Trading Strategy Robustness Across Markets Agent

Trading Strategy Robustness Across Markets Agent is a premium quantitative research and validation skill designed to determine whether a trading strategy represents a transferable market edge or merely a historical fit to one symbol, one timeframe, one volatility environment, or one optimization dataset.

Instead of optimizing a strategy repeatedly on a single chart, the agent evaluates whether the underlying trading logic remains economically coherent when the environment changes.

Its core research question is:

Does the strategy's edge survive outside the chart where it was created?

The agent can analyze the same trading logic across:

NQ

MNQ

ES

MES

YM

MYM

RTY

M2K Gold Futures Micro Gold Silver Crude Oil Treasury Futures Major Forex Pairs Cross-Currency Pairs Individual Stocks ETFs Crypto Spot Markets Crypto Perpetuals Crypto Futures Indices Commodities

It can also evaluate the same strategy across multiple:

Timeframes Historical Periods Market Cycles Volatility States Trading Sessions Directional Regimes Execution-Cost Assumptions

The agent supports TradingView and Pine Script workflows as well as exported backtest reports, trade lists, optimization tables, equity curves, CSV files, walk-forward results, parameter sweeps, and natural-language strategy specifications.

The operating methodology is:

Define Strategy Thesis → Identify Development Market → Freeze Core Strategy Logic → Define Test Universe → Normalize Instrument Economics → Run Same-Family Tests → Run Related-Market Tests → Run Cross-Asset Tests Where Economically Appropriate → Run Multi-Timeframe Tests → Run Multi-Period Tests → Run Regime Tests → Run Volatility Tests → Normalize Performance Metrics → Compare Drawdowns → Compare Trade Distributions → Compare Parameter Stability → Diagnose Failed Markets → Separate Universal and Market-Specific Parameters → Build Robustness Matrix → Produce Portability Classification

The agent does not begin by independently optimizing every symbol.

That approach can make almost any strategy appear portable.

Instead, the preferred baseline is:

Same Logic Same Rule Structure Same or Normalized Parameters Different Markets

Only after baseline transferability has been measured should market-specific adaptation be considered.

The skill first identifies the economic thesis of the strategy.

Examples include:

Trend Following

A strategy may attempt to exploit persistent directional order flow.

Possible transfer markets:

Equity Indices Commodities Forex Crypto

Mean Reversion

A strategy may attempt to exploit temporary price dislocations.

Its portability may depend much more heavily on:

Liquidity Session Market Microstructure Volatility Spread

Opening Range

A strategy may depend on behavior around a defined market open.

Its strongest transfer tests may therefore be among markets with comparable session structures.

Volatility Breakout

A strategy may depend on compression followed by expansion.

Such strategies may transfer more successfully when thresholds are expressed in normalized volatility units rather than fixed price points.

The agent applies an economic plausibility filter before interpreting failed markets.

A strategy does not need to work on every asset class to be robust.

For example:

A strategy that is robust across NQ, MNQ, ES, MES, RTY, and YM but fails in EURUSD and Gold may still deserve the classification:

FAMILY-ROBUST

rather than:

FRAGILE

if the strategy's economic thesis primarily relates to US equity-index behavior.

The agent therefore distinguishes several levels of testing.

TIER 1 — SAME MARKET FAMILY

Examples:

NQ / MNQ

ES / MES

YM / MYM

RTY / M2K

GC / MGC

Purpose:

Determine whether strategy behavior survives closely related contracts.

These tests are especially useful for detecting:

Commission Distortion Slippage Sensitivity Contract-Size Effects Data Differences Execution Assumption Differences

TIER 2 — RELATED MARKETS

Examples:

NQ / ES

ES / RTY

Gold / Silver

BTC / ETH

EURUSD / GBPUSD

Purpose:

Determine whether the economic mechanism survives related but non-identical markets.

TIER 3 — CROSS-ASSET TESTING

Examples:

Trend strategy tested across:

Equity Index Futures Gold Forex Crypto

Purpose:

Evaluate conceptual portability.

TIER 4 — UNRELATED STRESS MARKETS

Purpose:

Explore the boundaries of the strategy rather than demand universal profitability.

