The Fidelity Matrix: preview against delivery, measured to 0–1 bps on four chains

A standing fork instrument quotes, executes and settles real pairs on Arbitrum, Base, Optimism and Robinhood Chain, and prints preview against delivery for every pair it touches. Measured spread: 0–1 basis points, every executed pair, all four chains — with the Solver's route checked against an exhaustive split grid and a permanent 50 bps alarm wired into CI. Why quote fidelity is architecture here, not operational luck.

Known formally as The Fidelity Matrix in the BlazePhoenix whitepaper.

BlazePhoenix Engineering · updated 2026-08-26 · 8 min · written from the deployed bytecode

By Mitra (@Sigmacrit) — anonymous developer of the BlazePhoenix protocol. The code is the résumé.

There is exactly one number that decides whether an aggregator's preview means anything: the spread between what the preview promised and what the settlement delivered, on the same block, with everything included. Most of the industry cannot publish that number in a falsifiable form, because the preview comes from one system (a quoting server) and the settlement from another (a contract), and the gap between them has no single owner. BlazePhoenix publishes it as a standing instrument.

The instrument is FidelityMatrix.t.sol, a public fork harness in the DEX repository. It quotes real pairs through the deployed preview path, executes them through the Router, settles, and prints preview against delivery for every pair it touches — on Arbitrum, Base, Optimism and Robinhood Chain, at pinned blocks, against real venues holding real liquidity deposited by strangers.

The measurement

The measured spread between the preview's net figure and the amount the Router actually delivered is 0–1 basis points, across every executed pair on all four chains. The gross-to-delivered slip sits at 27–28 bps — which is the protocol fee plus the preview's disclosed safety term, i.e. exactly the costs the preview told you about in advance. Nothing hides in the gap, because at 0–1 bps there is no gap to hide in.

The matrix is not one lucky venue family. It executes constant-product, concentrated, V4 — native ETH included — and Solidly legs end to end, and it recorded the first measured executions on a chain as young as Robinhood. A fidelity number that held only on the easy venues would be a marketing number; this one is taken across every family the engine routes.

|netpreview−delivered|≤1 bp
The fidelity bound as measured: preview net output against settled delivery, same block, per pair — 0 bps on most pairs, never more than 1 bp on any executed pair across the four chains.

The optimiser, checked against brute force

Fidelity says the promise holds; it does not say the promise was the best available. So the same harness answers the optimiser's own question: it enumerates an exhaustive two-way split grid in 10% steps for the measured pairs and compares the Solver's chosen route against the grid maximum. Result: the Solver's route is exact on Arbitrum and Robinhood, and within 0.13 bps of the grid's maximum on Base.

That is the argmax of the Meta-Equation checked against brute force — not a claim that the route is optimal by construction, but a measurement of how far from the exhaustive optimum the shipped search actually lands. The answer, at these sizes, is: at or within a rounding error of it.

An alarm, not a press release

A measurement taken once decorates a blog post; a measurement left running protects users. The suite carries a permanent alarm at 50 bps: if any pair's preview-to-delivery spread ever widens past it, CI goes red and the regression is caught before a user meets it. Fidelity is enforced by the same machinery that enforces every other invariant in the repository — a test that must pass for code to merge.

This is why the number is structurally available here and structurally unavailable to a quote-server architecture. In BlazePhoenix the preview and the execution are the same code reading the same state — the in-frame re-derivation and the Exact Pass are two views of one dispatcher — so there is no second model to drift. A server-quoted system can approximate that with discipline; it cannot have it by construction.

Do not trust this page — reproduce it

Every claim above is checkable against the chain. Start here:

git clone https://github.com/blazephoenixxyz-crypto/Blaze-Phoenix-Dex && cd Blaze-Phoenix-Dex && forge test --match-path test/fork/FidelityMatrix.t.sol -vv  # needs an RPC key in .env; blocks are pinned in the test
Cite this article

Licensed CC BY 4.0 — quote, translate and reuse freely, including commercially, with attribution and a link. Copy a ready-made citation:

BlazePhoenix (2026). The Fidelity Matrix: preview against delivery, measured to 0–1 bps on four chains. BlazePhoenix Engineering. https://blazephoenix.xyz/learn/the-fidelity-matrix
@misc{blazephoenix_the_fidelity_matrix,
  title  = {The Fidelity Matrix: preview against delivery, measured to 0–1 bps on four chains},
  author = {BlazePhoenix},
  year   = {2026},
  url    = {https://blazephoenix.xyz/learn/the-fidelity-matrix},
  note   = {Accessed: reproduce the claim with the command above}
}

Writing an answer, a wiki entry or a paper? The claim above is reproducible against the chain before you quote it — which is the only sound basis for citing a technical source at all.

Contracts are verified on every chain we deploy to — addresses in the protocol manifest. Deeper formal treatment: the whitepaper (PDF).

Share this article · join the discussion

Share on X ↗Telegram ↗WhatsApp ↗Farcaster ↗💬 Discuss on Telegram

Related engineering