How it works

The rules change only between rounds, and a new round restarts every bot at 5 SOL, so results under different rules never mix. These are Round 2's rules (since Oct 8, 2026). Round 1 (Sep 30 to Oct 8) ran without the three rules marked new in Round 2 below; its results stay public on the Round 1 page.

The bots trade paper money, but nothing else about them is pretend. Every trade is priced with a real quote at the moment a real copy bot would trade, then quoted again a block later and filled at the worse of the two. Every network fee, priority fee and tip is charged.

That's exactly what a copy-trading bot would get. Three costs do the damage: the copy tax (paying more than the KOL because you buy after them), fees on every trade, and a low win rate once those are taken out.

The bot watches the wallet's balances, not which app the KOL used. KOLs trade through Axiom, GMGN, Fomo, pump.fun and Jupiter, and they all look the same from here:

• A token balance goes up while SOL or USDC goes down: that's a buy.
• A token balance goes down while SOL or USDC goes up: that's a sell.
• Tokens that arrive by transfer or airdrop, with nothing paid, are ignored.

The bot usually acts within about a second of the KOL's transaction landing, often before the full transaction is even available.

A KOL buy is skipped if any of these is true. Every skip is published with its reason.

• It's under 1 SOL (or the same in USDC).
• It's under 25% of the KOL's usual buy: the median of their buys of 1 SOL or more over the last 30 days. Small buys like that are often tests or bait.
• The pool is too thin for the bot's size: the buy would take more than 2% of the pool's liquidity.
• The token fails a safety check: its creator can still freeze tokens, it taxes transfers, it has another unsafe feature, or a test sell fails (a honeypot).
• It's $DUPEKOL itself. The bots never trade our own token.
• The bot already holds 10 positions, or buying would take its cash below 20% of its value.
• New in Round 2: the price has already run. If the bot's quote is more than 20% above what the KOL paid per token, it doesn't chase.
• New in Round 2: the fees are too big for the trade. A buy and its later sell pay fixed fees (0.003 SOL together on paper); if that's over 5% of the buy, it's skipped. In practice that means no buys under about 0.06 SOL.
• New in Round 2: the bot is K.O. A bot worth under $20 makes no new buys. It still follows the KOL's sells, so it gets out of what it holds.

• The starting point is 5% of the bot's current value (NAV).
• That's scaled from 0.5× to 2× by conviction: how big this buy is compared to the KOL's usual buy.
• Every buy is capped at 10% of NAV and at 5 SOL, whichever is smaller.
• Because size is a share of NAV, a growing bot buys bigger and a shrinking one buys smaller.

The bot sells the same share of its position that the KOL sold of theirs, counted in tokens. If the KOL sells 20% of their tokens, the bot sells 20% of its own. If the KOL sells everything, so does the bot.

If the KOL sells a token the bot doesn't hold (bought before the bot started, or skipped), there's nothing to copy.

A paused or K.O. bot makes no new buys but still follows the KOL's sells, so it never gets stuck in a coin the KOL has left.

Copying is never free, and every number on this site includes the costs:

• Copy tax: by the time the bot's trade lands, the KOL's buy (and everyone else copying them) has usually pushed the price up. The leaderboard shows this gap for every bot.
• Fees on every trade: the network fee, a priority fee and a Jito tip to land quickly, and a small refundable deposit (rent) for each new token account.
• A one-time 0.0018 SOL account the first time a wallet trades on PumpSwap. It's never refunded; Round 2 charges it on paper too.
• Price impact: bigger buys move the price more, especially in thin pools.

While the bots trade on paper, each fill is simulated the way a real one would go. The bot gets a real quote for its size, waits about one block for its transaction to land, quotes again, and takes the worse of the two, then pays the same fees.

By return: the change in each bot's value (NAV) since it started, or over the window you pick. Every bot started with 5 SOL of paper money, so the numbers compare directly.

NAV counts every open position at what selling all of it would actually return right now (a real sell quote), not the chart price. For memecoins with thin pools those can be very different.

Each bot's card also shows how the KOL did at their own prices: every coin they bought from zero and sold back to zero since tracking began, valued with the SOL they actually paid and received. It leaves out their own fees and any tokens they moved by transfer, so it's a close estimate, not their full account.

Some KOLs made money on those trades and some didn't. Their copy bots did worse in every case, and the gap is the point of this site. See Risks for why, and Transparency for what the admin can and can't do.