Wash trading: real transactions can still describe fake demand

The short answer

Wash trading is trading arranged so the same economic controller—or coordinating parties without genuinely independent interest—appears on both sides while creating misleading activity or volume. On a DEX, the swaps, fees, and pool reserve changes can all be real. What may be artificial is the appearance of independent demand, price discovery, or holder participation.

Legal definitions and enforcement vary by product and jurisdiction. This guide is an evidence workflow for market analysis, not a legal determination about a particular address or trade.

Volume counts flow, not independent conviction

Market dashboards aggregate swap value over a time window. Buys, sells, transactions, and maker counts can therefore rise when one actor uses several wallets, when routing creates multiple pool interactions, or when bots repeatedly recycle inventory. The number can be accurate as indexed activity and misleading as evidence of broad demand.

Always resolve the exact pair. The same token can trade through several pools and quote assets, and aggregators can split one user's route across venues. Compare the dashboard with the pool transactions and use the DEX Screener guide to separate pair-level observations from token-wide conclusions.

AMMs change the visible counterparty, not the economic question

In an AMM, a trader normally swaps against pool reserves rather than matching a named wallet's resting order. A controller can buy from a pool with one address and sell back through another. The pool is the immediate onchain counterparty in both events, while the controller relationship sits outside the swap instruction.

That means a direct “wallet A traded with wallet B” test misses important cases. Reconstruct funding, token transfers, fee-payer behavior, timing, and where the quote asset and token inventory end up. Include LP ownership when the suspected controller may recover part of the trading fees.

Look for a bundle of mutually reinforcing signals

• Repeated round trips: related wallets alternate buys and sells over short windows while returning near their starting token exposure.
• Common economic funding: the wallets receive SOL or quote assets from a common nonservice source, transfer inventory between one another, or consolidate proceeds to the same destination.
• Mechanical execution: repeated sizes, intervals, routes, priority settings, or transaction construction appear across the wallet set.
• High gross, low net change: reported volume is large relative to liquidity while the cluster's combined token and quote position changes little after accounting for transfers.
• Activity without broadening: transactions rise while independent funded participants, durable holders, and external inflows remain limited.

No item is decisive alone. Published DEX research has identified self-trade, two-account, and more complex structures in historical order-book DEX data. AMM analysis must adapt the controller and position logic to pool-based execution.

Reconstruct economics after fees, not just direction

Group only addresses supported by the entity-inference workflow. For each window, calculate starting and ending token balance, quote-asset balance, external deposits and withdrawals, token transfers, swap inflows and outflows, network fees, priority fees, tips, transfer fees, and estimated pool fees. Gross buy and sell volume can be huge while net exposure barely changes.

Cost does not make the activity organic. A manipulator may accept fees to obtain trending placement, apparent liquidity, attention, or a price path that attracts outside buyers. But meaningful cost and inventory risk are relevant evidence and should be measured rather than assumed away.

Compare volume with liquidity, makers, and external flow

Volume-to-liquidity is a context ratio, not a detector. A popular volatile pool can turn over its reserves many times legitimately. Add unique funded entities, repeat-participant share, median holding time, cluster share of volume, external net inflow, price divergence across pools, fee burden, and persistence after the suspicious wallets stop.

Avoid using dashboard “makers” as a proxy for people. One person can create many wallets, one custodial account can represent many people, and an aggregator may add addresses involved in execution. Define and document how the interface counts the field before making a participation claim.

A wash-activity investigation workflow

1. Fix the market and window. Save chain, token, pool, quote asset, start and end blocks, dashboard source, and its displayed volume definition.
2. Collect the underlying swaps. Record transaction, signer, fee payer, route, amounts, time, pool, and fees; separate routed legs where possible.
3. Build entity hypotheses. Trace funding, transfers, consolidation, service labels, and behavioral similarity with confidence levels.
4. Reconstruct positions. Calculate combined opening, external flows, swaps, internal transfers, closing balances, and total execution cost.
5. Test alternatives. Check arbitrage, market making, routing, incentives, DCA, liquidity management, and independent market events.
6. State the narrow conclusion. Separate confirmed self-trades, activity consistent with common control, unexplained repetition, and ordinary high turnover.

For a trader, uncertainty about demand changes the exit assumption

If a small linked set appears responsible for a large share of activity, do not size from headline volume. Model what liquidity, price, and route remain if that activity stops or reverses. Require a live exit quote, lower the assumed participation rate, and define invalidation around external maker growth rather than the same cluster printing more volume.

What belongs in the journal

Record the pair and window, raw swaps, route legs, dashboard volume, wallet and entity counts, funding and transfer evidence, confidence labels, combined opening and closing positions, external flows, fees, cluster share of activity, volume-to-liquidity context, alternative explanations, and the exact conclusion supported by the evidence.

Primary sources