Why it loses
Three costs stack: the spread you pay to enter, the premium you pay for convexity, and the compounding that turns a positive average into a negative outcome. Every figure here was measured, not assumed.
1. One trade, taken apart
The cheapest listed call on MARA, 5 days to expiry. Quoted 0.01 bid / 0.02 ask.
You pay the ask and you can only sell at the bid. On a one-cent-wide market where a penny is the premium, that is 50% gone the instant you click, before the stock has moved at all. MARA has to rise enough to double the option just to get you back to flat.
2. The cost ladder
Round-trip cost as a share of what you're buying. Log scale — each gridline is 10×.
SPY costs 0.26 basis points to round-trip. The cheapest listed option costs 13,300 — a factor of 51,000. And the ladder is monotonic: every step toward cheaper, more convex, more exciting is a step toward paying more of the instrument away on entry.
3. Now put a real edge against it
Measured edge per trade versus what the trade costs. Both in basis points.
The overnight-bounce signal is real — t = +4.07, 2,863 trades, survivorship-free. It is also 71 basis points. Any instrument above the $2-stock rung on the ladder costs more than the edge is worth. That is not a coincidence: an edge larger than its own spread gets traded down to its spread by someone with better execution.
4. And then compounding takes the rest
The same strategy. Same trades. The difference is that an average is additive and an account is multiplicative.
Win rate 68%, mean +0.71%, and the p10 is −7.6% with a worst case of −59%. Compounding weights the losses more heavily than the average does: a −50% needs a +100% to undo. With only ~3.6 positions a night, all crashing together, the left tail runs the account.
This is the same equation as the leveraged ETFs. A fund at leverage m loses (m²−m)/2 × σ² per year to variance. A strategy loses the same way: its geometric return sits below its arithmetic mean by roughly σ²/2.
Direction scales linearly with how much you bet. Cost scales with the square. That one fact is behind every negative result in this project — the 3x funds, the martingale, the dip rules, and this.
Table view
Spreads: Alpaca SIP quotes (equities) and indicative option quotes, measured 2024–2026. Edges: survivorship-free Massive panel, 2021-10 to 2026-10, split artifacts stripped. Portfolio simulation: equal split across each night's signals, 15 bps round trip.