The agent gives special attention to NQ/MNQ and ES/MES comparisons.

Because these contract families track closely related underlying markets, the strategy's signal behavior should normally remain highly similar when the same data methodology is used.

If performance differs dramatically, the agent investigates:

Tick Value Commission Slippage Liquidity Position Sizing Contract Multiplier Backtest Configuration Continuous Contract Construction Data Feed Session Settings Pine Script Logic Differences

For NQ/MNQ and ES/MES, the agent can compare:

Signal Timing Trade Count Entry Timing Exit Timing Expectancy in R Drawdown in R Profit Factor Cost-to-Gross-Edge Ratio Trade Distribution Long/Short Performance

The skill avoids raw-dollar bias.

A strategy may generate much larger dollar profit on NQ than MNQ simply because the contract multiplier is larger.

Therefore cross-market analysis should emphasize normalized metrics such as:

Return Percentage Expectancy in R Average Trade in R Drawdown in R Maximum Drawdown Percentage Return / Drawdown Profit Factor Sharpe Sortino Win Rate Payoff Ratio Exposure Trade Frequency

R-multiple normalization is especially useful when trades have predefined risk.

R can be calculated as:

Trade P&L / Initial Trade Risk

The agent can compare:

Mean R Median R R Distribution Maximum Drawdown in R Maximum Consecutive Losses Largest Winner in R Top Winner Concentration

This allows markets with different contract values to be evaluated on a more comparable basis.

The skill also uses volatility normalization.

A fixed:

20-point stop

has very different meaning on:

ES NQ Gold

BTC

A more transferable parameter may be:

1.5 ATR

The agent therefore distinguishes between:

Raw Price Parameters

and:

Normalized Parameters

Examples of normalized parameter design include:

ATR-Based Stops ATR-Based Targets Percentage Distance Volatility Percentiles Time-Based Durations Normalized Volume Thresholds

The skill builds a comprehensive parameter taxonomy.

UNIVERSAL PARAMETERS

Parameters whose values remain stable across a broad set of markets.

Example:

Reward/Risk = 2.0

when supported by multiple markets and timeframes.

NORMALIZED UNIVERSAL PARAMETERS

Parameters that become portable when expressed relative to market scale.

Example:

Stop = 1.5 ATR

rather than a fixed number of points.

MARKET-FAMILY PARAMETERS

Parameters that remain stable within a specific market family.

Example:

US equity-index futures session filter.

SYMBOL-SPECIFIC PARAMETERS

Parameters that require justified differences because of a market's structure.

Examples:

Liquidity Threshold Volume Threshold Contract-Specific Execution Setting

TIMEFRAME-SPECIFIC PARAMETERS

Parameters whose appropriate value changes because timeframe changes the effective duration or market microstructure.

EXECUTION-SPECIFIC PARAMETERS

Examples:

Commission Tick Size Spread Slippage Funding Contract Multiplier

These should not be mistaken for alpha parameters.

REGIME-SENSITIVE PARAMETERS

Parameters whose performance changes materially across volatility or directional regimes.

SUSPICIOUSLY OPTIMIZED PARAMETERS

Parameters that require a narrow, isolated value on a particular symbol without a convincing economic reason.

This is treated as an overfitting warning.

The agent can build a Parameter Portability Matrix such as:

Parameter Type NQ ES Gold

EURUSD

BTC

Stable Range Stability Recommendation

It does not merely compare the single best parameter value.

It examines parameter neighborhoods.

For example:

ATR Stop:

1.3 1.4 1.5 1.6 1.7 1.8

The goal is to find:

Broad Plateaus Stable Regions Overlapping Ranges

rather than isolated optimization spikes.

A strong result may look like:

NQ stable: 1.4–1.8

ES stable: 1.5–1.9

Gold stable: 1.4–1.7

BTC stable: 1.6–2.0

Possible universal stable range:

1.6–1.7

This is stronger evidence than discovering unrelated single-point optima.

The agent can identify a:

Universal Stable Range

for each major parameter.

It can also identify justified symbol-specific exceptions.

Example:

Universal trailing default:

2.0 ATR

BTC:

2.8 ATR

Possible justification:

Persistent volatility clustering and 24/7 market structure.

