"The strategy roughly breaks even, and subsidies push it positive" is the most common plan in this industry. It contains one arithmetic error: treating the subsidy as an addend. Subsidies are a multiplier — they improve a positive strategy and cannot rescue a negative one. Here is that shown with my own money.
| Per trade | |
|---|---|
| Taker fee (p≈0.5) | 1.75pp |
| Taker ladder (Gold) | 0.32pp |
| Referral | ≈ 0.18pp |
| Both maxed | 0.17–0.18pp |
Why not the sumbecause the ladder erodes referral's base (referral pays on the venue's net receipts) — the two partly overlap.
Because rebates only exist after you have paid a fee. They generate no new income; they return a fraction of a cost already incurred. So:
| If your gross edge is | No rebate | Rebates maxed | Verdict |
|---|---|---|---|
| +3.0pp | +1.25pp | +1.50pp | already profitable; 20% better |
| +1.8pp | +0.05pp | +0.30pp | break-even becomes a small profit |
| +1.5pp | −0.25pp | +0.00pp | marginal; just rescued |
| +0.8pp | −0.95pp | −0.70pp | far short; beyond rescue |
The pattern is clear: rebates can only save a strategy that is short by a hair. Their reach is a band about a quarter of a percentage point wide. A larger shortfall is out of range.
I built a line whose launch note said, in its own words: the model itself was expected to be "slightly negative to roughly break-even", and platform rebates would carry it into profit. Its entire case for profitability rested on the subsidy.
| Result | |
|---|---|
| Ran for | about 41 hours, roughly 496 windows |
| Net | −$496.51 |
| Observed gross edge | +1.56pp (research basis) / +0.78pp (settlement basis) |
| Cost line to clear | 2.25pp |
| Shortfall | 0.69pp / 1.47pp |
| Rebates could cover | about 0.30pp |
| Rebates actually received | $0 |
Two lessons stacked: ① even fully paid, the subsidy would not have closed the gap (0.30 against 0.69, let alone 1.47); ② and none of it arrived. Full ledger in the autopsy.
The criterion I wrote for that line set its cost threshold as:
1.75 (fee) − 0.30 (rebate stack) = 1.45pp
Reasonable-looking — except the 0.30pp assumed a particular tier, and not one cent of it materialised. Add it back and the real cost line is 1.75pp; with measured slippage, 2.55pp.
I had spent money I had not received while computing my own bar.
Rulea cost line may contain only items confirmed to occur. Writing expected income in as a cost deduction concedes ground before the start — and the error only ever lowers the bar, so it will never be caught by its own direction.
| Do | Do not |
|---|---|
| Verify the strategy is positive first | assume subsidies close the gap |
| Cost line holds only confirmed items | deduct expected rebates from cost |
| Spend two days making the subsidy column work | run the strategy and discover at the end that it is empty |
| Treat subsidies as a cushion | treat subsidies as the income |
| Size as if the subsidy could be zero tomorrow | model it as permanent |
The third row cost me the most: the strategy ran 41 hours and the subsidy column was never tested once — while testing it would have meant a few small orders and one day of waiting.
And the last row: a strategy fully dependent on a subsidy does not live or die by your judgement. It lives or dies by a product decision — rates, splits and eligible categories can all change, and the day they do it is negative.
Subsidies act on cost, not on revenue — so they rescue a strategy short by a hair, not one short by an order of magnitude. And before computing how much they could rescue, do something more basic: confirm they will actually arrive.
net = gross − cost×(1−rebate); recompute it yourself.