ADMIN
Internal · Kevin Artis Only
Fleet — all active books
One page for every book that holds capital: engine up or down, open positions marked off the last regular-session bar, realized split from signal-attributable, and TA 5.0 against its random-universe control. Start here
TA 6.0 — allocation arms
Four $25k books on one engine and one universe, differing only in how capital is committed. On TA 5.0’s own logs it funded 119 of 1,089 setups, and the ones it refused for want of cash scored 0.26R better (t −2.18): a tight stop buys more shares, so the good setup is the expensive one a 97%-invested book turns down. Whether that is cause or coincidence is what these three corrections and their control are running to settle. Live 2 Nov 2026; one verdict at 200 round trips, around March 2027
Option Trader 1.0 — dormant
Options research for the $25k book. The nightly chain collection still runs, but vertical spreads were found infeasible once real spreads and commissions were costed, so nothing is being built on it — data is accruing for a programme that is closed
Barbecue Sauce 1.0 — retired
$25k shorts-only book — VWAP_Reject, three stop-width arms (1.00x / 0.40x / 0.40x + 10-bar cap). Retired 25 Oct 2026, short of the committed three-month run: zero expectancy under its own mechanics, and 71% of the signal’s edge died at a realistic entry fill. Flat on all three arms; the page is kept for the record, not as a live book
Index 1.0 — RSI(2) reversion
Short-horizon mean reversion on index ETFs. 20 years of evidence, both halves agree; two slot-fair positions on a $25k book, no stop by design — the validated exit is the RSI condition alone
Dart Board 1.0
Control arm — a size-matched random universe run on the same engine as TA 5.0, testing whether the watchlist’s selection rule beats drawing names out of a hat
Experiments — themes and A/B forward tests
Every running test that had no page of its own: the three theme baskets (oil capex, Venezuela, shipbuilding) against their benchmarks, and the paired A/B books. Each exists to beat a control, so the page leads with the spread rather than the equity. Auto-discovers any state file it does not yet describe, and marks a test stale when it stops writing one
BCDA sandbox — limits lifted
A single-symbol tracking book, built 6 Nov 2026 on request. Runs TA 5.0’s engine, mounted read-only so nothing here can drift from the baseline, with three scope limits removed: the $10 minimum entry price, every Small-tier signal pause, and the uptrend regime block. Not a forward test — no control arm, a universe of one, and a signal set chosen by lifting pauses rather than by evidence, so several of the families it can now act on failed the October rescore. Read it as what happened to BCDA under the engine, never as evidence about a signal