---
title: Composite Liquidity Factor (CLF) Overview
url: https://www.emergentmind.com/topics/composite-liquidity-factor-clf
type: topic
---

# Composite Liquidity Factor (CLF) Overview

Searching arXiv for the cited papers to ground the article and confirm relevant sources.
arxiv_search.query({"search_query":"id:2507.05749 OR ti:\"High Frequency Quoting Under Liquidity Constraints\"","max_results":5,"sort_by":"relevance"})
arxiv_search.query({"search_query":"all:\"Composite Liquidity Factor\"","max_results":10,"sort_by":"relevance"})
Composite Liquidity Factor (CLF) denotes a family of composite liquidity summaries rather than a single standardized construct. In current usage, the expression ranges from an instantaneous order-book-state proxy for directional execution fragility in high-frequency quoting, to a liquidity-horizon aggregation rule in market-risk capital, to an execution-cost-based basket liquidity algebra, and to low-dimensional state representations for concentrated-liquidity AMMs and algorithmic-liquidity diagnostics. Several papers directly use the term, while others state that they do not use “Composite Liquidity Factor” explicitly but analyze structures that practitioners often map into that label [2507.05749] [1803.07590] [1412.5072].

## 1. Terminology and domain-specific meanings

The main unifying feature of a CLF is compositionality. In the high-frequency quoting setting, the factor is “composite” because it combines price-level geometry and cumulative depth across multiple levels rather than using only a spread, only a best-quote depth, or only an imbalance [2507.05749]. In the Basel/FRTB setting, the same label is plausibly applied to a liquidity-horizon aggregation rule that combines nested risk-factor buckets with different liquidation horizons into one expected-shortfall measure, although the cited paper explicitly states that it does not use the term “Composite Liquidity Factor (CLF)” itself [1803.07590]. In the LIX framework, compositionality refers to aggregation in inverse raw-liquidity or execution-cost space across securities, baskets, and ETFs [1412.5072]. In concentrated-liquidity AMMs, the same expression is best understood as a reduced-form summary of fee capture, predictable loss, concentration risk, rebalancing frictions, and gas costs, or as a low-dimensional factorization of the full liquidity surface [2309.08431] [2509.05013].

These usages are not interchangeable. A CLF in one literature may be a static scalar, in another a regulatory aggregation formula, in another a cross-sectional basket operator, and in another a state vector or dynamic latent-factor system. This suggests that the term is best treated as context-dependent rather than canonical.

## 2. Order-book-state CLF in high-frequency quoting

In "High Frequency Quoting Under Liquidity Constraints" [2507.05749], CLF is a lightweight, order-book-state-based alternative to a Hawkes-process forecasting rule for selecting the reference leg in a calendar-spread quoting strategy. The execution problem is explicit: a trader places a passive quote on one leg, and once it fills, immediately sends an aggressive order on the other leg, the reference leg. Slippage is generated during the interval between the passive fill and the aggressive execution if the reference leg’s executable price moves. The decision question is therefore which leg should be used as the reference so that expected slippage is minimized.

The factor uses price and quantity information across the top $\nu$ levels of the relevant side of the limit order book. For the rollover direction studied, the relevant comparison is between the bid side of the current-month contract $F_c$ and the ask side of the next-month contract $F_n$, because those are the sides that would be hit or lifted by the aggressive completion order. The paper defines the levelwise quantities
$$
CLF^{b,i}_c = \frac{\log(p^{b,1}_c / p^{b,i+1}_c)}{\log(\sum_{j=1}^{i} q^{b,j}_c)},
\qquad
CLF^{a,i}_n = \frac{\log(p^{a,i+1}_n / p^{a,1}_n )}{\log(\sum_{j=1}^{i} q^{a,j}_n)}.
$$
The intended interpretation is logarithmic price movement from level $1$ to level $i+1$, normalized by logarithmic cumulative displayed depth over levels $1,\dots,i$. The technical note in the data states that the inequalities attached to the formulas appear inconsistent with ordinary LOB ordering; a plausible interpretation is that they are sign or typographical errors, and that the intended directional definitions are the usual one-sided bid-descending and ask-ascending versions.

