The Real Cost of a Trade — Why Gross APR Is a Tourist Trap

Of 199 arb opportunities on the board right now, 43 — twenty-two percent — have positive gross spread and negative net APR. That's one in five "opportunities" that would pay you to hold them, then charge you more than the payment in fees. Gross APR is a tourist trap, and here's the math they don't want you to do.

Every screenshot on twitter quotes gross. Every landing page on every competing site quotes gross. Every "200% APR opportunity" you've ever scrolled past is gross. Strip the fees, strip the slippage, amortise over the hold you'll actually run, and most of those numbers evaporate. A lot of them go negative.

// THE FEES YOU'RE ACTUALLY PAYING

Every venue charges a taker fee on market orders. Every market order crosses the spread, which is slippage. A delta-neutral pair involves four taker bills — entry and exit on both legs — plus two slippage crossings. The Forge's cost table, pulled from backend/services/costs.py:

| Venue | Taker bps | Maker bps | Slippage bps | Funding interval | |--------------|-----------|-----------|--------------|------------------| | Hyperliquid | 3.5 | 1.0 | 3 | 1h | | dYdX | 5.0 | 2.0 | 5 | 1h | | GMX | 6.0 | 6.0 | 4 | 1h (continuous) | | Paradex | 3.0 | 0.5 | 5 | 8h | | Aster | 3.5 | 1.0 | 5 | 8h | | Lighter | 0 | 0 | 6 | 1h | | Vertex | 2.0 | 0 | 4 | 1h | | Bluefin | 7.5 | 1.0 | 8 | 1h | | Extended | 5.0 | 1.0 | 5 | 1h | | Backpack | 4.0 | 1.0 | 5 | 1h | | Pacifica | 5.0 | 1.0 | 6 | 1h | | Ethereal | 3.0 | 0 | 8 | 1h |

Look at Lighter. Zero taker, zero maker, six bps of slippage. That's not a gift — thin books eat more than thick ones, and Lighter's books are thin. The "zero fees" pitch is marketing. The slippage is where they get paid. You're still bleeding.

Look at GMX. Six taker, six maker — no maker discount, because GMX prices off a single pool instead of an order book. Every trade costs the same. If a venue's "fee" stays flat between maker and taker, that's not a fee schedule, that's a toll.

// OPEN THE COST LEDGER

// THE WORKED EXAMPLE

Top BTC arb on the board at write time: long Extended, short GMX. Gross spread of 29.43% APR. That's the number a screenshot trader would post.

Now do the math.

Extended taker: 5.0 bps × 2 (entry + exit) = 10 bps. Extended slippage: 5 bps. Leg cost: 15 bps.

GMX taker: 6.0 bps × 2 = 12 bps. GMX slippage: 4 bps. Leg cost: 16 bps.

Round trip: 31 bps. The scanner confirms it — round_trip_cost_bps: 31.0.

Amortised over a 7-day hold (the Forge's benchmark), 31 bps becomes 16.16% APR in break-even cost. The gross was 29.43%. Subtract. Net APR is 13.26%.

More than half of the "29% opportunity" vanishes. Not because the funding rate lied — it didn't. Because the round-trip fees, amortised honestly, chew through it.

The Forge puts the waterfall inside the strategy modal. Click any card, read the breakdown, watch the sage green bar carve into rust as every component takes its cut. The net bar at the bottom is what's left. If it's thin or negative, the play is a trap.

// THE AMORTISATION LIE

This is where most traders die. Same trade, same 31 bps of fees, different hold assumptions:

Three numbers. Same trade. Radically different conclusions.

On a 1-day hold, you need the funding spread to annualise at 113% just to cover the fees. For BTC with a 29% gross, that's a negative-84% trade. Day-trading funding arb is a mathematical suicide pact. The twitter tourists who screenshot their "50% APR play" and close it in 18 hours are paying themselves to lose money.

On a 30-day hold, the same 31 bps spread over thirty days of funding settlements shrinks the break-even to 3.77%. The 29% gross now looks like a 25.66% net. Juicy. Almost too juicy — the kind of number newsletter grifters put in their pitch deck. Thirty days is also the window where funding rates flip, venues die, and the trade you entered becomes a trade you can't exit without eating a different set of fees.

The Forge picks seven days. It's the honest middle. Long enough to amortise fees to realism, short enough that the spread is still recognisable to the pair you entered. Every backtest on the site uses 7d. Every net APR on every card uses 7d. If you're running a different hold window, do the math yourself — the strategy modal's profit calculator lets you pick 7d, 30d, or 90d and shows what the P&L looks like across each.

Hold windows matter more than most people understand. Most people don't understand them.

// RUN THE BACKTEST

// WHERE THE TRAPS LIVE

By gross APR bucket, on the live board right now:

| Gross APR | Count | Net < half gross | Net ≤ 0 | |----------------|-------|------------------|---------| | 0–10% | 23 | 23 | 23 | | 10–25% | 31 | 28 | 20 | | 25–50% | 27 | 3 | 0 | | 50–100% | 56 | 0 | 0 | | 100%+ | 62 | 0 | 0 |

Read that table. Every play with gross under 10% is a fee trap. Nearly every play in the 10–25% bucket is either a trap or bleeding half its gross to costs. The 25–50% bucket is where profitability starts to stabilise. Above 50% gross, fees are noise — you keep almost all of it.

The implication is brutal and specific: low-yield plays aren't "safe" plays. They're the MOST dangerous plays, because they look reasonable, look steady, look like something a prudent trader would pick. They're also where the fee structure eats you alive.

If the dashboard is sorted by gross APR, ignore the bottom half. If you sort by net APR — which is the Forge's default — the bottom half falls off naturally. The scanner is already doing the math. Let it.

// WHAT COSTS DO TO A STRATEGY

Three rules. Tattoo them somewhere:

One — fees don't shrink with position size. Your 31 bps round trip is 31 bps whether you size $1k or $1M. The absolute dollars scale, but the percentage is fixed. Anyone promising "cheaper per-unit fees at size" is either wrong or selling something.

Two — maker is fiction for funding arb. You can't sit as a maker on both venues simultaneously and guarantee fills on both. Leg risk eats any maker discount you thought you were saving. Take the taker hit. Enter in the same minute.

Three — slippage is proportional to book thinness, not to price. A 0.5-cent book on a $10 token is 50 bps slippage. A $5 book on a $1000 token is 5 bps. Sizing a funding arb on a microcap perp with a 50 bps book turns a 20% gross into a −30% net before you even touch the funding. Read the open interest column. Read the book. Read the cost waterfall in the modal. Then size.

// SORT BY NET APR

// THE CLOSING DISPATCH

Gross APR is a headline. Net APR is a paycheck. The Forge shows both because you deserve to see the lie before it gets you. One in five "opportunities" on any funding scanner on the internet is a pure fee trap. The other four hide the fee trap inside a hold-window assumption nobody tells you to question.

Do the math. Run the 7-day amortisation. Read the waterfall. If the net bar is thin, keep scrolling.

The warlords carve with the knife. The tourists bleed on the edge.

// ENTER THE FORGE →