategy, it should ideally achieve a 5-star rating by passing the following checks:

High Probability: Does the dominant scenario (bullish or bearish) occur in more than 70% of historical cases?
Significant Movement: Is the projected move strong enough to break out of the current price range, or is it merely sideways noise?
Correlation Alignment: Does the “Most Correlated Case” (the blue line) agree with the statistical average?
Pattern Consistency: Do at least 70% of the historical patterns display a consistent structure (e.g., a dip followed by a rally)?
Structural Similarity: Does the best match follow the same path? The system verifies if the price path of the top historical case is visually similar to the average structure of the dominant scenario.

Operational Tutorial: The Step-by-Step Strategy

With the theoretical foundation established, here is how to execute the strategy using the Quantum Screener and Seasonality tools.

Step 1: Screening for Opportunities

Manual searching is inefficient. The Quantum Screener automates the process: • Filter for a major index (e.g., S&P 500) to ensure liquidity.

Set the timeframe to 3 Months.
Look for opportunities on both the Long (buy) and Short (sell) sides to diversify exposure.

Example: The screener might highlight a stock like FactSet (Long) with 100% probability and 5- star robustness, or Western Digital (Short).

Step 2: Validation via Seasonality

A strong pattern alone is not enough. The trade must be validated by checking the asset’s seasonal trends. Open the Seasonality tab for the selected stock.

The “Traffic Light” Rule: Look for a historical timeframe (e.g., the last 5, 10, or 25 years) where the correlation between the current year’s movement and the historical average is above 70%.
If the correlation is low (e.g., 50%), the current price action is decoupling from its historical behavior. Discard the trade.

Example: If FactSet shows a 90% correlation over the last 25 years with a strong bullish trend in the coming weeks, the trade is validated.

Step 3: Setting Targets and Stop Losses

To manage risk objectively, avoid using the Projection targets, which can sometimes be extreme. Instead, rely on Seasonality averages for conservative exits.

Defining the Target (Take Profit): Select the upcoming period on the seasonality chart. If the stock has historically gained an average of +5% during this specific window over the last 25 years, set the Target at +5%.
Defining the Stop Loss: Use a fixed risk-to-reward ratio. A robust rule for this strategy is to set the Stop Loss at exactly half of the Target.

◦ Calculation: If the Target is +5%, the Stop Loss should be set at -2.5%.

◦ Short Example: If the historical data for a Short position indicates an average drop of 11%, the Target is set at 11%, and the Stop Loss is set at 5.5%.

Conclusion

This methodology removes emotional bias and subjective guessing from trading. Rather than attempting to predict macroeconomic shifts, this strategy adopts a quantitative approach based on the principle that financial markets tend to repeat historical behaviors.

Just as AI predicts the next word in a sentence based on billions of previous texts, this strategy identifies the most probable price action based on decades of market history.

Summary Checklist: