A Small Edge, Carefully Kept — and How It Connects to “Trading Big Bags”

In Trading Big Bags: Liquidity, Leverage, and the Architecture of Risk, I argued that size turns you from spectator into visible liquidity, and that survival depends on structure: distribution, alignment, and discipline. This post is the companion from the other end of the spectrum. Years ago (‘18-‘19), in a very specific BitMEX fee regime, I grew roughly $30 into about $2,000 in a week by posting passive liquidity and collecting maker rebates. The edge wasn’t in any single position — it was in repetition, running 60–120 trades a day, each one tiny, each one managed with care. It worked then because the context allowed a small, conditional edge to compound. This is not a victory lap; it’s a study in how microstructure, incentives, and humility can matter at small scale — and what breaks as you grow.

Humility first: Edges are conditional. The fee schedule, the microstructure, and my own small footprint all mattered. Change any of those and the edge can vanish.

What This Was (and Wasn’t)

• Was: A grind of 60–120 small trades a day, posting passive liquidity on both sides of the spread, earning maker rebates and capturing slivers of spread; inventory kept tight; quoting shut down in toxic or trending moments.

• Wasn’t: Big directional bets, outsized leverage, or “risk-free” carry. The real risks were adverse selection and inventory drifting the wrong way into a trend.

• Assumptions (historical): A negative maker fee (rebate) on fills, higher taker fee on crosses, a reliable post-only order flag, and an inverse, BTC-settled perpetual contract with predictable mechanics.

Where the PnL Came From (Small Edges Add Up)

• Maker rebate: Each maker fill paid a few bps of notional. Tiny per fill, meaningful across many cycles.

• Micro-spread capture: Selling slightly dearer and buying slightly cheaper than mid retained a sliver of the spread.

• Directional drift (sometimes): If inventory leaned with a short move, PnL improved; lean into a real trend and it hurt. Inventory control mattered more than drift windfalls.

• Funding/basis: Depending on regime and side, funding helped or hurt. The goal was to avoid letting funding bleed a small edge.

Back-of-Envelope (Edge Components)

One round-trip (one maker buy + one maker sell), ignoring direction:

$$\text{Rebate PnL} \approx 2 \times r_{\text{bps}} \times \text{notional}$$ $$\text{Spread PnL} \approx \delta_{\text{bps}} \times \text{notional}$$ $$E \approx (2r+\delta)\,N - S - C \pm F$$

• $E$ — Net edge per round-trip (USD)
• $r$ — Maker rebate per fill (bps)
• $\delta$ — Micro-spread capture (bps)
• $N$ — Notional traded (USD)
• $S$ — Adverse selection cost (USD)
• $C$ — Inventory carry / drift cost (USD)
• $F$ — Net funding impact (USD; positive if received, negative if paid)

Units: If $r$ and $\delta$ are in bps, the dollar contribution from the first term is $(2r+\delta)\times N \times 10^{-4}$.

Everything here lives and dies on adverse selection (getting lifted before price runs) and on hard limits to inventory.

Mechanics: Inverse Perp Math (Clarity, Not Nostalgia)

• Contracts / notional: $\text{contracts} = \text{BTC}_{\text{coll}} \times \text{entry} \times \text{leverage}$.
• Inverse PnL (BTC): $\text{PnL}_{\text{BTC}} = \text{contracts} \times \big(\frac{1}{\text{entry}} - \frac{1}{\text{exit}}\big)$.
• USD value: $\text{PnL}_{\text{USD}} = \text{PnL}_{\text{BTC}} \times \text{exit price}$.
• Rebates: Per maker fill ≈ $(\text{maker bps}) \times \text{filled notional}$.

Toy illustration: $100,000 notional per side; maker rebate 2.5 bps; one tick of spread capture ≈ 1.0 bps. If both orders fill as maker, Rebate ≈ $50 and Spread ≈ $10 — about $60 before drift/funding. The scale was “small bites, many repetitions, strict risk.” In practice, that meant dozens upon dozens of such round-trips each day — roughly 60–120 trades — where the law of large numbers let conditional edges show up in PnL.

The Playbook I Actually Used

• Post-only or don’t post: Guaranteed maker status; if a replace would cross, it cancelled.
• Small, many, both sides: Ladders of tiny orders above and below mid. One big order is easy prey; many small quotes are manageable and less visible.
• Inventory-aware quoting: Long inventory → tighten offers, widen bids (and vice versa) until flat.
• Hard inventory caps: Max long/short units scaled to recent volatility; hit the cap → stop quoting that side and work down risk.
• Kill-switches: Pull everything on trend breaks, spread collapse, funding flips against inventory, or venue wobble.
• Low effective leverage: Maker strategies die from liquidation, not fees. Sizing assumed routine swings.
• Cancel discipline: Throttled replaces to keep queue position and avoid turning makers into takers.

How This Ties Back to “Big Bags” (From Micro-Edge to Architecture)

In the big-bags piece, I wrote: “Survivability is the first alpha.” The same principle powered this small account. The difference is footprint. Small size can be invisible; big size becomes the liquidity. Here’s the mapping:

Small-Account Rule (This Post)	Big-Bag Architecture (Prior Post)
Post-only to earn rebates and avoid crossing	Participation caps and time-slicing to avoid impact and signaling
Hard inventory caps by recent vol	Gross/net exposure limits; VaR budgets; stress-tested drawdown guards
Kill-switch on trend/spread collapse	Circuit-breakers; cross-venue failover; session redundancy
Funding-aware inventory bias	Basis management; term structure alignment with mandate
Track edge components (rebate, spread, selection)	Attribution by venue/contract; slippage & impact analytics; funding PnL
Stay invisible (small clips, many levels)	Multi-venue distribution; lower footprint per book; queue management

Same principle, different scale: Structure — not bravado — is what lets you stay around long enough for edge to matter.

What Scales, What Breaks

• Scales conceptually: Kill-switches, inventory caps, and refusing toxic flow. These remain core, just implemented with bigger tooling (risk engines, venue routers).
• Breaks with size: Rebate-harvest at the inside quote. As notional grows, you become the queue; your fills get more toxic; the edge turns into impact.
• Context shift: If your mandate moves from BTC accumulation to USD returns/hedges, settlement alignment matters (inverse vs linear) — as discussed in the big-bags post.

When to Stop Scaling the Maker-Rebate Loop

• Participation too high: Your quotes consistently make up an outsized share of top-of-book depth.
• Edge attribution flips: Adverse selection + funding > rebates + spread over a meaningful window.
• Queue position degrades: More last-in-line fills; more toxic hits; rising cancel/replace churn.
• Venue risk rises: Outages, index anomalies, or fee tweaks begin to dominate PnL variance.

What I Took Away

• Small edges compound when protected: The craft is in refusing bad trades, not forcing good ones.
• Context governs edge: A modest process in the right microstructure can work. The same process in a different context can fail fast.
• From micro to macro: The big-bags toolkit is the small-account toolkit, scaled and made redundant — participation limits, distribution, alignment, discipline.

It wasn’t about turning $30 into $2,000. It was about learning how a tiny edge, guarded by discipline and repeated 60–120 times a day, can exist at all — and how quickly it disappears when discipline or context does.

Disclaimer: This is a personal account of a past approach in a specific environment. Markets change. This is not advice. If you experiment, paper-test against live order book data, assume the edge is thinner than it looks, and remember the first principle from the big-bags piece: survivability is the first alpha.