However, the BTC override is only accepted as legitimate if:

The difference is economically justified. The parameter remains stable across multiple BTC periods. Nearby values remain viable. The override improves robustness rather than merely peak historical profit.

Otherwise it is flagged as:

Potential Overfitting.

Timeframe robustness receives dedicated analysis.

The agent can compare:

1 Minute 3 Minutes 5 Minutes 15 Minutes 30 Minutes 1 Hour 4 Hours Daily

when those timeframes are economically relevant.

It distinguishes between:

Bar-Based Parameters Time-Based Parameters Volatility-Based Parameters

For example:

20 bars

represents:

20 minutes on 1-minute data

but:

20 hours on hourly data.

A more transferable definition might be:

60-minute lookback

which can then be represented differently across chart resolutions.

The agent can build a Timeframe Robustness Matrix:

Symbol 1m 5m 15m 1h 4h Daily

with statuses:

STRONG

ACCEPTABLE

WEAK

FAIL

INSUFFICIENT SAMPLE

NOT APPLICABLE

DATA REQUIRED

The skill also performs detailed regime analysis.

Core directional regimes include:

Bull Bear Sideways

Core volatility regimes include:

High Volatility Low Volatility

Additional regimes may include:

Crisis Recovery Trending Mean-Reverting Liquid Illiquid

Regime definitions should be objective and independent from strategy profitability.

Possible regime definitions include:

Long-Term Return Moving-Average Slope Benchmark Trend ATR Percentile Realized Volatility Percentile Rolling Standard Deviation Externally Supplied Regime Labels

The agent avoids circular definitions such as labeling every profitable strategy period as a favorable regime.

The Regime Robustness Matrix can contain:

Market Bull Bear Sideways High Volatility Low Volatility Overall

Each regime can report:

Trades Expectancy Profit Factor Maximum Drawdown Win Rate Average Trade Exposure

The agent interprets strategy behavior relative to the economic thesis.

For example:

A trend-following strategy may show:

Bull: Strong

Bear: Acceptable

Sideways: Weak

High Volatility: Strong

Low Volatility: Weak

This is not automatically a robustness failure.

The relevant question is whether the weakness is:

Expected Stable Bounded Understood Consistent across markets

The skill performs period robustness analysis.

Historical data can be segmented by:

Calendar Year Quarter Equal-Length Blocks Rolling 6 Months Rolling 12 Months Rolling 24 Months Pre/Post Structural Events Market Cycles

It can build a Period Matrix such as:

Symbol 2019 2020 2021 2022 2023 2024 2025

The agent looks for:

Long Deterioration One-Year Dependency Structural Breaks Recovery Performance Concentration

A strategy that generates most of its historical profit in one market during one year receives additional scrutiny.

The skill performs market-specific execution analysis.

For futures:

Continuous Contract Construction Back Adjustment Roll Methodology Tick Size Tick Value Commission Exchange Fees Slippage Session Contract Liquidity

For forex:

Spread Pip Value Rollover Session Broker Feed Currency Conversion News Slippage

For crypto:

Maker/Taker Fees Funding 24/7 Trading Exchange Differences Liquidity Weekend Behavior Perpetual vs Spot Differences

For stocks:

Survivorship Bias Delistings Corporate Actions Splits Dividends Liquidity Short-Borrow Constraints Sector Concentration

The agent avoids applying identical transaction costs across different markets without justification.

Cost robustness can be evaluated using scenarios such as:

Base Cost 1.5× Slippage 2× Slippage 3× Slippage

or custom scenarios supplied by the user.

A Cost Robustness Matrix can contain:

Market Base PF Moderate-Cost PF High-Cost PF Break-Even Cost Status

The agent can compare micro and mini contracts using:

Cost / Average Gross Trade

This is especially important for high-frequency strategies where commission represents a larger fraction of average profit on micro contracts.

The skill performs trade-distribution analysis across markets.

Possible metrics include:

Average Winner Average Loser Median Trade Largest Winner Largest Loss Top Winner Contribution Top 3 Winner Contribution Top 5 Winner Contribution Skew Maximum Consecutive Losses

The agent can perform winner-removal stress analysis:

Remove Largest Winner Remove Top 3 Winners Remove Top 5 Winners

A market whose entire historical edge disappears after removing a small number of trades receives a weaker robustness rating.