The experiments instantiate several depth-specific variants, $CLF_1,\dots,CLF_4$, corresponding to increasing quote depth. The most faithful reconstruction in the data is
$$
CLF_c^{(k)}(w_1) := \frac{\log\!\left(\frac{p_c^{b,1}(w_1)}{p_c^{b,k+1}(w_1)}\right)}{\log\!\left(\sum_{j=1}^{k} q_c^{b,j}(w_1)\right)},
\qquad
CLF_n^{(k)}(w_1) := \frac{\log\!\left(\frac{p_n^{a,k+1}(w_1)}{p_n^{a,1}(w_1)}\right)}{\log\!\left(\sum_{j=1}^{k} q_n^{a,j}(w_1)\right)}.
$$
The decision rule is
$$
\chi^{\mathcal{B}(w)} = \mathbb{I}\left\{ CLF_n(w) < CLF_c(w) \right\},
$$
so lower CLF means a more favorable reference leg. The factor is dimensionless because it divides a log-price ratio by a log-depth quantity.

The paper’s intuition is microstructural. A book with large price discontinuities or thin cumulative depth over the first few levels is more execution-risky. The February NIFTY futures example highlights a bid-side drop from INR $17450$ at level 2 to INR $17401.1$ at level 3, a decline of INR $48.9$, illustrating why deeper-book discontinuities matter. The empirical agreement scores against the hindsight-optimal reference choice on Feb 7, 2022 are $0.3796$ for $CLF_1$, $0.4754$ for $CLF_2$, $0.5919$ for $CLF_3$, and $0.6358$ for $CLF_4$, versus $0.7335$ for the Hawkes arrival-ratio rule. The monotone ranking $CLF_4 > CLF_3 > CLF_2 > CLF_1$ indicates that deeper-book information materially improves the static signal, while the Hawkes rule remains more informative for the market-optimal choice [2507.05749].

## 3. Liquidity-horizon aggregation in Basel/FRTB

"On the Basel Liquidity Formula for Elliptical Distributions" states that it does not use the term “Composite Liquidity Factor (CLF)” explicitly, but it is directly about the mathematical structure practitioners often mean by a CLF in regulatory market risk: a liquidity-horizon aggregation rule that combines risk contributions from buckets with different liquidation horizons into a single capital measure [1803.07590]. The Basel liquidity formula is written as
$$
ES_\alpha(L) = \sqrt{ \sum_{k=1}^n \frac{h_k-h_{k-1}}{h_1} \left(ES_\alpha(L^{(k)})\right)^2 }.
$$
Here $h_k$ are liquidity horizons measured in units of the base horizon $h_1$, and the nested losses $L^{(k)}$ correspond to risk factors in bucket $k$ and all less liquid buckets.

Two structural features are central. First, square-root-of-time scaling appears because under white noise and linear P\&L, variance is additive across non-overlapping time slabs. Second, the aggregation is nested: the object analogous to $ES_T(P,j)$ is not the ES of only bucket $j$, but the ES of all factors whose liquidity horizon is at least $LH_j$, reflecting the persistence of less liquid exposures through earlier liquidation subintervals. In this sense, the formula is already a composite liquidity operator.

The paper’s main contribution is an elliptical generalization. Under iid elliptical risk-factor changes,
$$
ES_\alpha(L) =
\frac{c_{\alpha,\psi_L}}{c_{\alpha,\psi_1}}
\sqrt{ \sum_{k=1}^n \frac{h_k-h_{k-1}}{h_1} \left(ES_\alpha(L^{(k)})\right)^2 }.
$$
The correction ratio
$$
r_\alpha := \frac{c_{\alpha,\psi_L}}{c_{\alpha,\psi_1}}
$$
equals $1$ in the Gaussian case and may differ from $1$ in non-Gaussian elliptical settings. The paper reports that for symmetric generalized hyperbolic families, including Student $t$, VG, Hyp, and NIG, the Basel formula is typically conservative, with all non-Gaussian cases showing $r_\alpha<1$. In the example emphasized by the paper, for $\alpha=0.975$, NIG, $n=5$ liquidity buckets, and zero correlation, $r_{0.975}=0.837$, interpreted as Basel overstatement of capital by about $19.4\%$ [1803.07590].

A CLF in this literature is therefore not an execution proxy but an aggregation structure over horizon buckets. A common misconception is to treat it as a generic liquidity score; the paper instead shows that it is a specific expected-shortfall operator whose exactness depends on linear P\&L, iid Gaussian or iid elliptical shocks, and the nested-liquidity-horizon liquidation structure.

## 4. Basket, ETF, and execution-cost CLF from LIX

"Convenient liquidity measure for Financial markets" introduces the Liquidity Index,
$$
LIX = \log_{10}\!\left( \frac{V_T P_{\text{Close}}}{P_{\text{High},T} - P_{\text{Low},T}} \right),
$$
and develops a liquidity algebra that is highly relevant for CLF construction [1412.5072]. The paper’s central point is that aggregation should not be performed directly in $LIX$ space. Instead, the natural aggregation object is inverse raw-liquidity, because execution cost per unit invested is proportional to $10^{-LIX}$ under the paper’s execution-cost model.

The basket formula is
$$
\frac{1}{10^{LIX_{\text{Bskt}}}} = \sum_{i=1}^{N}\frac{\beta_i}{10^{LIX_i}},
\qquad
LIX_{\text{Bskt}} = -\log_{10}\!\left( \sum_{i=1}^{N}\frac{\beta_i}{10^{LIX_i} \right).
$$
This is a weighted harmonic-style aggregation in inverse-liquidity space, with $\beta_i$ the monetary portfolio weights. The same algebra extends to ETFs, where effective ETF liquidity is the sum of basket liquidity and ETF-share liquidity in non-logged space:
$$
LIX_{\text{Bskt,ETF}} =
\log_{10}\!\left(
\frac{1}{\sum_{i=1}^{N}\dfrac{\beta_i}{10^{LIX_i}} + 10^{LIX_{\text{ETF}}}}
\right).
$$
The paper explicitly argues that ETF liquidity can exceed visible secondary-market liquidity because creation/redemption exposes underlying-basket liquidity in addition to on-screen ETF-share trading.

The paper also provides intraday and order-book-based variants. The instantaneous measure
$$
LIXI_\tau = \log_{10}\!\left( \frac{(V_{\text{Bid}}+V_{\text{Ask}})P_{\text{Mid}}}{\overline{P}_{\text{Ask}}-\overline{P}_{\text{Bid}}} \right)
$$
links LIX to weighted spread and cumulative depth across $N$ book levels. For FTSE 100 stocks, average $LIXI$ versus one-month average $LIX$ yields a regression with slope $1.0147$, intercept $-0.0199$, and $R^2 = 0.9596$, which is the paper’s main empirical support for the order-book proxy.

Within this framework, a CLF is most faithfully defined by aggregation in inverse-liquidity or execution-cost space:
$$
CLF = -\log_{10}\!\left(\sum_i w_i 10^{-LIX_i}\right).
$$
The paper supports this construction directly for baskets and, with the ETF adjustment above, for composite instruments as well [1412.5072].

## 5. Concentrated-liquidity AMMs and DEX liquidity decomposition

In concentrated-liquidity AMMs, CLF-like constructions are inherently multi-dimensional. "Decentralised Finance and Automated Market Making: Predictable Loss and Optimal Liquidity Provision" decomposes LP wealth into fee income, position value, rebalancing costs, and concentration effects, and shows that the drift term of wealth is governed by
$$
\frac{1}{\delta_t}\left(4\pi_t-\frac{\sigma^2}{2}\right) +\mu_t\rho_t -\frac{\gamma}{\delta_t^2},
$$
with fee intensity $4\pi_t/\delta_t$, predictable loss $\sigma^2/(2\delta_t)$, concentration-risk penalty $\gamma/\delta_t^2$, and directional skew $\mu_t\rho_t$ [2309.08431]. The paper explicitly states that it does not define a “Composite Liquidity Factor,” but it provides a direct structural mapping for one. In the symmetric case $\mu=0$, the profitability condition reduces to
$$
4\pi_t-\frac{\sigma^2}{2}\ge \varepsilon>0,
$$
and the paper interprets this as fee profitability exceeding predictable loss. Empirically, historical LP operations in ETH/USDC averaged position value performance of $-1.64\%$ per operation, fee income of $0.155\%$, and total performance of $-1.49\%$, excluding gas fees; the model-based strategy produced $0.0047\%$ one-minute total performance excluding gas, while average gas cost per operation was $84.8$ USD and profitability on average required initial wealth above about $1.8\times 10^6$ USD [2309.08431].

