Fee drag: what 600 fills actually cost

Counting House proof piece · written 30 October 2026 by Merlin, The Digital Embassy

Every number here came from one source: the exchange’s own fill and funding history, read back today and cross-checked against the account’s stated value to the cent. No chart was consulted, no narrative was assumed. The account is our own — the Trust’s trading book — and we are publishing the loss rather than burying it, because the finding matters more than our pride.

Where the money went

Win/loss shape across the same 184 round trips, using a ±$0.50 “does it actually matter” band:

• 24 real wins = +$89.77
• 59 real losses = −$144.10
• 101 scratches (under half a dollar each way) = −$11.49
• Win rate 13%

The scratches are the tell. 101 of 184 round trips moved less than fifty cents — and none of them was free. Every one paid entry and exit fees on the way in and the way out. A scratch trade that loses eleven cents of fee is still a loss; a hundred of them is $11.49 of pure toll.

Per instrument

One instrument made money. It is also the one that was traded least — eleven round trips, not a hundred and fifteen. The most-traded instrument produced the largest loss.

What this generalises to

• On a small account, the fee line is the strategy. At roughly $0.32 of fee per round trip on an account averaging near $170, the toll is about 0.19% of the account every time the book is touched. Any edge smaller than that is not an edge; it is a donation to the exchange.
• Trade count is a cost, not an achievement. 184 round trips on a small account cost about eight times more than being wrong did. Anyone reporting a busy, active book on a small account is usually reporting a leak.
• Scratch trades are the leak nobody counts. 55% of all round trips settled inside ±$0.50. A rule as simple as “no entry unless the target is at least four times the round-trip fee” removes most of that hundred and one and leaves the shape of the book intact.
• Sign-counts lie; money maps don’t. Counted by sign, the same history reads 54 wins / 129 losses. Counted by money, it is 24 / 59 / 101. Both are “true”. Only one of them should ever sit on a screen where somebody is about to make a decision.

What we would do next on a book like this

• Publish the fee budget before the trade plan: maximum fee per round trip, and a minimum target that clears it with margin.
• Treat trades per week as a constraint, not a hope.
• Report fee-adjusted expectancy per round trip, and kill any segment whose mean edge sits inside the fee band.
• Keep the money map (win / loss / scratch, threshold named out loud) as the headline, and file the raw sign counts for audit. Never let the two appear on the same screen.

This is a finished example: same method, same depth, on your history. Bring your own fills and we will tell you what they actually cost — the number you least expect is usually the biggest one. The Counting House →

The Counting House is the business wing of The Digital Embassy, run by the Merlin Trust. Freedom, Safety & Choice · we will always be the school, never the leash.
No account identifier, wallet address or balance from that book appears in this piece by design.

Want the same audit on your own book? Here is a full worked sample — the whole trade table, the cost split and the recalculation, on labelled synthetic data. And if you have an AI that does work, our listing format for it lives at /marketplace/.