Machine Learning / Longs

The Machine Learning / Longs indicator utilizes a pattern-recognition algorithm to identify potential long entry points by comparing current price action "images" against a stored memory of historically profitable structures. This tool aims to automate the identification of recurring 8-candle formations that have preceded user-defined price increases.

Usage

The Usage section describes how the script can be used to identify bullish setups based on historical similarity.

The script operates in two primary phases: Learning and Searching. During the Learning phase, the indicator scans historical data (starting from the last bar) to find 8-candle sequences followed by a profit increase equal to or greater than the Min Profit setting. These sequences are converted into 64-bit string "images" and stored in an internal array.

In the Searching phase, the tool generates an image of the most recent 8 candles and compares it to all images stored in its memory. If the similarity between the current pattern and a stored pattern exceeds the Similarity Rate, a "Matched" label appears along with a visualization of the historical pattern.

Examples of usage include:

• Scalping: Setting a lower Period and Min Profit (e.g., 0.2%) to find quick bullish continuation patterns on 1-minute or 5-minute charts.
• Trend Following: Using a higher Similarity Rate (90%+) on higher timeframes to identify highly specific bottoming or breakout structures.
• Visual Validation: Comparing the "Current Image" displayed on the chart against the "Matched" historical image to manually verify the context of the signal.

Details

The core of the script is the transformation of price action into an 8x8 grid. The indicator calculates the highest and lowest points of an 8-candle window and divides that vertical range into 8 equal zones. Each cell in the resulting 64-cell matrix is assigned a value based on its contents:

• 0: Empty
• 1: Wick only
• 2: Red body only
• 3: Green body only
• 4: Red body and wick
• 5: Green body and wick

When the price moves up by the Min Profit percentage within the defined Period without dropping below the pattern's low, the 64-character string is saved. The similarity check calculates the percentage of matching characters between strings; if the match meets the threshold, the indicator triggers a signal. Note that this script is experimental and processes memory on the fly; if a calculation timeout occurs, toggling the script visibility will reset the execution.

Settings

General Settings

• Period: Defines the look-ahead window (in bars) to check if the minimum profit target was achieved after a pattern.
• Min Profit: The minimum percentage price increase required for a pattern to be considered "successful" and added to memory.
• Similarity Rate: The percentage threshold required for a current pattern to be considered a match with a stored historical pattern.
• Max Memory Size: The maximum number of successful patterns the script can store in its array.
• Change Bar Color?: When enabled, highlights the price bars on the chart when a pattern match is identified.

FAQ

How do I use the "Memory Full" metric?

The Memory Full metric indicates how much of your Max Memory Size has been utilized. If it reaches 100%, the script will stop learning new patterns unless the size is increased or the script is reset.

Why are there no signals on historical bars?

The script is designed to perform its search logic on the most recent bar to optimize performance. To see how it would have performed historically, you can use the TradingView Bar Replay mode.

How can I access Machine Learning / Longs?

You can get access on the LuxAlgo Library for charting platforms like TradingView, MetaTrader (MT4/MT5), and NinjaTrader for free.