Risk models are designed to estimate how much a position can lose under adverse market conditions. In crypto, one of their most important assumptions is often less visible: that a position can still be closed at something close to the quoted market price when conditions deteriorate.
In a calm market that assumption looks reasonable. Order books are active, spreads are narrow, and recent trading data suggests that even a relatively large position can be sold without moving the market very far. Liquidity starts to look like a property of the asset itself.
The next serious crypto sell-off may challenge that assumption. Volatility can rise at the same moment that market depth falls: quotes get smaller, spreads widen, and an order that would normally be routine starts moving the price. A model built mainly on historical price behavior can then be broadly right about volatility and still badly underestimate the cost of getting out.
The point is not that a crash is imminent. The concern for 2025 is that market structure can amplify the next serious shock as leverage, automated liquidations, fragmented venues, and professional liquidity providers react at the same time.
I have spent years working across digital-asset and blockchain projects, including as the chief technology officer of a cryptocurrency exchange. One problem I kept returning to was the gap between quoted liquidity and executable liquidity. Portfolio models can make a deceptively simple assumption: that the liquidity visible in normal conditions will still be there when many market participants need it at once.
Today’s liquidity may not survive tomorrow’s market
Most risk systems begin with observable history: volatility, correlations, trading volume, spreads, and market depth. Those inputs are useful for estimating how an asset behaves under familiar conditions. The harder problem is deciding how much of today’s liquidity will still exist after conditions change.
Tomorrow’s liquidity may look very different from today’s.
A 2025 study in the Journal of Banking & Finance offers a useful reminder. The authors found that the economics of providing crypto liquidity vary with volatility, crash risk, tail risk, and other measures of market stress. Higher liquidity-provision premiums were associated with lower liquidity and trading volume, fewer transactions, more withdrawals, higher fees, and greater impermanent loss. Liquidity is therefore not a fixed backdrop to risk; it changes with the conditions that create the risk.
Market makers also have their own risk limits. When conditions become dangerous, they can reduce the capital committed to positions, quote smaller size, widen spreads, or leave a market altogether. These reactions can happen quickly because much of the process is automated. The order book used to estimate risk in the morning may therefore look very different by the time a stressed position actually has to be closed.
Price risk and liquidity risk can diverge sharply. A model can estimate the direction and magnitude of a price move reasonably well while still understating the slippage required to sell a large position into a deteriorating book.
Order-book depth can be more concentrated than it looks
A deep-looking order book can also overstate how many independent sources of capital stand behind the visible bids and offers.
A large number of orders does not necessarily mean that an equally large number of independent traders are willing to take the other side. On some venues, a meaningful share of quoted depth may come from a relatively small group of professional firms running automated strategies.
That concentration is useful in normal trading because market makers keep spreads tight and provide continuous quotes. It becomes a vulnerability when the same firms cut risk at the same time. The 2022 collapse of FTX and Alameda offers a precedent: Kaiko later called the persistent decline in order-book liquidity the “Alameda Gap.” As of late 2023, 1% market depth was still about 50% below its pre-FTX level, partly because market makers had suffered losses, left the market, or revised their risk policies.
The same pattern appears in credit and counterparty risk. The failure of Three Arrows Capital in 2022 did not remain an isolated hedge-fund problem. The Chicago Fed documented large exposures at Genesis, BlockFi and Voyager, while Celsius also had a smaller $75 million loan to 3AC. The crises were not simply a chain of failures caused by 3AC — Celsius, for example, was already facing its own liquidity problems — but overlapping exposures helped transmit stress across the market. These cases have a common feature: the same counterparties, venues, and liquidity providers often sit inside many portfolios, allowing one source of stress to reinforce another.
Because firms often respond to the same deterioration with similar risk controls, liquidity can disappear across several venues without any coordination between them.
The next shock could be amplified by the trades it triggers
This matters most when somebody has to sell rather than wants to sell. In a deep market, a large order can be absorbed across many price levels with limited slippage. If liquidity providers have already reduced their quotes, the same notional order can consume a much larger share of the book and start moving the market by itself.
Once the trade moves the price, stop orders and margin thresholds can be hit. Exchanges begin closing leveraged positions automatically. Those liquidations create new market orders that have to be absorbed by an order book that may already be thinner than it was minutes earlier.
The mechanism can become self-reinforcing. Rising volatility causes liquidity providers to reduce depth just as forced sellers need it most; their orders then move prices further, pushing additional leveraged positions through margin thresholds. The resulting liquidations send another wave of orders into an already thinner book, so each round of selling can make the next one larger and more expensive.
A future sell-off could therefore become much larger than the event that starts it. A macro shock, protocol failure, hack, or exchange problem does not have to be catastrophic on its own. If it arrives when leverage is high and liquidity is retreating, forced selling can turn the original shock into a self-reinforcing market event.
A risk model that treats future selling as mostly discretionary misses this mechanism. Under stress, part of the next wave of order flow may already be encoded in leverage and liquidation thresholds.
How to stress-test a crypto portfolio before the next shock
A useful crypto stress test should go beyond a price shock. Asking what happens if Bitcoin falls 10%, 20% or 30% is a start, but the answer changes dramatically if the same move happens in a thin order book, with wider spreads and impaired access to one of the venues holding collateral.
I use stress testing not to predict the next crash, but to find the assumptions a portfolio depends on and see what happens when several of them fail at once. One deliberately severe scenario might combine a 30% price decline with a 90% reduction in available depth, spreads ten times wider than normal, the withdrawal of a major liquidity provider, and one important exchange temporarily unavailable. These numbers are not a forecast. The point is to test whether the portfolio can survive a market that behaves very differently from its recent history.
The test should check whether any leveraged position reaches forced liquidation, whether collateral can still be moved where it is needed, whether the portfolio can be reduced or closed within a day, and how much slippage a crowded exit would generate. A fund can also set an explicit maximum tolerated exit cost — for example 3–5% of portfolio value — and size positions so that the stressed scenario stays within that limit.
Average daily volume is a weak substitute for this analysis. What matters in a crisis is the depth that remains near the market price and the amount of genuine risk-bearing capacity still willing to take the other side of an urgent transaction.
The warning signs to watch in 2025
Historical risk models still have value, provided liquidity is monitored as a changing market condition rather than entered once as a stable input.
The warning signs are mostly structural: shrinking market depth relative to position size, widening spreads, increasing divergence between venues, greater dependence on a small set of liquidity providers, rising leverage relative to available depth, and concentrations of liquidation levels around the same prices. Venue and collateral concentration deserve the same attention.
The next major crypto dislocation may begin with an ordinary trigger and become extraordinary because the exits disappear. When the next serious shock arrives, the availability and cost of liquidity may matter as much as the size of the initial price move. Portfolios built on the assumption of an orderly exit will be the most exposed.
