ResearchMarch 2026~8 min read
Toxic Flow in Low Volume Assets
When someone trades against a liquidity provider using information the provider does not have, the provider loses money. In high volume markets, these losses are absorbed across thousands of trades. In low volume tokenized assets, one informed trade can wipe out weeks of income.
1. The Problem
A market maker posts a price to buy and a price to sell. The gap between them is the spread. The spread is only profitable if the market maker can flip the position: buy from one counterparty, sell to another, pocket the difference.
That breaks when the counterparty knows something the market maker does not. An insider who knows a fund's NAV (net asset value) is about to drop sells at today's price. The market maker buys, unaware. The NAV drops. The market maker holds a position worth less than what was paid. The insider captured the difference.
How adverse selection works
1
Informed seller dumps position at current price.
2
Market maker buys at the quoted price, unaware of the news.
3
Bad news reaches the market. Price drops.
4
Market maker holds a loss. The informed seller captured the difference.
In liquid markets, this cost is manageable. A market maker on the S&P 500 loses on a handful of informed trades but earns the spread on thousands of uninformed ones. The wins cover the losses.
In a tokenized RWA that trades a few times a day, there are not enough uninformed trades to absorb the cost. One informed trade can erase weeks of spread income.
2. Why Existing Tools Do Not Work Here
Traditional finance has several methods for detecting informed flow. All of them require one thing tokenized RWAs do not have: volume.
Why standard detection methods fail for low volume assets
VPIN
Measures order flow imbalance in volume buckets.
Buckets take hours or days to fill. By the time it updates, the damage is done.
PIN
Estimates the proportion of informed traders from order arrival rates.
Backward looking. It tells you yesterday's toxicity, not today's.
Amihud Ratio
Measures how much price moves per unit of volume.
Measures fragility, not intent. A fragile market is not necessarily seeing informed flow.
Realized spread
Measures actual profit after the price moves post trade.
Only measurable after the loss has happened.
| Method | What it does | Why it fails for low volume |
|---|
| VPIN | Measures order flow imbalance in volume buckets | Buckets take hours or days to fill. Lags the event. |
| PIN | Estimates informed trader proportion from order arrival rates | Backward looking. Yesterday's estimate, not today's. |
| Amihud | Price impact per unit of volume | Measures fragility, not intent |
| Realized spread | Actual profit after post trade price movement | Only measurable after the loss |
On chain, the standard solution is an AMM (automated market maker), a pool that people trade against instead of a human counterparty. AMMs have their own version of the same problem. Every time the off chain NAV of a tokenized fund updates, an arbitrageur can trade against the stale on chain price and extract the difference from the pool. The liquidity provider is forced to sell at yesterday's price every time there is new information.
For low volume RWAs, standard AMMs function as a transfer of value from passive depositors to informed traders. The detection methods require volume that does not exist. The on chain pools require price freshness that illiquid assets cannot provide. A different mechanism is needed.
3. Silence as Signal
Every detection method above tries to analyze what is happening inside the flow: which trades are buys, which are sells, how imbalanced the flow is. That analysis requires thousands of trades to produce a meaningful result. In a market with three trades per day, the analysis is noise.
But there is a signal that does not require volume at all: the gap between trades. In thin markets, the only reason someone shows up after hours of silence is that something changed. An off chain NAV updated. A credit event occurred. A redemption gate was announced. The combination of prolonged silence followed by sudden activity is itself the indicator.
A smart contract can measure this directly. Time since the last trade is on chain state. Order size relative to pool depth is computable at execution. Both are available the moment the trade is submitted. No off chain data. No volume history. No proprietary classification.
The mechanism: scale the trading fee based on these two inputs. When both are elevated (long silence, large order), the fee goes up. When both are low (recent activity, small order), the fee stays at baseline. The fee is not a penalty. It is compensation to the liquidity providers for the elevated probability that the trade is informed.
How the fee responds to context
Small trade, recent activity. Normal fee. Nothing unusual about this trade.
Small trade, long silence. Slightly elevated. The silence raises baseline risk but small size limits damage.
Large trade, recent activity. Elevated. Size alone carries risk, but recent activity means the pool is turning over.
Large trade, long silence. Highest fee. This is the signature of informed flow in thin markets.
fee = base_fee × (1 + time_factor × hours_since_last_trade) × (1 + size_factor × order_size / pool_depth)
Same order, different context, different fee
Base fee: 0.3% · Order size: 30% of pool · Time factor: 0.1 · Size factor: 1.0
30 minutes after last trade
0.3% × (1 + 0.05) × (1 + 0.3)
0.41% fee
8 hours after last trade
0.3% × (1 + 0.8) × (1 + 0.3)
0.70% fee
Same order. 70% higher fee after 8 hours of silence. The difference stays in the pool.
The fee is deterministic. Every participant sees the same formula. There is no hidden market maker logic, no proprietary classification, no discretionary spread. The rules are in the contract. The higher fee goes directly into the liquidity pool, compensating the people who deposited capital for others to trade against.
The principle is not new. IEX, the exchange made famous by Michael Lewis's Flash Boys, delays incoming aggressive orders by 350 microseconds while processing cancellations instantly. Same idea: give the passive participant a structural edge. IEX does it with speed. This does it with fees. The difference is that fees produce revenue for liquidity providers rather than simply slowing everyone down.
This does not guarantee profit. If the underlying asset declines, liquidity providers still lose. The system addresses one specific category of loss: adverse selection from informed flow exploiting timing advantages. It does not eliminate market risk.
4. Where It Applies
Most markets do not need this. On Uniswap, ETH/USDC trades thousands of times per minute. Time since last trade is always near zero. The fee would never trigger. Centralized exchanges can freeze accounts, reverse trades, and ban bad actors. Liquidity providers on liquid AMMs can simply withdraw if toxic flow becomes a problem, letting the pool shrink until the effective spread naturally widens.
Time weighted fees apply where four conditions overlap: low volume, on chain execution, no centralized enforcement, and liquidity providers who cannot easily exit because the underlying asset has lockups or redemption queues. That intersection is specifically where tokenized RWAs sit.
The mechanism also requires trades to happen. It works when users sell, when liquidators buy at discounts, when any participant interacts with the pool. If the asset has zero activity for extended periods, there is no fee to collect and no signal to measure. Bootstrapping that initial activity is a precondition.
5. Open Problems
Calibration is per asset class. A tokenized credit fund accruing yield at basis points per day has different risk characteristics than tokenized real estate that reprices quarterly. The time factor that works for one will over or under charge on the other. A maximum fee cap prevents the formula from penalizing the first legitimate trade after a genuinely quiet period.
Informed participants will adapt. If large trades after quiet periods are penalized, traders will break orders into smaller pieces spread over time, mimicking benign flow. Tracking cumulative volume from a single address within a rolling window closes the most obvious evasion path but adds complexity.
The time weighted fee is the first mechanism designed specifically for markets where the absence of volume is itself the primary risk signal. Whether it performs in practice depends on calibration against live trading data, which does not yet exist in sufficient quantity for tokenized illiquid assets.