SERVICES / QUANT
EVIDENCED BY THE ARCHIVE
Simulation harnesses that regenerate byte-identically, with costs stated explicitly.
WHO THIS IS FOR — Researchers who have been burned by backtests they could not reproduce, and want a harness whose honesty is itself tested.
THE PROBLEM IT ADDRESSES
- Backtests that drift run to run because of implicit ordering, float noise or non-seeded randomness.
- Fill and fee assumptions buried in code, so reported results cannot be independently checked.
- Overfitting discovered only after capital was committed to the configuration.
EXPECTED DELIVERABLES
- A deterministic harness over the full historical corpus with explicit cost and slippage models.
- Reproducibility tests that deliberately break determinism and assert which false results appear.
- A written statement of what the harness proves, and what it structurally cannot prove.
OUT OF SCOPE
- Walk-forward tuning campaigns presented as proof of future performance.
- Overfitting-correction methods sold as a guarantee against overfitting.
TECHNICAL APPROACH
HOW THE WORK IS DONE
Deterministic simulation over historical bars with explicit fees and slippage — reproducible run to run.
SCHEMATIC · QUANTITATIVE RESEARCH
METHODS
Deterministic replay · Cost modelling · 11.9M+ bar harness
EVIDENCE
The apex-quant-engine harness simulates 11.9M+ bars with explicit cost modelling, and "Lookahead Bias as an Engineering Failure" documents the failure modes it guards against.
LIMITATIONS
SCOPE & LIMITS
- Reproducibility proves the run is honest, not that the past predicts the future.
- Multiple-comparison effects are documented, not eliminated — the harness makes them visible.
- Cost models approximate market impact; they cannot reconstruct liquidity that no longer exists.
WORKING PROCESS
HOW THIS ENGAGEMENT RUNS
- 01FIX THE HARNESS
- 02MODEL COSTS AND SLIPPAGE
- 03BREAK DETERMINISM ON PURPOSE
- 04STATE WHAT THE HARNESS CANNOT PROVE
Each stage produces something checkable. The full process, with what every stage produces across all work, lives on the process page. Every engagement starts the same way: describe the problem, the data, and the evidence that would count as success.