The agent separates long and short performance.

This helps identify hidden directional beta.

For example:

The strategy may appear robust across US equity indices but generate nearly all profit from long trades during bull markets.

The agent can therefore compare:

Long Expectancy Short Expectancy Long Drawdown Short Drawdown Bull Performance Bear Performance

For equity strategies, comparison against simple market exposure may be useful.

Possible benchmarks include:

Buy and Hold Simple Trend Filter Market Benchmark Simpler Strategy Variant

This helps determine whether the apparent strategy edge is mainly passive directional exposure.

Session dependence can also be tested.

Possible sessions include:

Regular Trading Hours Overnight New York London Asia

The agent can create:

Market Session Trades PF Expectancy Drawdown Status

The agent clearly separates:

Alpha Logic

from:

Execution Layer

A strategy's underlying concept may be transferable even when market-specific execution parameters differ.

For example:

Universal Logic:

Trend Pullback Momentum Confirmation ATR Stop R-Based Target

Market-Specific Execution:

Commission Spread Tick Size Slippage Trading Session

The agent extracts a universal Strategy Skeleton.

Possible components:

Trend Definition Setup Trigger Stop Principle Target Principle Session Principle Exit Principle

This enables the user to understand what is genuinely common across markets.

The agent supports several controlled adaptation types.

SCALE ADAPTATION

Example:

20-point stop →

1.5 ATR

TIME ADAPTATION

Example:

20 bars → 60-minute equivalent

SESSION ADAPTATION

Example:

Use each market's primary liquidity session.

COST ADAPTATION

Use instrument-specific transaction costs.

STRUCTURAL ADAPTATION

Allowed only when economically justified.

For every market-specific override, the agent can record:

Original Parameter Modified Parameter Reason Economic Justification Evidence Performance Impact Overfitting Risk

The skill explicitly rejects the following methodology:

For each market: Test thousands of parameter combinations. Select the best result. Declare the strategy robust.

That demonstrates fitting capacity rather than strategy portability.

Instead, the preferred hierarchy is:

Frozen Baseline → Normalized Shared Parameters → Related Markets → Market Holdouts → Controlled Adaptation

The skill can identify parameter drift.

Example:

NQ EMA:

18

ES EMA:

19

RTY EMA:

17

Gold EMA: 113

BTC EMA:

7

If the strategy requires radically unrelated values with narrow local peaks, the underlying model may not be portable.

The agent can calculate:

Median Preferred Parameter Interquartile Range Stable Overlap Region Markets Requiring Exceptions

A parameter becomes a strong universal candidate when:

Broad overlap exists. Nearby values remain viable. No market requires extreme precision. The economic interpretation is coherent.

The skill can build asset-class summaries:

Asset Class Markets Tested Pass Rate Median Expectancy Median Drawdown Verdict

It can build timeframe summaries:

Timeframe Markets Tested Pass Rate Stability Verdict

It can build regime summaries:

Regime Markets Tested Positive Weak Fail Verdict

The agent tracks evidence coverage.

Possible coverage metrics include:

Markets Tested Asset Classes Tested Timeframes Tested Historical Years Regimes Tested Total Trades OOS Trades Holdout Markets

Coverage is not treated as probability of future success.

The agent diagnoses failed markets rather than hiding them.

Possible failure types include:

Alpha Mechanism Absent Scale Mismatch Session Mismatch Cost Drag Insufficient Sample Volatility Mismatch Liquidity Mismatch Parameter Fragility Directional Bias Regime Concentration Data Quality Contract-Roll Issue Spread Funding Survivorship Bias Execution Assumption

Failed markets should not automatically be removed from the test universe.

The purpose is to map where the strategy works and where it does not.

The skill supports market holdouts.

Example:

Development: NQ

Validation: ES

Untouched Holdout: RTY and YM

The strategy is then tested once on the untouched markets.

This creates stronger evidence than repeated reoptimization.

The skill can divide markets into:

Development Markets Validation Markets Untouched Holdout Markets

It can also use timeframe holdouts and regime holdouts where appropriate.

