How to detect fake volume and wash trading on pump.fun

Volume is the most faked metric on pump.fun. Trending feeds sort by it, buyers screen by it, and a bot can manufacture it for pennies: buy, sell, repeat — the same SOL round-tripping through the same token, printing "activity" that never had a second party. This guide shows how to detect fake volume on pump.fun with data: measure wash trading, identify volume bots wallet by wallet, and check whether the first buyers of a launch are bots — using the Raiden API.

Why volume is the easiest metric to fake

Every discovery surface in the pump.fun ecosystem — trending tabs, "top movers" lists, volume screeners — ranks tokens by traded volume. That makes volume the one number an operator most wants to inflate, and it happens to be the cheapest one to inflate: unlike holders (which need funded wallets) or price (which needs net capital inflow), volume only needs the same capital to move back and forth. A wallet that buys 2 SOL and sells 2 SOL prints 4 SOL of volume and loses only the fees. Run that loop a few hundred times across a handful of wallets and a dead token looks like the hottest chart of the hour. Wash trading on Solana is especially cheap because transaction fees are fractions of a cent — the economics favor the botter.

One call: /tokens/{mint}/wash-traders

The detection endpoint scans the token's last 24 hours and flags round-trippers — wallets with at least 5 buys AND at least 5 sells in the window. A wallet that only accumulates is a buyer, not a wash trader; it is the repeated in-and-out cycling that inflates the tape:

curl -H "X-API-Key: $RAIDEN_KEY" \
  "https://terminal.raiden.wtf/api/tokens/9xQm…pump/wash-traders"

{
  "mint": "9xQm…pump",
  "window_h": 24,
  "flagged": 6,
  "shown": 6,
  "wash_vol_sol": 128.4,
  "total_vol_sol": 472.9,
  "wash_pct": 27.2,
  "data": [
    { "wallet": "6hh9…Nfew", "buys": 41, "sells": 39, "vol_sol": 88.2 }
  ]
}

Reading wash_pct honestly

Round-tripping is not automatically a scam. Scalpers cycle in and out of hot tokens; some market-making-ish behavior round-trips by design. The signal is not the existence of round-trippers — it is the share of volume they account for, and how the tape is shaped:

• Balanced counts — a wallet at 41 buys / 39 sells is cycling, not trading a view. Perfectly matched buy/sell counts across several wallets is the classic wash signature.
• Concentration — one wallet whose vol_sol is a large slice of total_vol_sol means the "market" is mostly one machine talking to itself.
• The share — a token with wash_pct in the single digits has a normal amount of churn. A token where a quarter of the volume is round-tripped deserves a closer look. A token where most of it is round-tripped is advertising activity that is not there.

How the trust score prices it in

GET /tokens/{mint}/trust — the live 0–100 risk score (higher = safer) — folds wash trading in as one of its five weighted signals, alongside dev reputation, known rug-dumpers in the token, bundled supply and dev exit. The wash weight is 0.15, and each signal reports its raw value plus its contribution (comp) to the final score:

{
  "mint": "9xQm…pump",
  "score": 45,
  "verdict": "caution",
  "signals": {
    "dev_score":   { "value": 0,    "comp": 8.4, "known": true },
    "bundled_pct": { "value": 52.0, "comp": 0.0, "known": true },
    "wash_pct":    { "value": 27.0, "comp": 4.3, "known": true }
  },
  "weights": { "dev": 0.35, "offenders": 0.25, "bundle": 0.15, "exit": 0.10, "wash": 0.15 }
}

The weight is only half the story. When wash_pct crosses 40%, a hard cap kicks in: the score is capped at 45 — inside the caution band (verdicts: clean ≥65, caution 40–64, likely_rug <40) — no matter how good the other signals look. A token whose tape is mostly fake can never rate as clean, even with a reputable dev and an unbundled launch. The full scoring model, signal by signal, is covered in the pump.fun rug check guide.

Cross-check the tape: /tokens/{mint}/swaps

Numbers should survive an eyeball test. Pull the raw tape and look for the same trader alternating sides within seconds:

curl -H "X-API-Key: $RAIDEN_KEY" \
  "https://terminal.raiden.wtf/api/tokens/9xQm…pump/swaps?order=desc&limit=50"

{
  "data": [
    { "time": "2026-07-23T14:02:11Z", "trader": "6hh9…Nfew", "is_buy": false,
      "sol_amount": "1998887301", "slot": 352114208, "block_index": 610,
      "router": "BSfD…bot", "sig": "3kPw…9dTa" },
    { "time": "2026-07-23T14:01:58Z", "trader": "6hh9…Nfew", "is_buy": true,
      "sol_amount": "2000000000", "slot": 352114177, "block_index": 402,
      "router": "BSfD…bot", "sig": "5mQx…hh2e" }
  ],
  "next_cursor": "2026-07-23T14:01:58Z"
}

The same wallet buys 2.000 SOL and sells ~1.999 SOL back thirteen seconds later: the round trip cost fees and printed ~4 SOL of "volume". Two more things to look at on the tape: the router field — the top-level program that orchestrated the swap (absent means a direct pump.fun / pump-amm call) — because bot volume tends to arrive through the same router program on every fill; and failed=1, which interleaves reverted attempts so you can see whether the "demand" ever competed for blockspace at all.

