Glossary

This glossary defines the most common terms used across the dashboard and API.

Trading Model

A Trading Model is a complete rule set for trading.
It combines signal generation and risk management into one system.

A Trading Model usually consists of:

• Entry Logic (how entries are generated)
• Position Manager (how stops/targets/trailing are handled)
• Regime Filter (when the Trading Model is allowed to trade)

Entry Logic

The entry engine.
It decides when to open a trade and in which direction (BUY/SELL), based on market structure and indicators.

Position Manager

The risk and exit engine.
It defines:

• initial stop distance
• trailing/exit logic
• optional take-profit logic
• max concurrent positions (if applicable)

Regime Filter

A gating layer that can block trades unless the market matches a certain condition (e.g. trend-only, volatility-only, etc.).

Regime

A high-level description of market structure.

• Trend Dominant: directional market, follow-through is likely
• Range Dominant: sideways / mean-reverting, fading extremes is often safer
• Mixed: timeframes disagree or structure is transitioning
• Unstable: regime flips frequently; edges degrade quickly

TrendMatrix

A multi-timeframe snapshot of direction and strength.

SupRes

Support/Resistance levels and related structures (e.g. multi-timeframe levels).

Fitness

The primary ranking score for Trading Models.
It combines performance and consistency over a recent backtest window.

Robustness Score

A stability estimate.
Higher values suggest the Trading Model is less sensitive to noise and fragile parameter settings.

Entry Move

Signed pip movement since the last entry, measured at request time.

Expected Value

Expected Value (EV) is the average expected result per trade.

Simple form:

EV = (Win Rate * Avg Win) - (Loss Rate * Avg Loss)

If EV is positive, the strategy has positive expectancy (before costs/slippage assumptions are violated).

Drawdown

Peak-to-trough loss during a losing phase (often expressed in pips or currency).

Simple relative formula:

Drawdown % = (Peak Equity - Current Equity) / Peak Equity * 100

Backtest

A backtest applies a trading strategy to historical data to estimate how it would have behaved in the past.

Important: a good backtest is not a guarantee for future returns. It is a stress test for ideas.

Overfitting

Overfitting means a strategy is tuned too closely to past data and learns noise instead of stable market behavior.

Typical warning signs:

• very strong in-sample performance
• large performance drop out-of-sample
• many parameters with narrow "sweet spots"

Win Rate

The share of winning trades:

Win Rate = Number of Winning Trades / Total Number of Trades

High Win Rate alone is not enough. It must be evaluated together with average win, average loss, and costs.

Risk Reward Ratio

The relationship between potential gain and potential loss of one trade.

Example: if risk is 1R and target is 2R, risk-reward ratio is 1:2.

Volatility

Volatility measures how strongly prices fluctuate.

A common proxy is the standard deviation of returns:

sigma = std(returns)

Higher volatility usually means both higher opportunity and higher risk.

Sharpe Ratio

A risk-adjusted return measure.

Basic form:

Sharpe = (Mean Portfolio Return - Risk-Free Rate) / Std(Returns)

Higher Sharpe means more return per unit of risk (volatility).

Sortino Ratio

Similar to Sharpe, but penalizes mainly downside volatility (bad volatility).

Simple form:

Sortino = (Mean Return - Target Return) / Downside Deviation

Useful when you want to separate harmful downside swings from upside variability.

Calmar Ratio

Compares return to maximum drawdown.

Simple form:

Calmar = Annualized Return / Max Drawdown

High Calmar means a strategy generated return with comparatively limited deep drawdowns.

Profit Factor

Shows how much gross profit is made per unit of gross loss.

Formula:

Profit Factor = Gross Profit / Gross Loss

Example: 1.5 means 1.5 units of profit per 1 unit of loss.

Stability Score

A composite score used in darwintIQ to summarize how stable and reliable model behavior is.

Interpretation:

• higher score: more stable profile over recent windows
• lower score: more instability, drift, or fragile behavior

Standard Deviation

A measure of dispersion around the average.

In trading it is often used on returns:

sigma = std(returns)

Higher standard deviation means outcomes vary more strongly around the mean.

Exposure

Exposure describes how much a model is actively in the market over time.

Simple idea:

Exposure = Time in market / Total observed time

Higher exposure can increase both opportunity and risk.

Return Stability

Return Stability describes how smooth or repeatable return behavior is across windows.

High return stability usually means less erratic swings in performance quality.

Drawdown Ratio

A drawdown-based risk metric (shown as DD Ratio in parts of the UI) that compares drawdown behavior to other model characteristics.

Practical meaning in darwintIQ: lower adverse drawdown pressure relative to performance is better.

Regime Fit

Regime Fit describes in which market environment a model tends to perform best (for example trend-dominant vs range-dominant phases).

It helps avoid applying the same model equally in all market states.

Trend Strength

Trend Strength measures how pronounced and directional a trend currently is.

Low trend strength often indicates choppy/ranging conditions; high trend strength indicates clearer directional structure.

Mutual Information

Mutual Information (MI) measures how much one variable tells us about another variable.

In trading, one use case is:
"How much does a signal distribution tell us about win/loss outcomes?"

Properties:

• MI = 0 means no information relationship
• MI > 0 means there is informational dependence
• higher MI means stronger informational connection

Discrete formula:

MI(X, Y) = sum_x sum_y p(x, y) * log( p(x, y) / (p(x) * p(y)) )

Beginner intuition:

• If signal and outcome are unrelated, joint probability is close to p(x) * p(y) and MI stays near 0.
• If some signal states are strongly linked to certain outcomes, MI increases.

Jensen-Shannon Divergence

Jensen-Shannon Divergence (JS) compares two probability distributions and measures how different they are.

In darwintIQ it is used in drift views to compare behavior between two windows.

Properties:

• bounded and symmetric
• 0 means distributions are identical
• larger values mean stronger distribution shift

Wasserstein Distance

Wasserstein Distance measures how much "mass" must be moved to transform one distribution into another.

Intuition: it captures how far distributions are apart, not only whether they overlap.

Kolmogorov-Smirnov Statistic

Kolmogorov-Smirnov (KS) statistic measures the largest vertical gap between two cumulative distributions.

Interpretation:

• KS = 0: distributions are the same
• larger KS: stronger distribution difference

Population Stability Index

Population Stability Index (PSI) quantifies how much a variable’s distribution has shifted between a reference and a current window.

In drift monitoring, higher PSI indicates stronger potential instability or regime change.