Ways to follow Polymarket wallets

Four approaches, honestly described. Three of them are free, and for some jobs they are the right answer.

We are not going to pretend the alternatives are useless — for a one-off look at a single address, several of them are faster than signing up for anything. What follows is what each approach is structurally good at, so you can tell whether you need this one.

The official leaderboard

Authoritative about the venue’s own accounting and free. It ranks for a period the venue chooses, and it is not designed to tell you what someone is doing right now or to be queried as data. If you want to know who won last month, start here.

Why our figures differ from Polymarket’s

Put the same wallet side by side and the numbers will not match exactly. That is not one side miscounting: the two answer different questions, for these reasons.

• Keyed by owner, not by contract. Polymarket trades go through a proxy contract. We resolve each proxy to the wallet that controls it and score the record under that owner, so a signer and its proxy are one entity, and proxies that share one owner are counted together. Wallets we cannot prove belong together stay separate — we do not guess.
• Every figure traces to the chain. Each number is computed from on-chain fills, transfers and resolutions inside a stated block window, so anyone can recompute it. A published run is never rewritten: what you saw yesterday is still there today.
• Ranked by realised profit by default. Unrealised PnL is shown separately (you can sort by it), so the default rank does not come from a price that has not been cashed yet.
• Windows end at a fixed block. Ours close a little behind the chain tip, so each run reads a fixed snapshot that a chain reorganisation cannot change; a rolling window that runs up to the present minute will differ by whatever was traded in between.
• Exclusions are shown, not hidden. A market we cannot account for correctly is left out of PnL rather than guessed at, and each entity shows how many of its markets were left out, next to the total.

One limit runs the other way: our ledger starts at a fixed block. A position opened before it has no cost basis in our data, so it is left out of PnL, and the entity page says so when a large share of its markets is affected. For a wallet that has traded far longer than our history, Polymarket’s own lifetime figure covers more.

Free wallet trackers

Fast for a single address, and often good at rendering one wallet’s history. The recurring limitation is that they rank addresses rather than traders, and frequently show a figure without saying what window it covers or whether it will look different tomorrow. That is fine for browsing and awkward for deciding.

SQL dashboards

The most flexible option by a distance: if you can express it in SQL against on-chain tables, you can compute it, and you are not taking anyone’s word for the method. The costs are real though — you own the correctness of your own fee handling, resolution logic and proxy clustering, the query re-runs from scratch every time someone loads the page, and nothing will ever push you a notification. Several of the hardest bugs we hit building this were in exactly those three places.

Hosted copy-trading bots

The other common kind of tool is a hosted copy-trading platform: you give it trading access to your account (a connected wallet or an authorised API key), it mirrors someone else’s orders for you, and it usually charges a flat monthly fee. The convenience is real. The costs are that a third party holds trading access to your account, and that how it decides which wallets are worth following is often a single score, without the window or the sample size behind it.

PMWallets keeps the two jobs apart: choose whom to follow from the checkable data above, and place the orders yourself — or with the open-source copy-trading bot (MIT) running on your own machine, where the private key never leaves your computer and the bot starts in dry-run.

What PMWallets adds

• A window you can quote. Every figure belongs to a stated block range.
• Numbers that stop moving. A published run is never rewritten, so a track record accumulates instead of being re-derived.
• Traders, not addresses. Signer and proxy are scored as one entity.
• Uncertainty shown. A win rate arrives with the interval that says how much to trust it.
• Push, not poll. Fills go out over WebSocket within about a second of the block, with the measured distribution published.

A profile that tells you whether a strategy can be copied

A leaderboard says who made money. Before you follow someone, the question is whether you could have made it too. Each entity’s profile is computed from every one of its fills, and answers the parts of that question the data can answer:

• Maker or taker, by value. A maker’s fills are resting orders that someone else hit. Chasing them afterwards with market orders is a different trade — you pay the spread and any move since — so the result may not carry over.
• Profit without the single best market. A record made by one lucky market is not a strategy. It is in the API as pnlExTop, and it is what an address’s unlock price is banded on.
• Pace and size. How often it trades and how big its fills are, including how uneven — a very fast, very small trader is hard to follow at any latency.
• Where they trade, and how the record built up. Categories by amount traded, and a PnL curve valued at each day's close (7 days to full history).
• How much the win rate is worth. The interval and the count of decided markets behind it.

If the profile looks copyable, the full fill history is available to backtest against, and the copy-trading bot starts in dry-run: it logs every decision it would make without trading, and trades only after you switch it to live.

When you should not pay for this

If you want to check one address once, use a free tracker. If you already run your own on-chain analysis and trust your own fee and resolution handling, a dashboard will serve you better and cost less. This is worth money when you want a stable ranking of traders rather than addresses, and want to know within a second when one of them moves.

If you need the market, not the traders

PMWallets is about who is trading. If what you need is the market itself — order books, Polymarket trade prints and the Chainlink settlement feeds behind crypto Up/Down markets, tick by tick, to backtest against — that is a different dataset, available from our partner OutcomeTick.

Method in detail: how we measure. Common questions: FAQ.