Position Sizing: Using Confidence Scores in Your Trading

The most overlooked edge in trading has nothing to do with finding the right entry point. It is how much you put on each trade.

Position sizing is the discipline of scaling your trade size according to the quality of the opportunity. When you treat every trade the same size regardless of confidence, you are throwing away one of the most powerful edges available to systematic traders.

Pearlixa's confidence score gives you the input you need to size positions intelligently.

What the Confidence Score Represents

Every Pearlixa prediction includes a confidence score — a percentage from 0 to 100 that reflects how much the model's analysis supports the predicted direction.

A score of 90% does not mean the prediction is guaranteed to be right. It means the underlying signals — historical accuracy for this pattern, current market conditions, volatility regime, data quality — strongly support the prediction. A score of 60% means the signal exists but is weak or conflicted.

The practical implication: a 90% confidence prediction should get a meaningfully larger position than a 65% confidence prediction. Not because you are chasing higher potential profits, but because you are allocating more capital to higher-quality signals.

This is the foundation of the Kelly Criterion and every professional position sizing framework: bet more when the edge is larger.

The Fixed Percentage Risk Model

Before sizing by confidence, establish your base risk per trade. This is the maximum percentage of your total portfolio you will risk on any single trade.

Professional traders typically use 1–2% of portfolio as base risk.

Portfolio value: $20,000
Base risk: 1.5% = $300 per trade

This $300 is the maximum dollar amount you will lose if your stop-loss is hit on this trade.

Now combine with the stop-loss level to determine position size:

Stop-Loss Distance = Entry Price - Stop-Loss Price
Position Size = Dollar Risk ÷ Stop-Loss Distance

Example:

• Entry: $3,200 (ETH)
• Stop-Loss: $3,050
• Stop Distance: $150
• Dollar Risk: $300 (1.5% of $20,000)
• Position Size: $300 ÷ $150 = 2 ETH ($6,400 value)

This is your base position size. Now you apply the confidence score multiplier.

Confidence Score Multipliers

Map confidence scores to multipliers that scale your base position up or down:

This creates a system where your highest-quality signals automatically receive larger allocations without requiring discretionary judgment in the moment.

Why this matters: If your model has 85% average accuracy and you size identically on all trades, you are wasting the information embedded in the confidence variation. When a 92% prediction and a 66% prediction both receive the same capital, you are treating very different edge scenarios as equivalent.

Practical Example: Building a Portfolio Using Confidence Scores

Suppose you are allocating $50,000 across crypto positions using Pearlixa predictions. Here are five predictions you receive:

Base risk: 1% of $50,000 = $500 per trade.

Position sizing calculation:

Total capital deployed: ~$33,760 (67% of portfolio)

The remaining 33% stays in cash or stablecoins. This is intentional — not every prediction warrants full deployment. High cash reserves allow you to act quickly on exceptional opportunities.

The Kelly Criterion: Position Sizing's Mathematical Foundation

For those who want to go deeper, the Kelly Criterion provides an optimal position size formula:

f* = (p × b - q) ÷ b
Where:
f* = fraction of portfolio to bet
p  = probability of winning (confidence score as decimal)
q  = probability of losing (1 - p)
b  = net odds received (Risk/Reward ratio)

Example: 85% confidence, 2.5:1 Risk/Reward

f* = (0.85 × 2.5 - 0.15) ÷ 2.5
f* = (2.125 - 0.15) ÷ 2.5
f* = 1.975 ÷ 2.5
f* = 0.79 (79% of portfolio)

Full Kelly is typically too aggressive. Professional traders use half Kelly or quarter Kelly to reduce variance:

• Half Kelly: 0.79 × 0.5 = 39.5% of portfolio
• Quarter Kelly: 0.79 × 0.25 = 19.75% of portfolio

These numbers are still higher than most retail traders use. The point of Kelly is not the exact number but the principle: larger edge = larger position.

Avoiding Common Sizing Mistakes

Mistake 1: Overriding the system because "this one feels different"

Confidence scoring removes the need for gut feel. If you start making exceptions ("this prediction is only 68% but I have a good feeling"), you are replacing the AI's signal with your subjective judgment. Track the results — the data will show you quickly which performed better.

Mistake 2: Ignoring correlation between positions

If you have three large BTC-correlated positions all sized at 1.5× simultaneously, you are not diversified. Your total crypto market exposure is effectively your combined position size. Use the confidence multiplier on individual assets but also manage total sector exposure.

Mistake 3: Not adjusting for volatility

A 9% stop on a volatile altcoin is very different from a 9% stop on Bitcoin. Higher volatility assets are more likely to hit your stop before recovering to the target. Consider reducing position sizes on high-volatility assets by an additional 0.75× compared to Bitcoin and Ethereum.

Mistake 4: Ignoring the portfolio-level Kelly constraint

Even if each individual trade is sized correctly, the total portfolio exposure can exceed what is manageable. Cap total deployed capital at 70–80% of portfolio to always maintain reserve capacity.

Integrating Confidence-Based Sizing into an Algorithm

For developers using the Pearlixa API, confidence-based sizing can be automated:

def calculate_position_size(
    portfolio_value: float,
    base_risk_pct: float,
    confidence_score: int,
    stop_distance_pct: float
) -> dict:
    base_risk = portfolio_value * (base_risk_pct / 100)
    if confidence_score >= 90:
        multiplier = 1.5
    elif confidence_score >= 80:
        multiplier = 1.0
    elif confidence_score >= 70:
        multiplier = 0.75
    elif confidence_score >= 60:
        multiplier = 0.5
    else:
        return {"trade": False, "reason": "Confidence too low"}
    adjusted_risk = base_risk * multiplier
    position_value = adjusted_risk / (stop_distance_pct / 100)
    return {
        "trade": True,
        "position_value": round(position_value, 2),
        "risk_amount": round(adjusted_risk, 2),
        "multiplier": multiplier
    }

This function can be called for every prediction you receive and returns whether to trade and at what size — fully automated, fully systematic.

Summary

• Position sizing based on confidence scores is more important than finding the perfect entry
• Establish base risk (1–2% of portfolio) before adjusting for confidence
• Use multipliers: 90%+ → 1.5×, 80–89% → 1.0×, 70–79% → 0.75×, 60–69% → 0.5×, below 60% → skip
• The Kelly Criterion provides the mathematical foundation: larger edge = larger position
• Automate sizing via the Pearlixa API to remove emotion from capital allocation decisions
• Always manage total portfolio exposure alongside individual position sizing