Pattern Matching

Pattern matching is the center of PyPScript.

Everything you calculate in the strategy block eventually becomes a pattern decision:

• when should the script emit UP
• when should it emit DOWN
• when should it stay HOLD

Core Form

patterns {
    condition -> SIGNAL(confidence: 0.80)
    default -> HOLD(confidence: 0.95)
}

A pattern has four parts:

1. condition
2. arrow ->
3. signal direction
4. confidence

Valid Signal Directions

• UP
• DOWN
• HOLD

Typical meaning:

• UP = bullish or long bias
• DOWN = bearish or short bias
• HOLD = no actionable edge right now

First Match Wins

Patterns are evaluated from top to bottom.

This matters a lot.

patterns {
    above(rsi_value, 50) -> UP(confidence: 0.65)
    above(rsi_value, 80) -> UP(confidence: 0.92)
    default -> HOLD(confidence: 0.95)
}

The 80 rule is never reached because anything above 80 is already above 50.

Correct ordering:

patterns {
    above(rsi_value, 80) -> UP(confidence: 0.92)
    above(rsi_value, 50) -> UP(confidence: 0.65)
    default -> HOLD(confidence: 0.95)
}

Simple Patterns

patterns {
    above(close, trend_ma) -> UP(confidence: 0.72)
    below(close, trend_ma) -> DOWN(confidence: 0.72)
    default -> HOLD(confidence: 0.95)
}

Multi-Condition Patterns

Use and, or, and not to build more precise conditions.

and

patterns {
    trend_up and momentum_up -> UP(confidence: 0.80)
    default -> HOLD(confidence: 0.95)
}

or

patterns {
    oversold or breakout -> UP(confidence: 0.72)
    default -> HOLD(confidence: 0.95)
}

not

patterns {
    trend_up and not overbought -> UP(confidence: 0.76)
    default -> HOLD(confidence: 0.95)
}

Parentheses

Use parentheses whenever grouping matters.

patterns {
    (trend_up and momentum_up) or breakout -> UP(confidence: 0.82)
    default -> HOLD(confidence: 0.95)
}

Common Pattern Functions

Most public PyPScript strategies use a small reliable set of functions:

• above(a, b)
• below(a, b)
• between(value, min, max)
• cross(a, b)

Examples:

above(close, trend_ma)
below(rsi_value, 30)
between(rsi_value, 45, 60)
cross(fast_ma, slow_ma)

Confidence

Confidence is required on every emitted signal.

UP(confidence: 0.84)
DOWN(confidence: 0.79)
HOLD(confidence: 0.95)

Practical guidance:

• 0.90+ = very selective
• 0.75 - 0.89 = strong
• 0.60 - 0.74 = moderate
• 0.50 - 0.59 = weak, usually too loose unless intentional

Do not assign 0.95 just because a condition looks fancy. Confidence should reflect selectivity and quality, not ego.

The Default Rule

Every strategy should end with:

default -> HOLD(confidence: 0.95)

That makes the no-trade state explicit.

HOLD is not a failure. It is part of the strategy.

Good strategies often spend a lot of time doing nothing.

Strong Pattern Design

Strong patterns usually have:

• one trend component
• one momentum component
• one trigger component
• an explicit hold fallback

Example:

trend_up = above(fast_ma, slow_ma)
momentum_up = above(rsi_value, 55)
entry_trigger = cross(close, fast_ma)

patterns {
    trend_up and momentum_up and entry_trigger -> UP(confidence: 0.85)
    default -> HOLD(confidence: 0.95)
}

Symmetry And Asymmetry

Some strategies should be symmetric:

patterns {
    trend_up and momentum_up and long_trigger -> UP(confidence: 0.84)
    trend_down and momentum_down and short_trigger -> DOWN(confidence: 0.84)
    default -> HOLD(confidence: 0.96)
}

Others should not.

If a market behaves differently on the long side and short side, your patterns should reflect that. Do not force fake symmetry.

Pattern Design Mistakes

1. Overlapping Rules

patterns {
    trend_up -> UP(confidence: 0.60)
    trend_up and momentum_up -> UP(confidence: 0.85)
    default -> HOLD(confidence: 0.95)
}

The stronger rule is unreachable.

2. Contradictory Conditions

patterns {
    above(rsi_value, 70) and below(rsi_value, 30) -> UP(confidence: 0.90)
    default -> HOLD(confidence: 0.95)
}

That condition can never be true.

3. No True Hold State

If everything maps to UP or DOWN, the strategy often becomes noisy and unrealistic.

4. Giant Opaque Expressions

If the pattern cannot be explained in plain English, break it into named variables first.

Good Pattern Example

#!pyp/1.0
engine: pyp-1.0

import indicators.sma
import indicators.rsi
import indicators.atr
import functions.above
import functions.below
import functions.cross

strategy "pattern_quality_example" {
    fast_ma = sma.evaluate(close, 20)
    slow_ma = sma.evaluate(close, 50)
    rsi_value = rsi.evaluate(close, 14)
    atr_value = atr.evaluate(high, low, close, 14)

    trend_up = above(fast_ma, slow_ma)
    trend_down = below(fast_ma, slow_ma)
    long_momentum = above(rsi_value, 55)
    short_momentum = below(rsi_value, 45)
    volatility_ok = above(atr_value, 0.0)
    long_trigger = cross(close, fast_ma)
    short_trigger = cross(fast_ma, close)

    patterns {
        trend_up and long_momentum and long_trigger and volatility_ok -> UP(confidence: 0.86)
        trend_down and short_momentum and short_trigger and volatility_ok -> DOWN(confidence: 0.86)
        default -> HOLD(confidence: 0.96)
    }
}

Checklist For Better Patterns

• strongest rules first
• weaker rules later
• explicit default -> HOLD
• no contradictory logic
• no hidden unreachable branches
• confidence matches selectivity

Related Pages

• Variables & Data
• Advanced Techniques
• Best Practices