6.3.5 Failure Modes & Practice

Basic Prerequisites: 04 · Quote Skewing.

---

1. Intuition & Practical Objective

The inventory problem is mathematically clean and operationally unforgiving. This page names the ways inventory control fails in money terms, tied to first principles. The objective is not cynicism - it is knowing exactly where the position is at risk so it can be bounded.

The three failures, in one line each:

1. Inventory limits are binding in the tail. The skew encourages flattening but never forbids a position; an uncontrolled maker breaches the position cap and is force-liquidated at adverse prices.
2. Forced liquidation converts paper risk into realized loss. Once $|I|$ hits the risk limit, the desk must pay the spread to exit - at exactly the worst time (a trend). In our simulation $92\%$ of uncontrolled paths breach the cap in a trending market, versus $0.2\%$ with a skew.
3. Risk limits interact with the trend. A cap is not free: it can force you out right before the trend turns, so the sizing of the cap is itself a risk decision.

The one-line takeaway. “An unskewed maker in a trending market accumulates inventory until the position limit force-liquidates him at adverse prices; quote skewing cuts the breach rate from $92\%$ to $0.2\%$ and collapses the P&L standard deviation - the inventory limit is the guarantee, the skew is the cause of never needing it.”

---

2. Mathematical Ground Truth & Derivations

The breach condition. Inventory evolves as a random walk with reversion speed $\kappa$ set by the skew strength $\alpha =\gamma {\sigma }^2\tau$ and fill elasticity $k$. For weak skew ($\alpha \to 0$) the position is a pure random walk and the probability of ever reaching $\pm Q$ by time $T$ is

$$\mathbb{P}(\underset{0\le t\le T}{\max }|q_t|\ge Q)\approx 4\Phi (-\frac{Q}{{\sigma }_q\sqrt{T}})$$

(a reflected-walk / first-passage bound, ${\sigma }_q^2=$ per-step inventory variance). Stronger skew shrinks ${\sigma }_q$, and the breach probability collapses. Our simulation confirms this monotone dependence.

The forced-liquidation cost. When $|q|=Q$ the desk liquidates at market, paying the half-spread per share:

$$\text{forced-loss per share}\approx \frac{1}{2}{s}_{\text{market}}+\text{adverse move realized at that instant}.$$

The expected loss of a forced liquidation is the adverse price move $\mathbb{E}[\Delta S\mid \text{trend},\text{breach time}]$ - which, in a trending market, is large because breaches happen when the trend is running against the position. This is why the tail (P&L std) explodes under a cap without skew.

Risk limits are a variance budget. A desk with position limit $Q$ and per-period variance ${\sigma }_q^2$ has inventory-risk VaR $\approx z_{\beta }{\sigma }_q\sqrt{T}$ (a $Q$-period variance budget). Choosing $Q$ is choosing how much of that tail to accept; skewing shrinks ${\sigma }_q$ so a given $Q$ is breached far less often.

---

3. Computational Implementation - the trending-market failure

We simulate a trending, volatile market ($\mu =+3$, $\sigma =3$) with position cap $Q=6$, and sweep the skew strength $\alpha$ (0 = no control). A breach force-liquidates the position at market. Reported: mean P&L, P&L std, and the fraction of paths that ever breached the cap.

This is the failure made concrete. With no inventory control ($\alpha =0$) $91.8\%$ of paths breach the position cap in the trending market, and the forced-liquidating maker carries P&L std $9.58$ (driven by the realized losses at breach). A modest skew ($\alpha =0.20$) cuts the breach rate to $0.22\%$ and the P&L std to $6.57$, while giving up only $\sim$ $2 of mean ($55.59$ vs $57.73$). Stronger skew ($\alpha =0.50$, $1.0$) drives breaches to zero and P&L std down to $5.21$ and $3.67$ - at a rising cost in mean ($50.12$, $34.96$) because over-skewing forgoes spread flow. The cap bounds the position; the skew decides how often the cap is ever touched.

---

4. Failure Modes & First-Principles Breakdowns (numbered)

1. Uncontrolled inventory = guaranteed breach in a trend. A random-walking position in a trending market reaches $\pm Q$ with near-certainty ($91.8\%$ here). The first line of defence is not a tighter cap - it is a skew strong enough that the cap is rarely binding.
2. Forced liquidation realizes paper losses at the worst moment. Breach happens when the trend is against the position; liquidating then converts mark-to-market risk into realized loss. This is why P&L std explodes ($9.58$) under a cap without control.
3. The cap itself is a risk decision. A tight $Q$ forces exits that a wider $Q$ (or stronger skew) would have ridden out. Sizing $Q$ from a risk budget - not from “make it small” - is the correct first-principles answer.
4. Overnight / gap risk bypasses the skew. Inventory held into the close gaps over a jump; intraday skewing cannot protect a position when the market is closed. Flatten (or hedge) before the close.
5. Adverse selection is orthogonal. The inventory limit controls the position; informed flow that picks off stale quotes needs an information-side defence (Toxic Order Flow & VPIN).

---

5. References

• Ho & Stoll (1981), Optimal dealer pricing under transactions and return uncertainty, JFE 9(1)
• Garman (1976), Market microstructure, JFE 3(3)
• Guéant, Lehalle & Fernandez-Tapia (2013), Dealing with the inventory risk, Math. & Financial Econ. 7(4)
• Hendershott & Menkveld (2014), Price pressures, Journal of Financial Economics 114(3)

---

6. Connected Graph Bridges

• Back: 04 · Quote Skewing · Index Hub
• Forward: 06 · Advanced Extensions
• Sibling: Toxic Order Flow & VPIN · VaR & Expected Shortfall · Liquidity Risk & Margin Spirals