Counting the room…
Counting the room…
Simulated Simulated results. Paper trades only, no real money. Past and simulated results do not predict future results. Not affiliated with, endorsed by or sponsored by Polymarket.
GainBuddy Picks watches the 150 most profitable traders in Polymarket's all-time leaderboard and lists the outcomes several of them are buying at the same time. A separate simulation (“paper trading”) then pretends to follow those lists with fake money, so we can see what would have happened.
We have not shown that following these traders makes money. So far the evidence points the other way (see Honest status). Nothing here is advice, and every performance figure is simulated.
For a chosen time window, we look at each market outcome (for example “Team A to win”) and count how many different tracked traders entered it: their net position grew in the window, and they bought at least a minimum amount. Outcomes are then ranked by how many traders agree, and then by total dollars spent. That count is the whole idea of consensus.
The paper-trading desk uses a strict version: at least 2 different traders, each spending at least $50, entering the same outcome within 6 hours, with market makers removed and the price between 10¢ and 85¢. The Picks board is a wider view, with smaller minimums and a wider price band, so you can see more of what is happening. Its “Consensus” label means 2+ traders (what the desk acts on, before its price and spread checks), “Strong consensus” means 3+, and anything below is “Watchlist”. The simulation only acts on the strict version.
If tracked traders buy opposite outcomes of the same market, the market is marked as disputed and the simulation skips it. Outcomes whose last price is close to 0 or 1 are hidden, because the market is effectively decided.
Consensus is not common. Replay (simulated, as of 29 October 2026): under the earlier, stricter rule (3 or more traders within 3 hours), a 45-day look back found it fired 69 times, which is why on 29 October 2026 we widened the desk to 2+ traders within 6 hours so it can produce enough trades to judge. A few traders may also be the same person using several wallets, which would overstate agreement; we cannot tell.
Some wallets buy both sides of many markets, to earn the spread or to hedge. Their purchases say nothing about which outcome they expect to win, so counting them would inflate agreement. We treat a wallet as a market maker when it has traded at least 5 markets and bought both outcomes in 25% or more of them. Their trades are ignored when counting consensus.
This is a rule of thumb. Some real directional traders get dropped by it, and some hedgers slip through.
The table below is generated from the same settings the simulation runs on, so it cannot drift from the code. The rules are fixed for the length of the test. If we change any rule, the public record restarts from a visible date; we will not quietly tune them to look better. The last change was on 29 October 2026 (signal window 3 to 6 hours, minimum traders 3 to 2). Results are counted from that date, and the replay figures under “Honest status” describe the earlier, stricter rule, not this one.
In words: when 2 or more traders agree, the desk looks at the live order book. If the price is in range, the spread is at most 4¢, and the price has not already run more than 3¢ above the traders' average entry, it “buys” $25 of paper shares at the price the order book would really have given. It then sells at 6¢ profit, at a 5¢ loss, when 2 of the signal traders sell, or after 24 hours, whichever comes first. It holds at most 10 positions and $250 at a time.
| Rule | Value | What it does |
|---|---|---|
| Signal window | 6 hours | Buying only counts if it happened within this long before now. |
| Minimum traders (consensus) | 2 traders | Distinct tracked traders, market makers excluded, who entered the same outcome. |
| Minimum spend per trader | $50 | A trader counts only if they bought at least this much of the outcome in the window. |
| Lowest entry price | 10¢ | Outcomes priced below this are skipped (long shots). |
| Highest entry price | 85¢ | Outcomes priced above this are skipped (little room left, large loss if wrong). |
| Maximum chase | 3¢ | Skip if our simulated fill would be more than this above the traders' average entry price. |
| Maximum spread | 4¢ | Skip if the gap between best buy and best sell price is wider than this. |
| Stake per trade | $25 | Paper money, flat per position. |
| Maximum open positions | 10 | No new paper trades while this many are open. |
| Maximum total exposure | $250 | Cap on paper money in open positions at once. |
| Daily loss limit | $50 | No new positions after this much realised loss in the last 24 hours. |
| Take-profit | +6¢ | Sell when the best bid is this many cents above our entry price. |
| Stop-loss | -5¢ | Sell when the best bid is this many cents below our entry price. |
| Maximum hold | 24 hours | Sell at the bid after this long if nothing else triggered. |
| Exit when traders sell | 2 traders | Sell if this many of the traders behind the signal have sold since we entered. |
| Resolved-high threshold | 97¢ | A best bid at or above this is treated as the market having been decided in our favour. |
| Resolved-low threshold | 3¢ | A best bid at or below this is treated as the market having been decided against us. |
How to read the results. A “win” is a closed paper trade that ended above its entry price. The “Agent success rate (simulated)” is simply the share of closed simulated trades that ended in profit. It always appears with its sample size, and it is greyed out and marked “early days” until there are at least 30 closed trades, because with fewer the number mostly reflects luck. Profit and loss is always called “Simulated P&L”: it is not revenue and not money anyone earned. The dashboard shows it for this week, this month and overall.
The simulation is kinder to itself than real trading would be. It leaves out, or only roughly handles:
No statistical edge has been demonstrated yet. The paper test must run for at least 8 weeks and reach 150+ closed trades before any conclusion is drawn, and until then nothing on this site should be read as evidence that the approach works.
What we know from testing on historical data (hypothetical, and using hindsight about which traders to follow):
We are staying on paper. Before anyone, including us, should consider real money, the simulation would have to clear all of these, with fees and realistic fills included:
Even then, the first real step would be a tiny, labelled, separate test. Real and simulated results would never be added together. That is a bar for taking the idea seriously, not a promise that it will be met.
No. GainBuddy Picks reports what other people have publicly traded and a simulation of following them. It is not a recommendation to buy or sell anything. Prediction markets can lose you your whole stake.
That is your decision and your risk. The traders are not obliged to be right, we see their trades late, you would pay fees and a worse price than the simulation assumes, and nothing here shows that copying them profits. Check that Polymarket is available and lawful where you live before doing anything.
Because we do not yet know whether the approach works, and testing it with pretend money costs nothing. The simulation uses real order books so its fills are not invented, but no real money is involved.
With fewer than 30 closed trades, a win rate swings wildly: 7 wins out of 10 is consistent with the true rate being anywhere from about 35% to 93%.
This site places no real orders and shows no real-money results. The operator may trade the same markets on their own account, before, alongside or against the picks. If real trading is ever shown here, it will be stated clearly and kept separate from the simulated results.
The full address is not needed to understand a pick, and these are real people's accounts. We show the public name each trader chose and a shortened address.
See also the Terms of Use and Privacy notice.