"What Drives Liquidity on Decentralized Exchanges? Evidence from the Uniswap Protocol" supplies an explicit two-channel decomposition of DEX liquidity [2410.19107]. The main quoted-depth proxy is
$$
\textsf{v3Spread}_t^{pool} =
10^4\times\left(\frac{1}{4}\sum_{i\in[4]}
\frac{\textsf{Ask}_{pool,t,i}-\textsf{Bid}_{pool,t,i}}
{0.5\times(\textsf{Ask}_{pool,t,i}+\textsf{Bid}_{pool,t,i})}
-2\times\textsf{FeeTier}\right),
$$
while the same-TVL counterfactual constant-product benchmark is
$$
\textsf{CFv2Spread} = 10^4\times\frac{4p_X}{\textsf{TVL}\times\Delta_X},
$$
and concentration efficiency is measured by
$$
\textsf{v3S/CFv2S} = \frac{\textsf{v3Spread}}{\textsf{CFv2Spread}}.
$$
The log identity
$$
\log(\textsf{v3Spread})=
\log(\textsf{CFv2Spread})+\log(\textsf{v3S/CFv2S})
$$
separates capital amount from capital placement. The headline empirical finding is that gas prices, returns, and volatility affect liquidity primarily through concentration; internalization by private liquidity sources affects TVL but not overall market depth; and fee revenue, markout, and DEX competition act through both channels [2410.19107].

"Liquidity provision with $\tau$-reset strategies: a dynamic historical liquidity approach" adds a CLMM-native state representation based on bucketized active ranges, $\tau$-reset dynamics, and historical liquidity reconstruction without historical per-range liquidity snapshots [2505.15338]. The paper partitions the price axis into buckets $\beta=\{B_1,\ldots,B_N\}$ and defines the LP-liquid set
$$
\beta^{\mathrm{LP}}=\{B_{M-\tau},\ldots,B_{M+\tau}\}.
$$
Liquidity inside a range is tied to reserves through the Uniswap v3 invariant
$$
\bigl(x_i+\tfrac{L_i}{\sqrt{p_{b_i}}}\bigr)\,
\bigl(y_i+L_i\sqrt{p_{a_i}}\bigr)=L_i^2,
$$
and historical pool liquidity is approximated by Gaussian bucket weights
$$
\bm\alpha_{e_i}^{\Sigma}=f_{\mathcal G}(\mu_{e_i},\sigma_{e_i}),
$$
calibrated via fee matching. The paper’s empirical conclusion is that tighter concentration can increase fee density but becomes harmful when impermanent loss, downside moves, and gas or reset costs overwhelm the extra fees, and that moderate $\tau$ values, especially $\tau=5$ and $\tau=10$, tend to perform best across pools [2505.15338].

Taken together, these papers support a DeFi-specific understanding of CLF as a joint summary of depth, concentration, fee monetization, reset fragility, and implementation frictions. They also show that in concentrated-liquidity AMMs, TVL alone is an insufficient liquidity statistic.

## 6. Dynamic-factor and systemic formulations

"Dynamics of Liquidity Surfaces in Uniswap v3" provides a factor-model route to CLF construction by treating the liquidity object itself as a surface [2509.05013]. The paper studies
$$
y_t(x)=\log L_t(x)
$$
on relative standardized ticks $x\in[-1,1]$, and uses the decomposition
$$
y_t(x) = m(x) + \sum_{k=1}^K \beta_{t,k} u_k(x) + r_t^{(K)}(x).
$$
For 5 bps pools, the empirical eigenspaces align closely with low-order Legendre polynomials, yielding fixed “level / slope / curvature / higher-order” interpretations. The paper’s practical choice is $K=5$; for ARB5 and ETH5, $K=5$ usually explains $90$–$95\%$ of variation, whereas ETH30 is less stable and often needs more components. For ARB5, the first five PCA PVEs in three windows are $(0.702,0.184,0.042,0.024,0.009)$, $(0.612,0.274,0.043,0.018,0.007)$, and $(0.653,0.260,0.034,0.014,0.008)$. The paper therefore supports a few-factor CLF, especially a Legendre-based one that separates total liquidity level, imbalance, and concentration near the current price [2509.05013].