Cross-check the crowd: volume without holders

Real volume implies real people. GET /tokens/{mint}/holder-stats is one cheap call:

{ "holders": 61, "held": "812400000000000", "top10": "703900000000000", "top25": "798100000000000" }

Hundreds of SOL of daily volume on a token with 61 holders — where the top 10 hold almost everything — is not a crowd, it is a cast. GET /tokens/{mint}/activity adds the flow view: each window (5m/1h/24h) carries not just buys and sells but distinct buyers and sellers. Thousands of transactions from a few dozen distinct wallets is the same story from the other side.

How to see if the first buyers are bots

Fake volume rarely starts at hour six — tokens that wash usually launch coordinated. GET /tokens/{mint}/launch groups the token's first swaps into bundles: runs of buys landing in the same slot at consecutive block indexes — the bot ladder, pre-funded wallets submitted together before any human could react:

{
  "mint": "9xQm…pump",
  "creator": "5Q5q…Funr",
  "bundled_pct_supply": 23.4,
  "dev_bundle_pct_supply": 4.9,
  "bundle_count": 3,
  "bundles": [
    { "slot": 352114100, "start_index": 12, "size": 6, "is_bundle": true,
      "has_dev": true, "wallets": ["5Q5q…Funr", "8psN…VRtf", "…"],
      "sol_total": "2988888890", "pct_supply": 14.2 }
  ]
}

bundled_pct_supply is the share of supply grabbed by bundles at launch, and has_dev marks the bundle the creator was in. A token born with a bot ladder tends to keep its bots: the same wallets that sniped the creation slot are frequent round-trippers later. The full launch forensics — how bundles are detected and what the percentages mean — is in the bundle checker guide.

Screen for fake volume in Python

Put it together: take a screener page, run every token through the wash detector, and drop what fails your threshold. The threshold is yours to pick — the API reports the measurement, not a verdict:

import requests

BASE = "https://terminal.raiden.wtf/api"
H = {"X-API-Key": "YOUR_KEY"}
WASH_MAX = 25.0  # your tolerance: max % of 24h volume from round-trippers

hot = requests.get(f"{BASE}/tokens/trending",
                   params={"window": "1h", "limit": 50}, headers=H).json()

for t in hot["data"]:
    w = requests.get(f"{BASE}/tokens/{t['mint']}/wash-traders", headers=H).json()
    if w["wash_pct"] >= WASH_MAX:
        top = w["data"][0] if w["data"] else None
        line = (f"{t['symbol']:<10} wash={w['wash_pct']:5.1f}%  "
                f"{w['wash_vol_sol']:.1f}/{w['total_vol_sol']:.1f} SOL  "
                f"({w['flagged']} round-trippers)")
        if top:
            line += f"  top: {top['wallet']} {top['buys']}B/{top['sells']}S"
        print(line)

# PUMPY      wash= 61.3%  289.1/471.6 SOL  (14 round-trippers)  top: 6hh9…Nfew 41B/39S

For a deeper look at anything the screen flags, GET /tokens/{mint}/pack returns the whole dossier in one call — trust, wash traders, launch, holder stats, funding clusters and more, each key the exact response of the standalone endpoint.

The honest summary

• Volume is the most faked metric because it is the cheapest to fake and every trending feed rewards it.
• One number to check first: wash_pct — the share of 24h volume from wallets with ≥5 buys and ≥5 sells. Share, not existence: some round-tripping is organic.
• The trust score already prices it in — weight 0.15, and a hard cap at 45 once wash crosses 40% of volume.
• Cross-check with the raw tape (same trader alternating sides, same router), the crowd (holders and distinct buyers vs volume), and the launch (bot ladders in the creation slot).

Every endpoint here is documented with full response shapes in the API reference. The data is a complete record since May 1, 2026, kept on a rolling 6-month window — enough to check any token the trending tab throws at you, and the wallets behind it: when a round-tripper looks familiar, the wallet tracking guide shows how to profile it across every token it ever touched.