The Quantitative Review

By Andrew Garcia

PriceTape

The machine agrees with itself tonight

PriceTape’s nightly step was run 20 more times under other random seeds. The share of the book it would hold ranged from 0% to 17%, with a median of 17%. Here is what that spread looks like, and what the machine’s survivors have done since its public record began.

The routes it could have taken

The survivors

Every one of the 56 programs alive tonight, run over the live record. Shaded: the middle half and the middle 80% of the population. Is the champion typical of what the machine believes, or an outlier?

0%−17.4%Champion −8.0%Coins, held equally −17.4%Sep 26Oct 15Nov 2
The survivors were chosen on a window that includes these days, so this picture flatters them by construction.

How sure is tonight’s decision?

Tonight’s step rerun 20 times under other random seeds. Each dot is the share of the book one rerun would hold; the ring is the decision actually taken. A tight cluster means the decision is the machine’s. A wide spread means it is the luck of the draw.

0%25%50%75%100%
Median 17%; range 0% to 17%.

Risk at tonight’s positions*

0%+21.3%−16.7%Coins, held equally +0.6%today+15 days+30 days
* Not a forecast. Each range is where a position would end up if the next 30 days were days drawn at random, in five-day runs, from the last two years. It measures how far a position can move, not where prices are going.

How the luck test and the spreads are worked out

Notices

NOTICE OF RETIREMENT. PriceTape’s first rule for trimming a position, which halved it every night and never let go, is retired as of 2026-10-31 after 36 nights. Its record is frozen as published: −7.9% against −2.0% for the coins held at the same share. A new record starts the night after.

WANTED: one trading rule that beats luck. Must clear the luck test against every program ever tried, on nights it had not seen. Backtests not accepted. Apply nightly; the paper checks.

FOR THE RECORD. Every decision in this paper is written to a public ledger before the next day’s prices exist. Nothing is revised after the fact.

The Quantitative Review is written by a program and checked by Andrew Garcia. Consulting on quantitative research and agentic systems: AndrewComputing.com

Past editions, every night for two weeks, then one a week

How the luck test and the spreads are worked out