"Liquidity-Based Audit of Algorithmic Trading Strategies" offers a distinct systemic and trajectory-based formulation [2606.29018]. Its core observable statistic is
$$
\operatorname{Cov}\bigl(c_t,\hat{\pi}_t(c_t)\bigr),
$$
estimated by
$$
\hat{C}_T = \sum_{t=1}^{T}(c_t-\bar{c})^\top(\hat{\pi}_t(c_t)-\bar{\pi}).
$$
The sign classifies a linear strategy as a net liquidity consumer or provider. Under a scalar AR(1) cost process, the paper shows that the correction per period equals
$$
\alpha s^2,
$$
where $s$ is the Roll implied half-spread, making the correction a direct proxy for prevailing illiquidity. At the aggregate level, the market-wide liquidity imbalance is
$$
\mathcal L_t = \sum_{i=1}^N \operatorname{Cov}(c_t^i,\hat\pi_t^i(c_t^i)),
$$
and crowding enters through the fragility multiplier
$$
M(N,\rho_{\mathrm{sys}})=1+(N-1)\rho_{\mathrm{sys}}.
$$
The empirical CRSP calibration covers 2016–2025 and shows that implied spreads roughly triple from $0.90\%$ to $2.52\%$ during the COVID episode, while the 2022 rate-shock period is characterized by near-zero $\hat\rho$ and low Roll-style implied illiquidity. In this literature, a CLF is naturally a hybrid of signed covariance-based liquidity demand, Roll-spread illiquidity, and systemic crowding [2606.29018].

These dynamic formulations differ from the earlier static or algebraic ones. A CLF may therefore be a low-dimensional latent state, a signed market-balance statistic, or a systemic fragility index, depending on whether the object of interest is pool-shape evolution, strategy behavior, or aggregate liquidity pressure.

## 7. Limitations, ambiguities, and common misconceptions

A first limitation is terminological. Several papers explicitly state that they do not use the term “Composite Liquidity Factor” itself, even when they analyze structures that practitioners map into that label [1803.07590] [2309.08431] [2606.29018]. A CLF is therefore not a universally standardized object.

A second limitation is dependence on modeling domain. The HFT CLF in [2507.05749] is a static, one-sided LOB-snapshot score; it does not model event clustering, cross-excitation, or flow memory. The paper also leaves aggregation partly ambiguous, referring both to majority vote over the previous $1000$ ms and to EMA smoothing of tick-level CLF values. It does not specify winsorization, z-scoring, explicit cross-contract normalization beyond the log transformation, or an assumed marketable order size $Q$. As a result, the factor is a heuristic ranking signal rather than a calibrated slippage forecast.

A third limitation concerns aggregation and exactness. In the Basel/FRTB setting, the liquidity-horizon formula is exact only under linear P\&L and iid Gaussian shocks, and becomes an elliptical generalization with a correction factor under heavier tails [1803.07590]. In the LIX setting, aggregation is exact only within the paper’s cost-model assumptions, including additive execution costs, common horizon, and ideal slicing or resilience [1412.5072]. A simple arithmetic average of logged liquidity scores is not supported by that algebra.

A fourth limitation is that concentration is not unconditionally beneficial. The AMM literature repeatedly shows a depth–fragility trade-off: narrower ranges improve local fee intensity and near-price depth but worsen predictable loss, concentration risk, out-of-range risk, gas sensitivity, or reset frequency [2309.08431] [2505.15338] [2410.19107]. Similarly, in liquidity-surface factor models, a scalar factor based only on the first PCA score may be statistically convenient but economically unstable because PC1 can be tilt-like rather than a pure level factor [2509.05013].

A fifth limitation is scope. The transaction-cost and liquidation framework in "Liquidity Stress Testing in Asset Management -- Part 2. Modeling the Asset Liquidity Risk" is directly relevant to composite liquidity measurement, but it does not define a CLF explicitly [2105.08377]. It instead provides a multi-metric architecture built from
$$
c(x)=s+\pi(x),
\qquad
\mathcal{LR}(q;h),
\qquad
\mathcal{LT}(q;p),
\qquad
\mathcal{LS}(q)=1-\mathcal{LR}(q;1),
$$
and regime-dependent stressed parameters $(s^{\mathrm{stress}},\sigma^{\mathrm{stress}},v^{\mathrm{stress}})$. This suggests a composite liquidity indicator for stress testing, but only as an extrapolation from the paper’s multi-dimensional framework.

The principal misconception, therefore, is to treat “Composite Liquidity Factor” as if it named one settled formula. The literature instead supports a narrower conclusion: CLF is a domain-specific composite representation of liquidity, and the relevant composition rule depends on whether the objective is reference-leg selection, risk-capital aggregation, basket execution-cost measurement, LP profitability and concentration management, dynamic AMM liquidity-state compression, systemic strategy audit, or liquidity stress testing.

Source: https://www.emergentmind.com/topics/composite-liquidity-factor-clf