Economics

Where this product costs money, stated as plainly as where it makes it. The premium is roughly fair; the spread is not, and rolling it every window is what turns a fair price into a ruinous one.

On this page

At-the-money is the expensive end
The spread is the structural cost
Frequency multiplies it
Why 4h and 24h
The sample caveat, at full length

At-the-money is the expensive end#

Because the strike is the window’s opening price, the book is always at-the-money. There is no cheaper out-of-the-money strike to buy instead — see the strike finding — so you are forced to pay for the very likely small moves as well as the tail you actually want covered. This is the most expensive cover this instrument offers, and it is the only cover it offers.

The spread is the structural cost#

The premium itself is roughly fair in expectation: q is the market’s own probability of the Down outcome, so on average you pay about what the risk is worth. The cost that does not average away is the spread, and it is paid on every single roll.

Observed spreads on testnet run 2.2% to 15.4% of mid — wide, because the books are thin. Every roll crosses that spread again.

Frequency multiplies it#

Rolling a 250 bps make-whole point every window, priced against the live books:

Interval	Windows / day	Spread cost / day	Spread cost / year
60 s	1440	939% – 9720%	astronomically ruinous
900 s	96	~5%	~1900%
3600 s	24	15% – 25%	5500% – 9200%
14400 s (4 h)	6	0.2% – 0.3%	79% – 110%
86400 s (24 h)	1	~0.1%	19% – 21%

The 1 h row is bad in this snapshot only because q happened to sit near 0.975 on those two markets. The window length is fine; the price was not — which is precisely why the premium ceiling is a per-window check rather than a configuration constant.

Why 4h and 24h#

The product defaults to the four-hour and twenty-four-hour windows, and sixty seconds is unusable. Not marginal — unusable. At 1,440 rolls a day the spread alone consumes multiples of the position every year, whatever happens to the price.

Treat the absolute figures as structure, not forecastTestnet books are thin and erratic. What transfers to a real venue is the structure — premium proportional to x*/(1−q), multiplied by windows per day, with the spread as the real drag — rather than these particular percentages.

The sample caveat, at full length#

The recorded history shows 128 settled positions with a positive net. That figure is on the dashboard and on the landing page, and it should be read with this attached:

A sample, not a resultThose 128 positions are short windows on a thin testnet book — 40 one-minute, 71 five-minute, 13 fifteen-minute and 4 one-hour, recorded over 1–11 September 2026. The economics on this page say plainly that rolling cover that often is ruinous over any real horizon — the table above puts the spread alone at thousands of percent a year for every one of those intervals, which is why the product defaults to four-hour and twenty-four-hour windows. A favourable run of 128 does not contradict that. It is what a small sample looks like, and reading it as a return would be reading it wrong.

This is also why the dashboard shows cumulative premium paid against protection actually used, rather than the winners alone. A reader will do that arithmetic; better that we did it first.