Advanced cross-validation can include:

Leave-One-Market-Out

For each iteration:

Train or select parameters on all but one market. Test on the held-out market. Repeat.

This helps quantify transferability.

The skill can also support:

Leave-One-Period-Out Cross-Market Walk-Forward Aggregated OOS Analysis

Aggregated out-of-sample analysis should not simply sum raw dollars.

Instead, the agent can use:

Median OOS Expectancy OOS Pass Rate Median OOS Drawdown OOS Profit Factor Distribution OOS R Distribution

The agent can build the central Robustness Matrix with fields such as:

Market Timeframe Trades Net Return Profit Factor Expectancy Expectancy in R Maximum Drawdown Sharpe Sortino Bull Regime Bear Regime Sideways Regime High Volatility Low Volatility Parameter Stability Cost Robustness Overall Status

Status values include:

STRONG

ACCEPTABLE

WEAK

FAIL

INSUFFICIENT SAMPLE

NOT APPLICABLE

DATA REQUIRED

The agent interprets the matrix rather than simply displaying it.

It asks:

Is performance concentrated in one market? Does MNQ confirm NQ? Does MES confirm ES? Does the strategy survive related indices? Does normalized expectancy remain positive? Does drawdown remain economically coherent? Do nearby parameter values remain viable? Does transaction-cost sensitivity differ materially? Do regime results match the strategy thesis? Are weak markets economically understandable? Are market-specific adaptations justified? Are any exceptions suspiciously optimized?

The final portability classification is one of:

BROADLY PORTABLE

The strategy shows coherent behavior across several market families or asset classes, multiple regimes, and multiple periods using stable or normalized parameter ranges.

FAMILY-ROBUST

The strategy is robust within a well-defined market family.

Example:

US Equity Index Futures

PARTIALLY PORTABLE

The strategy transfers to several markets but requires limited, economically justified adaptations.

SYMBOL-DEPENDENT

The strategy appears primarily dependent on one or very few symbols.

REGIME-DEPENDENT

Performance depends strongly on a narrow market state.

TIMEFRAME-DEPENDENT

Performance survives only at a narrow sampling resolution.

FRAGILE / NON-ROBUST

The strategy loses economic coherence when moved away from its development environment.

The agent also reports:

Confidence: High Medium Low

Confidence depends on:

Sample Size Market Coverage Timeframe Coverage Regime Coverage Out-of-Sample Evidence Holdout-Market Evidence Parameter Stability Cost Realism Execution Realism Data Quality

An optional Robustness Scorecard can rate:

Same-Family Portability Cross-Family Portability Timeframe Robustness Period Robustness Regime Robustness Parameter Stability Cost Robustness Drawdown Stability Sample Adequacy Execution Realism

The score is not presented as a probability of future profitability.

The final deliverable can include:

Executive Verdict Strategy Thesis Development Market Development Timeframe Test Universe Market Families Asset Classes Timeframes Historical Periods Regime Definitions Volatility Definitions Execution Assumptions Cost Model Sample Adequacy Trade Frequency Normalized Performance Profit Factor Expectancy Expectancy in R Drawdown Sharpe Sortino Consecutive Losses Profit Concentration Long/Short Decomposition Bull Performance Bear Performance Sideways Performance High-Volatility Performance Low-Volatility Performance Period Stability Rolling Stability Cost Sensitivity Same-Family Portability Cross-Family Portability Cross-Asset Portability Timeframe Portability Universal Logic Universal Parameter Candidates Normalized Universal Parameters Market-Family Parameters Symbol-Specific Parameters Timeframe-Specific Parameters Execution-Specific Parameters Regime-Sensitive Parameters Suspicious Optimization Stable Parameter Ranges Parameter Drift Holdout-Market Evidence OOS Evidence Failure Map Robustness Matrix Regime Matrix Period Matrix Timeframe Matrix Cost Matrix Parameter Portability Matrix Asset-Class Summary Required Retests Portability Classification Confidence

The central commercial value of the skill is that it prevents traders from mistaking one highly optimized chart for a genuinely transferable trading edge.

It turns multi-market testing into a structured robustness audit rather than a search for whichever settings make every symbol profitable.