Methodology
This page explains, in plain terms, how MarketSignals is put together. It covers where the numbers come from, how the trader leaderboard is assembled, how the mispricing heuristic decides which markets are worth a second look, how often everything refreshes, and the limitations you should keep in mind when reading any figure on this site. The short version: everything here is built from public data and presented as an observation, never as guidance.
Data sources
All data on this site is drawn from Polymarket's public interfaces. Market metadata, such as a market's question, outcomes, current prices, and volume, comes from the public Gamma API. Position and trade activity comes from the public data API. Neither requires authentication, an account, or a private key, because both surface information that is already public, much of it settling on a public blockchain.
We do not connect to any private feed, and we hold no special access. Anything you see here, you could in principle reconstruct yourself from the same public endpoints. We pull politely, with a deliberate pause between requests so as not to strain the upstream services.
How the trader leaderboard is built
The leaderboard is an ordered view of wallet addresses that have transacted large notional amounts on Polymarket. For each wallet we read publicly visible figures: traded volume, the positions currently open, and realized profit and loss where the public data exposes it. These come from the same public on-chain and public-API records that anyone can query.
Wallets are ranked primarily by traded volume, which is simply a measure of how much activity an address has moved through the platform. A wallet appearing near the top means it has been active at scale; it does not imply skill, endorsement, or any judgment about future results. We identify wallets only by their public addresses and the display names Polymarket itself has already published. We do not attempt to tie an address to a real-world person.
How the mispricing heuristic works
A subset of markets is flagged as worth a second look. This happens when the current price diverges from a simple sanity heuristic. The heuristic is deliberately plain. It compares a market's present price against straightforward reference points such as how close the market is to its resolution date, how active trading still is, and how far the price sits into either tail of the range. When a market trades deep toward one outcome while volume remains busy and resolution is near, the heuristic notes the combination as unusual.
That is the whole of it. The heuristic does not model the underlying event, read the news, or estimate a fair price. It surfaces patterns in the public price and volume data that a reader might find interesting enough to examine more closely. Each flagged entry carries a short editorial rationale describing what the heuristic observed, written to report the observation and nothing more. A flag is an invitation to look, not a statement that the market is wrong.
Refresh cadence
The site renders from a snapshot of the public data rather than querying live on every page view. That snapshot is refreshed periodically, on a roughly daily or on-demand basis, after which the site is rebuilt so the pages reflect the latest pull. Because of this design, every figure you read is as of the most recent snapshot, not the current second. Prices and positions on Polymarket move continuously, so the live platform may already show different numbers than the ones captured here.
Limitations
- Data lag. Everything is shown as of the last snapshot. Live prices, volumes, and positions will differ from what is displayed here, sometimes substantially.
- Public data only.We see what the public APIs and the public chain expose. We have no insight into anyone's reasoning, off-platform activity, or intentions.
- A heuristic is not a model. The mispricing flag is a simple sanity check, not a valuation. It can flag markets that are priced perfectly sensibly and miss ones that are not.
- Not advice. Nothing on this site is financial advice, a tip, a recommendation, or an offer of any kind. It is editorial observation of public data.
- No guarantees. Data can be incomplete, delayed, or wrong upstream. We make no warranty as to accuracy or completeness, and past activity says nothing about future outcomes.
If you spot a figure that looks off, the editorial desk welcomes corrections. Read the data, read the market, and form your own view.