Blog
Blog
What Healthier Markets Look Like Under the Surface
A tight spread is not a healthy market. It is evidence that a market maker was willing to quote one.
Traders talk about liquidity constantly, usually right after a fill goes worse than the chart suggested. But when pressed to define liquidity precisely, most fall back on spread. The spread is the menu price. The actual execution depends on everything behind it.
A microstructure report card measures five distinct dimensions. A market can look healthy on one and fragile on the rest. The common mistake is treating them as a single variable.
Dimension 1: Depth
The spread tells you the price at which a market-maker is willing to trade one unit. Depth tells you how many units are available at each price level before the market-maker steps back and reprices.
Consider an illustrative comparison. Market A quotes a 0.02% spread with $2,000 resting at the best ask. Market B quotes a 0.08% spread with $400,000 resting across the first ten levels. A retail order for $500 fits comfortably in Market A. An institutional order for $50,000 drains Market A's best ask in under two seconds and walks up the book. Market B is the liquid one. The spread suggested otherwise.
Depth is measured at multiple levels and weighted by distance from the mid-price. Levels close to execution carry more weight than levels far away, because orders at the extremes are sometimes not genuine. Resting depth that never gets hit and never gets cancelled is often cosmetic. Depth that consistently absorbs flow without moving is the real measure.
Dimension 2: Resilience
A healthy market bounces. An unhealthy market sags.
When a large market order consumes the best ask and price ticks up, two things can happen. Resilient markets see new limit orders arrive within seconds, refilling the consumed depth and returning the spread to normal. This is the signal that the market-maker community is active, confident in their models, and willing to re-engage at the new price. In fragile markets, price ticks up, depth does not refill, and the next small buy order walks price further. The original move was not driven by information. The follow-through was driven by emptiness.
Measuring resilience requires watching the orderbook at sub-second resolution after a large fill event. Time to refill the top of book. Rate at which depth returns. Whether the spread widened and stayed wide or contracted back. These numbers differ enormously between venues and between times of day. The same instrument on the same exchange at 3 AM versus 2 PM can show resilience that differs by a factor of five.
Resilience is the structural variable that separates exchanges that look liquid from exchanges that are liquid when it matters.
Dimension 3: Order Flow Imbalance
Depth and resilience measure the resting state of the market. Order Flow Imbalance measures what is happening dynamically, at the moment of decision.
OFI counts aggressive buying minus aggressive selling over a rolling window: the volume of traders crossing the spread to act now, directionally signed. Sustained positive OFI in a flat market is pressure that has not yet resolved. Sustained negative OFI into a thinning bid is the last seconds before a sharp move down.
Cont, Kukanov, and Stoikov established in 2014 (Journal of Financial Econometrics, 12(1):47-88) that OFI has statistically significant explanatory power for mid-price changes at horizons of seconds to minutes, beyond what price itself provides. The chart was the lagging indicator. OFI was the leading one.
A healthy market's OFI pattern is mean-reverting under normal conditions. Bursts of buying or selling are absorbed, the imbalance fades, and the orderbook resets. Markets in distress show persistent OFI in one direction for minutes at a time, which is the signature of uninformed panic on one side or large informed flow on the other. Distinguishing between the two requires the fourth dimension.
Dimension 4: Informed Flow Probability
Not all flow is equal. A broker executing a pension fund's quarterly rebalance creates order flow. A hedge fund with better information about the next two hours creates order flow. Both show up as signed volume. Only one predicts where price goes next.
VPIN (Volume-Synchronized Probability of Informed Trading), developed by Easley, Lopez de Prado, and O'Hara, was designed to estimate what fraction of recent volume represents genuinely directional, risk-bearing participants rather than noise, arb, or fabricated activity. High VPIN is the market's signal that someone who knows something is transacting. Market-makers respond by widening spreads. The cost of adverse selection goes up because the probability of being on the wrong side of an informed trade goes up.
