Why 88% Accuracy Still Loses 0.42 bps — free 3-page PDF

Free 3-page tear sheet: a lead-lag model that was right 87.9% of the time and lost 0.42 bps a trade, and the cost hurdle that explains it.

Digital file

At a glance

Delivery
3-page PDF tear sheet, instant download
Type
Digital file
Free
Get it free

3-page PDF tear sheet, instant download

About

A lead-lag model can call direction correctly four times in five and still lose money on every trade it takes. This three-page tear sheet shows the measurement where that happened and the arithmetic that explains it, both taken from the sample capture that ships with the Hyperliquid Multi-Venue Research Kit: 412,299 records over 119.7 seconds, public endpoints only, no account and no key. Page one is the measurement, with the hit rate by window and horizon and the mean forward return beside it; at a 1,000 ms window and a 500 ms horizon the sign was right 87.9% of the time and the mean forward move was minus 0.42 basis points, because a right call moved a little and a wrong one moved a lot. Page two is the hurdle: 1.5 bps maker, 4.5 bps taker, a one basis point buffer, and the break-even hit rate formula that says a 5 bps average win cannot pay a 7 bps round trip at any skill level, followed by displacement, where rank correlation runs from 0.62 to 0.84 and the strongest bucket still predicts only 1.5 bps against a 6 bps round trip. Page three opens the instrument: eight modules, forty tests, a two-minute BTC and SOL capture from four venues, and a 20-page guide whose every number is read from the run it describes. It measures; it does not trade, and it contains no signal service and no promise of returns. The kit is $100 on its own and $129 with the gold report in the Market-Structure Bundle.

Reviews

No reviews yet. Buyers can review after their order.

Questions about this file

Not answered here? Support

When do I get the file?

The download button appears on the order page the moment payment confirms, and the same link arrives by email. The link stays usable for a year, and a fresh one can be issued at any time from the order page.

Can I read part of it before paying?

Not the file itself, but enough to judge it. Each product page says what the file contains and what it does not, down to page and module counts, and every paid file has a free three-page tear sheet in front of it: The 73-Hour Gold Gap for the report, and Why 88% Accuracy Still Loses 0.42 bps for the kit, both at $0 with no account.

What if the file is not what this page says?

Reply to your receipt within 7 days and you get your money back. Every number in the gold report is tied to a named source, and every figure in the kit's guide can be reproduced from the sample capture on your own machine, so a question about a number gets an answer with the page or the command behind it.

Where do I get help?

Write to [email protected]. The reply comes from the person who wrote the file, usually within a day, and a question about a number, a source or a command gets a specific answer rather than a link elsewhere.

Need help?

Need help?

Send us a message and we reply by email, usually within a few hours.

Open the full support page