Backtests designed to disappoint you honestly
Most backtests are marketing. Quantivo models the costs, fills, and survivorship effects that turn a beautiful equity curve into a realistic one, and applies statistical controls that penalise strategies found by searching too hard.
What it does
Point-in-time everything
Index membership, fundamentals, and corporate actions are reconstructed as they were known on the simulated date. No lookahead, no survivorship bias.
Realistic execution
Commission, spread, slippage, and partial fills are modelled per instrument class instead of assumed away.
Walk-forward validation
Parameters are fit in-sample and evaluated out-of-sample on a rolling basis, with the degradation between the two reported prominently.
Overfit controls
Multiple-testing adjustments and deflated performance metrics account for how many variations were tried before a result was found.
How it works
Reconstruct history
The universe, prices, and corporate actions are rebuilt for the simulated date from lineage-tracked sources.
Simulate execution
Orders fill against modelled liquidity with explicit costs, including full options lifecycle handling.
Validate out-of-sample
Rolling walk-forward windows measure how much of the in-sample result survives.
Report with caveats
Results ship with sample size, trial count, and the statistical adjustments that were applied.
Where this actually stands
Quantivo is being built in the open. This is what is running today and what is being worked on next — no roadmap items dressed up as shipped features.
Shipped today
- Point-in-time universe reconstruction with sourced membership history
In progress next
- Backtest reality models for fills, costs, liquidity, and options lifecycle (QVP-157)
- Walk-forward and out-of-sample validation with overfit controls (QVP-158)
- Strategy registry with live drift monitoring and a decision kill switch (QVP-166)