Custom search · the whole pump.fun trading history

19 million wallets.
For pump.fun traders, beginner to advanced.

A search engine over every Solana wallet that has ever traded on pump.fun. You set the criteria, on any of 79 columns. The base hands you everything it has.

For instance: win rate > 80% > 50 trades entry under $10k MCap active within 48 h gain per trade > +30% …and 74 more columns

And behind every line: we replay their trades with your entry simulated at three latencies — same block, +1, +2. Fees and slippage counted both ways.

Open the base See the method

Telegram sign-in · 0.5 SOL a month · no query limits

108,830,081positions replayed last night
800,368wallets scored
79columns to pick from
GTT

The tool, the moment you sign in

enlarge Base Wallets: the whole table, 19,342,830 rows
No filter by defaultthe base as it is — you add the bounds you want
18 of 79 columns shownyou choose which ones, and the order
Export up to 50,000 rowsyour selection leaves with you, as a CSV

Figures

Reading taken 18 September 2026 65 new rows a second, right now
809,495,597
trades recorded straight off the Solana stream. Not a sample, not a selection: the stream.
19,342,830
wallets in the table — bots, launch factories and one-trade addresses included.
4,419,746
tokens tracked from creation, migration included.

The tokens drifting behind this page are real — 89 of them, each labelled with the multiple it reached.

What every other leaderboard gets wrong

They rank the wallet. We simulate you.

GMGN, the scanners, the "top trader" boards all show the same thing: what the wallet made. Nobody can have that number — nobody was in its block.

The worst case is the launch bundle. A wallet buying inside it posts numbers no copier can reach: by the time you are in, the whole bundle has gone through and the price with it. We do not need a blacklist for them — simulating your entry after the fact settles it. Those wallets show a median ×1.00 at their own price, and ×0.86 once copied at block +2. The first-block buys column names them: 2.5% of judged wallets.

The same trade, depending on when you get in
What their board shows
+150%
what the wallet banked — its price, not yours
Block 0
+100%
the same block as them — co-located bots only
Block +1
+70%
a fast bot, on a good day
Block +2
+50%
you saw it, you reacted, yours landed
+2 seconds
−50%
the token has already turned — you are buying the exit

Inside the tool

Backtested at all three blocks after the wallet

enlarge Three entry prices side by side: same block, +1 block, +2 blocks
Inside, every single trade carries its three entries. Same block, +1, +2 — and what the token did after each. Other boards show only the first column.

On every wallet sheet

Which take-profit, and how far they actually went

Every exit level tested one by one, gain against win rate — and the shape of what each token did after the entry. The best level is almost never the one that wins most often, and an average hides that completely.

enlarge Take-profit curve: gain per trade and win rate at each level
Every exit level, tested one by one. Gain per trade on the left, win rate on the right. They pull in opposite directions.
enlarge How far its tokens go: 60% flop, 15% double, 1% reach x10
And how far its tokens actually went. 60% flop, 15% double, 1% reach ×10. An average hides that shape.

The method, in the trade's own terms

Backtested on four periods, not one

Every wallet is tuned and measured on each quarter of its own history. A target that holds from one quarter to the next means something; one that jumps around means nothing. And the published score is out-of-sample: the tuning never saw the trades that judge it.

Holds +200%+250%+250%+250% the same target, quarter after quarter
Means nothing +50%+400%+25%+300% tuned on noise — it will not hold tomorrow
Wallets examined last night16,013,361
Active enough to be judged1,302,089
Scored on both sides of the split800,368

Six stages, in this order

  • Execution lag — entry simulated at +0, +1, +2 blocks.
  • Outlier filtering — one impossible tick skews everything.
  • Fill realism — the third level reached, not the wick.
  • No look-ahead — the exit is found on the past, then frozen.
  • Costs and slippage — 1% both sides, curve slippage included.
  • Out-of-sample split — chronological, never random.

The AI connector

Your AI, plugged into 19 million wallets

Plug your own assistant straight into the base and let it do the digging. It reads the 79 columns itself — you never write a query, and there is no quota.

Claude ChatGPT Codex Cursor and any MCP client

One key, generated from your subscription. Revoked the day it lapses. The Lab does the same from the site, with nothing to install.

AI connector · Lab
Find me wallets active this week that enter within 30 seconds, with under 15% of their buys on rugged tokens.
Querying the base… 412 wallets match. Top ten by copyable gain at block +2:
9M3pKqcFscpVrK6Y6D6EEk7xQSCa7jzB38iGKNEDLAGZ — 218 tokens, ×1.30 copied, 13 trades at ×10.

The rest of the house

Six more tools, one key

Smart Money16.09
05:36$windin at 3.5k · ×4
04:21$BAPin at 4.5k · ×4
02:36$IMRANin at 5.2k · ×4
01:05$habibiin at 4.1k · ×4

Smart Money

Who just bought what, a second or two after the block — real entries from last night, at 3.5k to 5.2k of market cap.

Rugger radar
Launch factory121 devs
Tokens issued342
Wallets in the army731
Funded fromone exchange deposit

Rugger radar

Whole factories mapped from a single funding address — one of them ran 342 tokens.

Rugs / Mooners
$Family×283
$Tusk×206
$NYEHEHE×204
$BID×178

Rugs / Mooners

What went up and what collapsed, day by day, with the multiple each one reached.

Top Insider

Addresses whose trades do ×10 after a copier gets in, not before.

Optimal TP

A wallet's full sheet: targets, selling speed, gain curve.

Tools

Scan a token, compare two wallets, find an address's twins.

Access

One subscription, no add-ons

You sign in with Telegram and pay in SOL from the bot. Everything opens at once: the site, the alerts and the AI connector. There is no higher tier to sell you, no query quota, no commitment — the subscription stops when you stop paying.

  • The raw base, its 79 columns and its export
  • The test bench and the sheet for every scored wallet
  • The AI connector for Claude and Codex, plus the Lab
  • Smart Money alerts in real time
  • Your first month's price stays yours if it rises later
0.5 SOL / month

Paid from the Telegram bot in one transaction. Nothing to cancel: just don't renew.

Sign in with Telegram

You keep access until the last day you paid for.

The base is open. The sorting is yours.

19,342,830 wallets, 787,571,505 trades, 79 columns and a measure of what a copy would really have paid. Nobody picks for you any more.

Sign in with Telegram
GTT Signals Free alerts on Telegram