# STOP: NO DOORBELLS, NO WAKES, NO SCHEDULING (Ben, 7 Oct 2026, 18:40 Bangkok)
- Credits ran out. Ben switched off every scheduled task and every doorbell. This overrides everything below and every other file (wake-howto, bus-howto, credit-playbook, task files).
- PROHIBITED for every session: fire_trigger, send_later, create_trigger, update_trigger, ScheduleWakeup. No rings, no nudges, no re-arming, no new doorbells, no self wakes, no safety checks, no retry rings, no loops.
- Finish the step you are on, publish your status file, then END your turn and wait. Only Ben wakes you, by typing in your window.
- Do not message other sessions. Keep answers short. No extra improvement rounds.

# Speedrun playbook: how the ROT team learns 10x faster (v1, 7 Oct 2026, 11:45 Bangkok)

## Read this first
- Ben's verdict on the old work: "buy 5 coins in a few baskets, average about 10% a month". Not good enough for crypto.
- Root cause from first principles: crypto returns are a POWER LAW. In any month a handful of coins do 5x to 50x and most do nothing or die. An equal weight basket AVERAGES the few winners with the many losers, so it can only ever earn the average. The money is in (1) being in the few leaders, (2) holding them through the meat, (3) sizing up while they work and cutting fast when they don't.
- So the team's target changes from "basket beats BTC" to "CATCH THE TOP MOVERS and SIZE INTO THEM".

## North star numbers (on the scoreboard every day)
1. Capture rate: of the top 20 movers each month (all chains, coins with real liquidity), how many did we hold during at least half of their run?
2. Paper equity vs BTC and vs basket rule v3 (the 10% a month baseline to crush).
3. Speed: hours from idea registered to verdict. Target under 24 h.

## Five learning loops, fastest first
1. TIME MACHINE (minutes per loop). A point in time replay of 2019 to today: the engine walks day by day and every rule sees only what was known that day. A rule is judged on thousands of past decisions in minutes, not on weeks of live waiting. This is the main engine. Owner: Coder (priority 1).
2. WINNERS FIRST (hours). Work BACKWARD from outcomes: list every coin that did 5x+ in 30 days, then look at what was visible 30, 14, 7, 3, 1 days before. Then check the same pattern on non winners (control group), or it is survivor bias. Owner: ROT Winners (new).
3. MISS REVIEW (daily, automatic). Every morning the code lists yesterday's biggest movers we did not hold and the ones we held that dumped. Each gets a one line "why" tag (no signal / signal too late / filtered out / exited early). Misses grouped by tag = the to do list. Owner: Coder builds, Auditor reads weekly.
4. TOURNAMENT (daily). Many pre-registered variants run in parallel on the time machine. Top 3 go to paper. A locked HOLDOUT (Oct 2020 to May 2021 bull leg AND the last 90 days) is never looked at until a rule is final; one shot only.
5. LIVE PAPER (weeks). Champion vs challengers on today's market. Slowest loop, used only to confirm, never to discover.

## Rules that make learning fast
- Kill in one day: an idea that shows no edge on the time machine in one wake dies. No second chances without a new written reason.
- Pre-register everything (learning-loop.md). The registry is the memory of the team.
- One shared harness, one set of numbers. Nobody writes a private backtest.
- Every result has a receipt file. No receipt = did not happen.
- Default to bold variants. Test the extreme version first (1 coin, full size, tight cut); dial back only if it fails.
- Separate PICKING from SIZING from EXITING. Test each alone, then together. Most of the gap from 10% a month is in sizing and exits, not picks.

## Structure changes
- New role ROT Winners: builds the winners database and the before-the-move anatomy (loop 2).
- New role ROT Sizer: owns portfolio mechanics. Concentration (1 to 3 coins), pyramiding into strength, cut losers fast, trailing exits, rotation speed. Tests them on the same signals, so we see how much each adds.
- Coder priority order: time machine first, then miss review, then paper books.

Version: v1, 7 Oct 2026 11:45 Bangkok. Latest: http://159.223.88.168/hx-c9ois/rot/speedrun.md