Market Forcast

3 hari lalu

Core Concept:
It analyzes a recent segment of price data (user-defined length).

It searches backwards in historical data for similar sequences using one of several similarity methods (e.g., Pearson, Cosine, Euclidean, etc.).

Once a match is found, it projects the future movement from that past pattern onto the present, visualizing how the price may move next.

⚙️ How It Works:
1. Data Preparation:
Gathers historical price or percent change data into arrays.

Defines the "recent" segment (length = Correlation Length) as the pattern to match.

Searches backward over Bars Back To Search length to find a matching sequence.

2. Similarity Matching:
Compares each past sequence to the current one using the selected similarity method:

Cosine Similarity (default): Angle between vectors.

Pearson / Spearman / Kendall: Statistical correlation.

Euclidean: Distance between points.

MSE (Mean Squared Error): Variance difference.

3. Forecasting:
Once the most similar historical sequence is found, the price movement that followed it is used as a forecast.

Forecast is drawn as:

A dotted line (ponly) showing projected price direction.

An optional ZigZag (zonly) drawn using projected highs and lows.

Optionally, a linear regression channel around the projected line.

4. Visualization:
Draws:

Two boxes: One around the original pattern found in history, and one for the forecasted period.

Optional ZigZag projections connecting the high/low points of the forecast.

Optional Linear Regression Channel with standard deviation bounds.

Forecast line via dotted segments.

📊 User Inputs Include:
Correlation Length: Size of the recent pattern to match.

Forecast Length: How far into the future to project.

Similarity Method: Choice of similarity algorithm.

Search Range: How many past bars to search through.

Visual toggles for ZigZag, price line, and regression channel.

📈 Use Case:
This script is intended for pattern forecasting or cycle recognition, useful in markets with repetitive behavior. It’s experimental and not for blind execution, but can offer visual guidance for anticipating movement based on historical patterns.

3 hari lalu

Nota Keluaran

Logic Simplification
Feature v5 Script v6 Script
🔍 Similarity Matching Implemented using method calls (result.spearmen(), etc.) and a complex matrix structure Rewritten with simple spearmanScore() and cosineScore() functions using plain arrays
📊 Forecast Path Complex multi-branch line logic with dynamic zigzag Clean loop: forecast lines drawn from stored similarity match
🧮 Statistical Math Heavy use of matrix, method, and map Replaced with clean loops and array math (easier to debug)

✏️ Renamed & Cleaned Inputs
v5 Label v6 Label
"How Long Should The Correlating Sequences Be? (Bars Back)" "Correlation Length"
"How Long Should The Forecast Be?" "Forecast Length"
"Similarity Calculation" (same) "Similarity Calculation" (but options simplified)
"Show Projected Zig Zag" "Show ZigZag Forecast"
"Show Projected Price Path" "Show Price Forecast"
"Lin Reg σ" "Regression Std Dev"

🧰 Removed / Replaced Features
Removed from v5 Replaced or Omitted
method keyword usage Replaced with regular functions for readability
matrix, map, and custom line.updateZig() logic Removed — simplified with native Pine features
Advanced ZigZag projection logic (projectZig) Omitted — now shows only real ZigZag lines and forecast path

✨ Added in v6
Feature Description
✅ Clean forecast block Draws lines based on historical similarity, without overengineering
✅ Real-time label Shows best similarity score as a label
✅ Visual forecast boxes Clean boxes marking correlation period and forecast period
✅ Deletion of old lines Script automatically deletes excess lines (>400) for performance

🧹 Performance Optimizations
Optimization Result
📉 Removed unused arrays at runtime Less memory overhead
🔄 Efficient array slicing Avoids repeatedly slicing the same arrays unnecessarily
🧾 Simpler loop structures Reduced risk of runtime errors, improves script readability