Crypto markets create a specific problem for any informed flow metric. Wash trading inflates the denominator. Volume-synchronized methods that assume all volume represents real inventory risk are miscalibrated on exchanges where a meaningful fraction of volume never changes hands in any genuine economic sense. Healthy markets exhibit VPIN patterns that match expected flow distributions. Markets with significant wash trading show VPIN values that are structurally low regardless of conditions, which is the tell: real informed flow exists in every liquid market, and an exchange where VPIN never spikes is an exchange where the denominator is being padded.
Dimension 5: Price Impact
The previous four dimensions describe what the market looks like. The fifth measures what it does to you.
Kyle's Lambda, formalised in the original market microstructure paper by Albert Kyle in 1985, estimates how much each unit of signed volume moves price. In a deep, resilient, balanced market, Lambda is low: large trades move price minimally because the available liquidity absorbs them. In thin, fragile markets, Lambda is high: even moderate-sized trades cause disproportionate slippage.
Lambda is not stable. The same instrument on the same exchange shows Lambda that varies by hour of day, by volatility regime, by day of week, and by the current state of the orderbook. A position sizing model built on a 30-day rolling Lambda will systematically underestimate cost during volatility spikes, because Lambda rises when orderbooks thin and volatility rises simultaneously. That is precisely when accurate position sizing matters most.
An institution that sizes positions using stale Lambda estimates discovers the error in the trade. A system that updates Lambda estimates continuously can detect when market conditions have shifted from the regime in which the original sizing decision was made.
What a Healthy Market's Report Card Looks Like
A healthy market passes all five dimensions simultaneously. The spread is tight because depth is real and resilient. OFI is mean-reverting because no single participant is overwhelming the market. VPIN is responsive because informed traders are present but not dominant. Lambda is low because depth is absorbing flow efficiently.
These five numbers rarely all tell the same story at the same time. That divergence is the information.
Tight spread with low depth is a market-maker holding a position they want to exit. Mean-reverting OFI with rising Lambda is depth that is present but shallow, where the available resting orders are smaller than the flow that is arriving. High VPIN with tight spread is a market-maker who has not yet adjusted to the informed flow that is building. Each combination tells a different story about what the next few minutes are likely to look like.
The honest limitation: no single combination of these metrics is a reliable directional signal in isolation. Markets are adaptive. Patterns that persist become known, and known patterns attract participants who trade against them until the pattern disappears. These five dimensions are most useful as filters, as context for the signals you already have, and as regime detectors that tell you when the market's plumbing is normal versus when something structural is shifting.
That is why this report-card framing belongs beside what market depth actually measures, because depth is one dimension, not the whole answer. It also belongs beside a technical framework for exchange market quality, where the same structural questions become an exchange-selection problem rather than only a trading one. And when the informed-flow dimension matters most, VPIN explained gives the narrower lens.
The Ring Closes
Every trader who talks about liquidity is describing a market. These five dimensions are the measurement of that description.
Healthy markets have tight spreads because the depth underneath them is real. Resilience confirms that market-makers trust their models enough to re-engage after a hit. Mean-reverting OFI confirms that no participant is running the book. VPIN confirms that informed flow is present but not overwhelming. Low Lambda confirms that the structure is absorbing activity without breaking.
The alternative is a market that quotes tight and breaks on contact. You have traded in one. The spread looked fine until the moment it mattered, and then the book was empty three ticks deep and your fill was wherever the next resting order happened to be.
That gap between the spread and the reality is measurable. The engineering is non-trivial, but the measurement exists. Traders who know the spread and nothing else are reading the menu price. Traders who know all five dimensions know what they are actually paying.
Because spread only shows the first quoted price. It does not show whether deeper liquidity is real, whether the book refills, or whether price impact is rising.
Real depth, resilience after a hit, balanced flow, responsive informed-flow signals, and low price impact relative to traded size.
Yes. A market can quote tightly while deeper liquidity is shallow or while informed flow is already stressing the book.
Because the quality of the book determines whether a larger order fills cleanly or walks into a thin and expensive auction.
Because divergence between them often tells the most useful story about whether the current calm is structural or only cosmetic.