Methodology
Where every number comes from.
ML Data computes market-rate reference prices from on-chain liquidity: aggregated across constituent venues, weighted by depth, filtered for manipulation. The methodology, step by step.
Why on-chain assets need on-chain reference rates
Established reference-rate methodologies are built around centralized exchange order books. That input hierarchy is well suited to major assets. It is a poor fit for the assets DeFi actually settles on.
A liquid staking token, a yield-bearing stablecoin or a wrapped BTC variant does most of its trading in on-chain pools. Its centralized listing, where one exists at all, is often thin, lagging, or absent during stress. A rate computed from that venue surfaces its error at the worst possible moment: a depeg, a liquidation cascade, a redemption run.
The methodology treats the on-chain market as the primary source of record for these assets, not a fallback.
From pool to published price
Identify the constituent venues
Each pair maps to the set of on-chain pools that constitute its market. wstETH/USD draws on 16 active pools; ETH/USD on 193. We observe reserves, executed trades plus liquidity depth across those venues on the chains of record.
Aggregate, liquidity-weighted
A single-venue price is an open door to manipulation. The published rate is aggregated across every active venue for the pair, weighted by real liquidity, so shallow pools carry proportionally little weight. The rate reflects the full market, not its shallowest corner.
Validate before publication
Anomalous prints, stale pools plus suspected manipulation are detected and excluded before the rate is published. The design intent is that a spike confined to a single pool does not propagate into the published rate.
Publish point-in-time
Every rate is pinned to a single block: one consistent, timestamped snapshot, verifiable against the chain that produced it. Published over API, scoped to the pairs you need.
Methodology FAQ
What does DEX-native pricing mean?
The rate is computed from decentralized exchange pools rather than centralized exchange listings. For assets like liquid staking tokens, yield-bearing stablecoins or wrapped BTC, those pools hold most of the tradable liquidity, which makes them the market that sets the price.
How does ML Data resist price manipulation?
Resilience is addressed at three levels: aggregation across every constituent venue so no single pool sets the rate, liquidity weighting so shallow pools do not dominate, plus outlier detection that excludes anomalous prints before publication. The design goal is that moving the published rate requires moving the underlying market itself.
Do you provide the market rate or the exchange rate for LSTs?
The market rate: what the token actually trades for. In-protocol exchange rates (like the wstETH/stETH contract ratio) can be read on-chain directly. Risk lives in the gap between the two, which is exactly what our feeds capture.
How are feeds delivered?
Over a REST API, scoped to the pairs you need. The service is operated by Matrixedlabs Technologies - FZCO. The underlying blockchain infrastructure is operated by Matrixedlink B.V. (Matrixed.Link), an affiliated company: an official Chainlink node operator running 500+ live price feeds in production.