Execution realismDaily Backtests: Put the Signal Before the Fill
A daily bar can hide the most important assumption in the simulation: when the order could actually have been sent.
Daily strategiesExecutionSignal timing
4 min read
Research notes · 18 field notes
The machinery behind a defensible backtest.
Validraft notes on evidence quality: from point-in-time data and execution realism to statistical gates, immutable reporting, and post-validation monitoring.
Execution realism
Statistical validationTrade Count and the Illusion of Independent Evidence
A long trade log may contain repeated exposure to the same days, the same market shock, and the same underlying bet.
Sample sizeDependenceUncertainty
4 min read
MonitoringPost-Validation Monitoring: When the Live World Moves
A validation verdict is dated evidence, not a permanent license to trust a strategy.
MonitoringDriftGovernance
6 min read
Research governanceA Research Result Is Not Yet a Validated Strategy
Evidence about a signal and evidence about a tradable implementation are different deliverables.
ResearchValidationGovernance
6 min read
Alternative dataAlternative Data: More Features, More Ways to Leak the Future
Novel datasets can add information, but their publication history is often harder to reconstruct than their values.
Alternative dataLeakagePoint-in-time
6 min read
Futures dataContinuous Futures: The Roll Is Part of the Strategy
A smooth historical series can hide a chain of contract choices, adjustments, and executable roll decisions.
FuturesData engineeringExecution
6 min read
Data infrastructureMarket Data Lineage: Every Bar Needs a Provenance Trail
Clean data is not enough; a reviewer must be able to reconstruct where each transformation came from.
Data lineageReproducibilityInfrastructure
6 min read
Research infrastructureReproducible Backtests: From Notebook to Evidence Package
A result is reproducible only when another run can recover the same inputs, decisions, and outputs.
ReproducibilityArtifactsGovernance
6 min read
BenchmarkingChoose a Benchmark That Can Falsify the Claim
A benchmark should challenge the proposed edge, not provide an easy line to beat.
BenchmarkAttributionValidation
5 min read
Statistical validationDeflated Sharpe Ratio: Paying the Multiple-Testing Tax
The best Sharpe selected from many attempts is not comparable with a Sharpe produced by one preregistered test.
Deflated SharpeMultiple testingStatistics
6 min read
Robustness testingMonte Carlo for Strategies: Rearranging Risk, Not Inventing Returns
Simulation is most useful when it perturbs a declared mechanism and least useful when it fabricates precision.
Monte CarloDrawdownRobustness
6 min read
Stress testingWhat Regime Stress Testing Should Actually Test
A stress test should challenge the economic dependency of a strategy, not merely divide history into colorful labels.
RegimesStress testingRisk
6 min read
Out-of-sample validationWalk-Forward Validation Without the Ritual
Repeated train-and-test windows are useful only when they reproduce the decisions the strategy would really make.
Walk-forwardOut-of-sampleValidation
6 min read
Execution modelingSlippage Is a Distribution, Not a Constant
A fixed basis-point haircut can be a useful baseline, but it cannot represent every order, regime, and level of urgency.
SlippageMarket microstructureExecution
6 min read
Execution realismTransaction Costs: Where Paper Alpha Goes to Die
Gross performance describes a signal; net performance determines whether the implementation still has an economic claim.
Transaction costsTurnoverCapacity
6 min read
Universe integritySurvivorship Bias: The Companies Your Dataset Forgot
Testing today's survivors in yesterday's market gives the strategy information that an investor never had.
Survivorship biasEquitiesData quality
5 min read
Point-in-time dataPoint-in-Time Data: The Difference Between a Test and a Time Machine
A historical value is usable only if the simulation knows when that value became available.
Point-in-timeData leakageFundamentals
6 min read
Backtest overfittingHow to Tell If Your Backtest Is Overfit
A practical checklist for spotting curve-fitting before a beautiful equity curve becomes an expensive mistake.
BacktestingOverfittingValidation
7 min read