VWAP Mean Reversion Backtest: The Logic, the Edge, and the Failure Modes

Repository reference: cutebacktests

Abstract

VWAP mean reversion is one of the most common intraday ideas because it has a clean market intuition. Short-horizon dislocations away from a central intraday benchmark may snap back once the move becomes stretched. The difficulty is not the intuition. The difficulty is turning that intuition into a strategy that preserves both quality and enough sample size to matter.

In this repository, the best example is c36, the option-native descendant of the c18 VWAP mean-reversion family. In Episode 8, the quality version, c36_vwap_mr_option_native_quality_v1, produced +16004 PnL on 15 trades with DSR 0.6400, while failing only trades_per_week_ok. The opportunity version reached 85 trades and +2987 PnL, but the quality shape decayed. That is a scientifically useful result because it shows a real edge and a real bottleneck at the same time.

For the VWAP family, read this with VWAP Mean Reversion Signal Quality and Density, Stock Aggregates and Indicators, and Backtesting Execution Realism.

Question

The practical question is not whether VWAP mean reversion makes sense. It is whether the edge survives once the setup is defined tightly enough to be causal and monetized honestly through options.

That is the question the c36 branch answers well. It is not a loose discretionary fade. It is a constrained intraday mean-reversion model with explicit VWAP residual z-scores, bounded VWAP slope, sigma controls, relative-volume filtering, and short holding periods. The research value of the branch is that it makes the selectivity versus density tradeoff visible.

Method: how the VWAP mean reversion backtest was structured

As described in Episode 8, c36 keeps the same core signal family while varying the degree of selectivity.

The quality version raises the entry threshold, requires stronger relative volume, narrows acceptable sigma and slope conditions, and cuts the time-in-trade budget. In plain language, it asks for cleaner dislocations and exits them quickly. The opportunity version loosens those requirements so more trades can form, even if the average setup is less extreme.

The branch is then monetized through quote-aware single-leg option execution in the 0-2DTE window. This is an important detail because it keeps the comparison focused. The underlying mean-reversion family stays conceptually stable, and the main experimental question becomes whether widening the opportunity set preserves enough quality to justify the extra trades.

Evidence and results

The repository's summary now gives a clean comparison:

• c36_vwap_mr_option_native_quality_v1: +16004 PnL, 15 trades, DSR 0.6400
• failed only trades_per_week_ok
• c36_vwap_mr_option_native_opportunity_v1: 85 trades and +2987 PnL
• the denser branch lost enough quality that it did not replace the higher-quality profile

This is one of the most useful negative-positive pairs in the repo. The quality branch says the edge is not imaginary. The opportunity branch says density cannot be purchased for free. The repo ended with a strategy that was interesting enough to keep and not strong enough to promote.

What worked

What worked was the signal logic itself. The repo did not find a dead branch here. It found a branch with a real positive profile that survived stricter evaluation than many other ideas in the same period. That is why c36 remains backup_candidate or open_paper_only in the portfolio map rather than being closed.

This also makes c36 a strong public case study. Many strategy writeups only show complete failures or obvious winners. This one shows something much closer to real research: a credible signal with a real operational weakness.

What failed

What failed was density. The best-quality version simply did not trade often enough to satisfy the repo's portfolio admission bar. The exact failed condition was trades_per_week_ok. That is not a cosmetic failure. It means the branch could make money and still fail the job it needed to do as a component of a diversified portfolio.

The opportunity version then showed why loosening the filters was not an easy repair. Trade count rose sharply, but the quality of the branch did not remain strong enough. The branch could be selective and sparse or denser and weaker, but it did not yet find the middle ground that would justify promotion.

Takeaway

The c36 result is one of the best examples in this repo of how a good backtest can still stop short of deployment. The strategy had real signal and real profits in its quality form. It also had a real density problem. That combination is precisely why VWAP mean reversion remains a live research question here rather than a closed one.

If you want the options-expression angle of this tradeoff, Intraday Mean Reversion Options: Why Signal Quality Drops When You Chase Density goes one step further. If you want the c36 decision itself, VWAP Z-Score Strategy: How We Evaluated c36 and Why It Still Was Not Promoted focuses on the admission bar. Join the research log to get the next backtest and failure report.

How the terminology applies

For VWAP Mean Reversion Backtest: The Logic, the Edge, and the Failure Modes, the backtesting workflow should treat Point-in-time contracts, Quote-aware fills, Reject reasons, Replay artifact, Cache key, and Signal timestamp as operational state rather than glossary decoration. That framing keeps the research claim causal: the strategy can only select instruments, prices, and labels that existed at the decision time.

A developer implementing this Case Study idea should persist Look-ahead leakage, Walk-forward test, Slippage model, Same-bar fill, Promotion gate, and Options data API beside the result, instead of leaving those words in a term card. It also turns attractive performance into an auditable record where fills, skips, thresholds, and replay inputs can be challenged independently.

The review artifact for VWAP Mean Reversion Backtest: The Logic, the Edge, and the Failure Modes becomes more useful when OPRA-originating data, OCC option symbol, Bid/ask spread, Midpoint, Quote/trade condition, and Quote vs trade semantics appear in the same body of evidence as the selected rows. When a result is promoted, these fields should appear in the run manifest, rather than a prose summary or final equity curve.

In production notes for this backtesting workflow, REST snapshot, WebSocket stream, Entitlement gate, Quote freshness, Timestamp semantics, and Pagination cursor define the checks that decide whether the workflow is reproducible. The result is a backtest that can be rerun, compared across threshold families, and rejected when the evidence is not strong enough.

For VWAP Mean Reversion Backtest: The Logic, the Edge, and the Failure Modes, the practical acceptance test is simple: another developer should be able to read the body, identify the exact inputs, reproduce the request sequence, and explain the accepted and rejected rows without relying on the bottom terminology grid. If a phrase appears in the page vocabulary, it should correspond to a stored field, a validation check, a replay step, or an implementation decision in the backtesting workflow.

This is also the reason the article should not measure success only by the final chart, table, or headline metric. The better standard is whether the data path, timing model, entitlement state, and evidence trail survive review. When those pieces are written directly into the body, the terminology becomes part of the workflow readers can implement.