Guide
Do Polymarket bots make money?
Some do, but leaderboard profit is not profit you can collect. We screened 4,125 Polymarket wallets: 590 kept a profitable record over at least 90 days, and 100 of the 587 we stored trade like machines. We then simulated copying 31 wallets at a delay. 2 passed, and for both a later check found far less cash profit than the leaderboard showed.
What we measured
We read Polymarket's public leaderboard and account data (snapshot of 2026-10-26) and ran the cheapest test first on each wallet: one request for its daily profit record. The 3,627 candidates we had not reached yet are not counted below.
| Stage | Wallets |
|---|---|
| Made at least 1,000 USD in 30 days (overall board) | 7,752 |
| Screened for a persistent record | 4,125 |
| Passed the persistence tests | 590 (14%) |
| Stored: trade like machines, human or slow-script pace | 100, 487 |
| Copy-simulated at a delay | 31 |
Source: our stored captures of Polymarket's public API, snapshot of 2026-10-26. The research continues, so later counts will be larger.
What persistent means
A wallet passes if its daily profit record is at least 90 days long, at least 55% of its days are winning days, no single day is more than 40% of its lifetime profit, and its lifetime profit is above zero. A wallet can fail several tests. Among the 4,125 screened:
- 2,748 had too few winning days.
- 1,556 owed too much of their profit to one day.
- 1,446 had too short a record.
- 913 had no positive lifetime profit.
- 21 had no usable daily record.
We call a wallet machine-like if it trades in all 24 hours of one day and either keeps a median gap under 60 seconds between orders or sustains 1,000 or more fills a day. That is timing only. It cannot tell an AI agent from a rules-based bot.
Can you copy them?
We replayed each wallet's stored trades as a late copier: a taker with 10,000 USD, entering 30 seconds, 2 minutes, 10 minutes after the wallet. Of 31 wallets simulated, 2 passed our test, which needs a positive result at the short delays that barely moves as the delay grows. None of the 27 wallets from our widened search passed. 11 lost everything at every delay because costs exceeded the wallet's gross edge, and 8 gave swings too large to read as an edge.
The pattern: persistent winners are mostly buyers of near-certain outcomes with a high win rate. Taker fees and price movement during the delay eat that thin edge.
Limits of this test. The simulation uses past fills. It penalises high-turnover wallets by design, and the 27 were the most persistent wallets plus weather-market specialists. None of the 429 human-paced wallets that passed the persistence tests was simulated, so "none passed" means none of those tested.
Why leaderboard profit misleads
- Cash is not the board figure. For the 2 wallets that passed the simulation, a later check found cash profit of about 7% and 27% of the figure reported, because some position-level profit fields carry accounting artefacts. We now take dollars from the account's daily series.
- The two sources disagree often enough to matter. 77 of 500 wallets on our shortlist carry a flag because the board figure is not confirmed by the account series.
- Wallets are not independent. One group of 12 wallets moved almost together (correlation up to 0.97). That hints at one operator or strategy, not proof, so we count a group as one lead.
What this does not tell you
It does not say who runs a wallet or whether software, an AI agent or a person trades it. Stories of small accounts that turned into six figures are claims unless the account series shows it, and we tag them that way. We have not measured arbitrage bots as a group. This is research on other people's results, not a way to make money, and not advice: see Not financial advice. For the method, read how we check. For the wider question, see can you make money with AI agents.
Questions people ask
Do Polymarket bots make money?
Some wallets that trade like bots show a profitable record: 100 of the 587 persistent wallets we stored. That is profit on the account, not proof that a copier or a new bot would keep it. Of 31 wallets we simulated copying at a delay, 2 passed our test.
Are most top Polymarket wallets bots?
Not in our sample. 487 of the 587 persistent wallets we stored traded at a human or slow-script pace, and 100 like machines. Order timing cannot tell an AI agent from a rules-based bot or a person with a script.
Can you copy trade Polymarket wallets profitably?
In our simulations, no for the 27 wallets from our widened search: none passed. 11 lost everything at every delay because costs exceeded the wallet's gross edge. The simulation uses past fills and a late taker copy, so it is a test of speed and costs, not a forecast.
How do you tell a profitable wallet from a lucky one?
We look for a long record, steady winning days and no single day carrying the total, using the account's daily profit series rather than the leaderboard figure. A wallet must have at least 90 days of record, at least 55% winning days, and no day above 40% of its lifetime profit.
Is Polymarket available everywhere?
No. Polymarket restricts users in some countries and has its own terms, so check your local rules first. We read public data only and do not trade.
Other people's results, not a promise that you will earn anything. This is research and education, not advice: see Not financial advice.
Last checked 2026-10-31. Published by Opus Vault (ABN 85 543 288 407).