We pulled every recent callout from pump.fun's top-12 weekly PnL wallets and computed what their public calls are actually worth. The result should rewire how you copy-trade: even the best callers on the platform see barely half their calls reach 2x — and the leaderboard's eye-popping averages are carried by a handful of lottery-ticket outliers, not repeatable skill you can mirror.
This is the dataset behind Xanguard's caller track-record feature. Methodology at the bottom; the numbers first.
The data: 307 calls, 12 whale wallets
We took pump.fun's weekly PnL leaderboard (the wallets that made the most money trading this week), paged through each wallet's public callout history (up to 30 most recent calls each), and computed per-caller statistics: median multiplier, mean multiplier, and the share of calls that reached 2x or better. Data pulled September 2, 2026.
| Rank (weekly PnL) | PnL (SOL) | Calls | Median call | Mean call | Calls that 2x+ |
|---|---|---|---|---|---|
| #1 | 13,289 | 30 | 2.51x | 261.0x | 63% |
| #2 | 6,317 | 3 | 2.26x | 9.8x | 67% |
| #3 | 5,650 | 30 | 2.84x | 15.9x | 60% |
| #4 | 5,118 | 30 | 1.90x | 2.8x | 47% |
| #5 | 4,859 | 2 | 2.23x | 2.2x | 50% |
| #6 | 4,741 | 30 | 1.67x | 3.1x | 40% |
| #7 | 4,422 | 30 | 1.89x | 4.2x | 43% |
| #8 | 4,003 | 30 | 1.71x | 2.4x | 43% |
| #9 | 3,907 | 17 | 1.87x | 839.6x | 47% |
| #10 | 3,623 | 3 | 2.19x | 4.0x | 67% |
| #11 | 3,095 | 30 | 2.30x | 4.9x | 63% |
| #12 | 2,950 | 30 | 1.76x | 3.8x | 40% |
Finding 1: Even the best callers are a coin flip on any single call
Median call multipliers across the top-12 weekly earners: 1.67x to 2.84x. Half of their calls do worse than that. The share of calls reaching 2x or better runs 40–67% — meaning the best traders on pump.fun, the wallets up thousands of SOL this week, post calls that fail to double 33–60% of the time.
If your mental model is "whale called it, so it's going up" — the data says you're taking a slightly-better-than-coin-flip. That doesn't mean don't mirror callers. It means position sizing against the base rate is the entire game.
Finding 2: The mean is a lie
Look at the #1 wallet: 13,289 SOL in weekly PnL, and a mean call multiplier of 261x. The median call: 2.51x. The gap is the story — one or two absurd outliers (a 100x, a 1000x) drag the average to a number no follower will ever experience. Wallet #9 is the extreme case: 839.6x mean against a 1.87x median.
Leaderboards rank on total PnL. Total PnL is made of outliers. A caller's leaderboard position tells you they got lucky at least once. Their median call tells you what happens if you copy the next one. These are not the same number, and the difference is where followers get farmed.
Finding 3: Profitable wallet ≠ profitable caller
PnL measures what they made trading. Callouts measure what they told the public to buy. Our earlier research found stretches where top-PnL wallets had median public calls below 1x — actively losing money for everyone who mirrored them while their own PnL printed. This week's crop is healthier, but the structural warning stands: the wallet's trading skill and the wallet's public calls are different products, and only one of them is sold to you.
What to do with this
- Judge callers on median and win rate, never the leaderboard rank. A 60%+ rate at 2x+ with a 2.5x median is a mirroring candidate. A 43% rate with a fat mean is a lottery ticket seller.
- Size for the base rate. If even elite callers 2x half the time, full-port apes are a donation. The Kelly-flavored version: small, repeatable, survivable entries.
- Speed still matters most. The callers who do print are front-run by their own followers within seconds — the entry you get depends on how fast you see the call, not just who's calling.
The tooling answer
Doing this analysis by hand means paging through callout histories and multiplying. We built it into the alert itself: every Xanguard callout alert carries the caller's win rate, median multiple, call count, and follower count — computed from their full callout history, refreshed continuously. You see whether the caller is worth mirroring in the same message that tells you what they called.
👤 Cupsey called at $14.6K · token 47m old
💰 Holding 8.4M (0.84%) · $126.99 cost basis
🏆 his calls 2x+ 61% of the time (23 calls) · 12.4K followers
🪙 Curve 12.4 SOL · ATH $18.2K · bonding
Milestone alerts (2x/5x/10x with time-to-hit) track the call after entry. Plans from $100/month, SOL, 60-second setup in @PF_Xanguard_bot with /call. Bots get the same data over WebSocket — details on the pump.fun callout alerts page.
Methodology
Source: pump.fun weekly PnL leaderboard (top 12 wallets by PnL, week ending 2026-09-02) and each wallet's public callout history via pump.fun's own API (up to 30 most recent callouts per wallet, 307 calls total; wallets with fewer than 30 lifetime calls analyzed as-is). Multiplier = the callout's maximum observed multiple (maxMultiplier). Win rate = share of calls with maxMultiplier ≥ 2.0. Small-sample wallets (#2, #5, #10) are marked by their call counts. This is a point-in-time snapshot — caller performance drifts week to week, which is exactly why the track record belongs inside the alert, recomputed continuously.