Copy or not
The copy check: Glimz replays a wallet's trades as a copier would, one minute late at the real market price, and calls Copy or Don't copy with a score and the reasons.
Most busy wallets aren't worth copying. Some win only because they buy in the first second of a launch, some flip faster than anyone can follow, and some made all their money on one lucky coin. Paste a wallet into Glimz (or ask Glimz AI) and the copy check answers the only question that matters: would you have made money copying it?
The answer is Copy or Don't copy, with a confidence (low, medium or high), a score from 0 to 100, four pillars, and the reasons. It is rule-based, deterministic and not financial advice.
How it works
- Read the history. Glimz reads up to the wallet's last 400 transactions and rebuilds every buy and sell from the balance changes in each one (Wallet trades).
- Value each trade when it happened. A trade paid in SOL, ETH or BNB is valued in USD at that coin's price at the hour of the trade, not today's. Stablecoin trades (USDC, USDT, USD1, DAI) are taken as is.
- Build positions. One position per coin: what it paid, what it got back, and what it still holds at today's price. Whatever it holds in a pool with less than $1,000 of liquidity counts as zero: it can't be sold.
- Replay it as a copier. For its most recent settled positions, Glimz buys and sells the same coins 0, 1, 3 and 5 minutes after the wallet did, at the price the market actually traded at then, sells the same share of the bag at each of its sells, and takes off 1.5% per round trip for fees, tips and slippage. Each trade is $100, so one big bet can't decide the result.
- Score and call. The results feed four pillars and fixed rules, below.
Where the prices come from
- Base, Ethereum and BNB Chain: the pool's own swap events, read from the chain. Exact executions, no third party.
- Solana: the pool's own swaps too: Glimz finds the swaps in that pool right after the wallet's trade and reads the price from the pool's vaults, so it leaves out other traders' fees and account rent. One-minute candles from GeckoTerminal fill in when the chain can't be read.
- When no price is available around a trade, that trade is left out of the replay and the check says how many were left out. It never guesses a price.
The pillars
| Pillar | What it measures |
|---|---|
| Profit | Win rate, profit factor (money won ÷ money lost) and the net result at the prices it traded at. |
| Copyable | Whether a copier one minute late still made money, how much of the edge survives the delay, and whether it snipes the first minute of a pool, flips within 3 minutes or trades like a bot. |
| Consistency | Profit without its single best trade, and whether its older and newer trades both made money. |
| Risk control | Coins that died in its hands, its worst losses and its drawdown. |
The score weighs them 35% profit, 30% copyable, 20% consistency and 15% risk control.
The rules
A position is settled once it has an outcome: sold (95% or more of the bag), mostly sold (half or more), dead, or held for 12 hours or more and judged at today's price.
| Situation | Call |
|---|---|
| The address is a contract | Don't copy, high confidence. |
| More than 40 swaps in a single hour | Don't copy, high confidence. A bot you can't copy in time. |
| Down overall over 6 or more settled positions | Don't copy, high confidence. |
| Copying it 1 minute late lost money (4 or more trades replayed) | Don't copy. It may win, but a copier doesn't. |
| Half or more of its entries within a minute of the pool opening, and no price data to show a copier keeps up | Don't copy. |
| Fewer than 6 settled positions | Don't copy, low confidence. Not enough evidence yet: track it and check again. |
| All of the bars below cleared | Copy |
| Anything else | Don't copy |
A Copy has to clear every bar, because most busy wallets aren't worth copying:
- A score of 65 or more.
- 10 or more settled positions.
- Still profitable without its single best trade.
- At least 4 trades replayed, and copying them 1 minute late made money,
- with some certainty: the average return of the copied trades minus one standard error is still above zero, so a couple of lucky trades can't carry it,
- and copying still made money without the copier's best trade.
Glimz never calls Copy without a replay. When no prices around its trades could be read, the call is Don't copy and says why.
Confidence
- A Copy is high confidence with a score of 78 or more, 15 or more settled positions and 8 or more trades replayed. Medium otherwise.
- A Don't copy is high confidence when a hard rule above decides it with enough evidence; otherwise it follows the score (under 45 high, under 57 medium, else low).
How to copy it
A Copy comes with a plan:
- Latest delay that still worked. The longest of 0, 1, 3 and 5 minutes at which copying made money.
- Stake size. A size that keeps you from moving its pools, from how deep they are.
- What to skip. When its trades in thin pools (under $50K of liquidity) lose and the rest win, it says to skip those.
- Follow its sells. Its exits are part of its edge; the replay copies them.
Track the wallet to get its next trades in your inbox within about a minute, then copy from your own wallet with Trading.
Copy radar
Glimz doesn't wait for you to find wallets. Every 10 minutes it picks a trader from a fresh launch (a pool under 7 days old: the most profitable traders on it when Glimz's CoinGecko plan has them, otherwise active ones), runs the same copy check, and ranks it on the Copy radar. Wallets you search are added too. Ask Glimz AI "which wallets should I copy?" and it reads the radar.
Grading
Every decisive wallet call (6 or more settled positions, or a hard rule like a bot) is logged on the signal ledger and graded 7 days later on what copying it one minute late would have made on the trades it opened after the call: a Copy is right if that made money, a Don't copy is right if it didn't. With fewer than 3 new trades to judge, the call is graded unknown.
Glimz's own calibration does the same walk-forward test on real wallets: make the call from trades that were opened and closed before a cutoff, then measure copying it after.
Coverage
| Chain | Copy check |
|---|---|
| Solana | Live |
| Ethereum | Live |
| Base | Basic |
| BNB Chain | Basic |
On Base and BNB Chain the wallet's history needs Glimz's Alchemy connection. Without it, the app says the wallet can't be read yet.
How to read it
- Past trades, not the future. A wallet that passed can lose the next trade. The replay shows what copying would have made, not what it will make.
- Recent history. Up to 400 transactions, often days for an active trader. The check always shows the window it covers.
- Swaps between two coins (no SOL, ETH, BNB or stablecoin on either side) aren't valued, and coins that arrived by transfer aren't scored: their cost is unknown.