Convexly Enterprise / Market Quality Audit
Venue-level audit of coherence, capacity, slippage, and audit-chain reproducibility.
The Convexly Market Quality Audit (MQA) is a planned productized institutional report scoped to a single prediction-market or crypto-perpetual venue. It is the artifact a compliance team, market-maker desk, or research office can hand to its risk committee and act on. Pilot scope and timing depend on V0.1.6 validation; today this page is the spec + pilot-discussion surface. No sample report exists yet.
Scope
Four audit primitives, scoped to a single venue.
Each MQA bundles four independent measurements; pilots can subset to one or two if the venue prefers a tighter first scope. The Coherent Markets Engine and Edge Score methodology pieces underlie all four. See research / start-here for the methodology entry point.
Coherence violations
Cross-event probability axiom checks (mutually-exclusive sums, conditional Bayes, binary additivity). Per-violation detail with timestamped CLOB inputs and BH-FDR-corrected significance.
Capacity decay
Per-signal maximum-deployable notional walked against full-ladder orderbook depth, reported at $100 / $1K / $10K / $100K aggregate tiers. Distinguishes top-of-book capacity from realized capacity above book.
Slippage sensitivity
Cohort net-EV sweep at additional-slippage haircuts above the venue's published spread baseline. Median per-signal breakeven additional-slippage in basis points.
Audit-chain verification
Optional: third-party reproducibility check on the venue's internal signal pipeline. SHA-256 hash chain from raw inputs to published outputs; any divergence flagged with the input-pair that produced it.
Inputs
What we need from the venue.
- 01Public CLOB or orderbook snapshot feed — read-only access. For most prediction-market venues the public REST endpoint is sufficient; for permissioned venues a daily JSONL dump works.
- 02Event / market metadata — slugs, end-dates, resolution oracles, fee schedules, min-order sizes.
- 03Internal signal pipeline outputs — only required if the venue wants the audit-chain verification primitive. JSON or CSV of signals with input-hash references.
- 04Scoping document — which primitive(s) to run, target sample window, any contractual confidentiality requirements (NDA available).
Deliverable
What the venue receives.
- AAudit PDF on Convexly Research letterhead — narrative findings + every chart with CIs / p-values inline + scoping caveats. Designed to drop into a compliance file.
- BJSON appendix — every numeric in the PDF, machine-readable. Lets the venue's analytics team re-derive any chart or run the same audit primitive on a different sample window.
- CHash manifest — SHA-256 of every input artifact + every output. Anyone can verify bit-for-bit reproducibility against the audit chain at /research/verify.
- D30-minute walkthrough call — primary author walks the venue's risk committee through findings + answers questions + scopes follow-on work if any.
Limitations
What an MQA does NOT do.
- Does NOT guarantee the absence of coherence violations. The audit measures a sample window; the venue's live state can diverge between samples. Sample size + window length are stated explicitly.
- Does NOT certify the venue. Convexly is not a credit-rating agency or a regulator. The MQA reports measurements; the regulator or risk committee makes determinations.
- Does NOT replace internal compliance review. The audit is a third-party numerical layer; internal procedural review still applies.
- Does NOT cover counterparty credit, custody, or operational risk. Scope is market-microstructure-only.
Next step
Discuss a pilot.
The MQA is in the planned-pilot phase. We are talking with potential partners now to scope the first one. If your venue, desk, or research office wants to be part of that scoping conversation, apply below for a 30-minute call slot with the founder.