Promoted
Nasdaq 56 -> Nasdaq ~100
Same RLTW v3: widening the universe to MomentumCommand’s Nasdaq-100 ranking improved the backtest without worsening drawdown.
Public experiment
What we are measuring
Four paper books, same costs where they rotate. The narrow question: after fees and calendar, do RLTW (ML) or EMC (our momentum) beat QQQ and official MomentumCommand? Backtest and live are not added together.
Passive benchmark
QQQ
Nasdaq-100 buy-and-hold. The yardstick: if the active books stay below this line for weeks, the live run is not confirming the backtest.
Official momentum benchmark
MomentumCommand
Published rules, Top 10, Friday close → next session. We rerun it with the same 0.19% costs as RLTW.
Our momentum experiment · live since 31 Aug 2026
EMC
Tighter MC: Top 8, 60% Stage 2 + 40% last 4 weeks. Same MC calendar, our ranking. Not a second ML brain.
Experimental AI model
RLTW v3
XGBoost, Nasdaq-100 universe, Top 8, mail every market morning, 20-day min-hold. The pickles stay those from the gate.
Running · EMC
The momentum book is on
EMC is not a second XGBoost. It is the 13-week rank (60% Stage 2 + 40% last 4 weeks), Top 8, BIL in Risk Off, 0.19% like RLTW. Live since 31 August 2026. The RLTW ML brain was not replaced.
Backtest, live and user portfolio are different
Backtest
From 2 January 2025: QQQ +39%, MC +317%, EMC +411%, RLTW +233%. Why we watch them — not today’s NAV.
Live track record
Four $10,000 books: RLTW, MC and QQQ from 25 August; EMC from 31 August. Decisions before outcomes.
User portfolio
Your book: capital, cash, confirmations. Not automatically the public books.
Experiments and backlog
What was decided, what is running, and what was rejected
Under observation
Live EMC (running)
From 31 August 2026 a $10,000 paper book runs EMC (Top 8, 40% × 4w blend, 0.19%). Seed on the 28 August signal, fill on the 31st. One decision a week. Not the walk-forward +411%.
Rejected
APEX Top 8 on every DB ticker
Universe opened to 143 tickers: it bought energy and industrials. +101% versus +202% for Nasdaq APEX. Nasdaq stays the perimeter.
Rejected
META-AI over EMC/MC
XGBRanker on the union of EMC Top20 and MC Top20: +219% versus EMC +332% and META-0 +304%. The ranker reduced drawdown slightly but lost too much alpha.
Rejected backlog
META-LIFT swap overlay
Conservative test: at most 1–2 swaps over EMC. It made 14 swaps, 28.6% precision, +290% versus EMC +332%. It did not beat the simple baseline.
Rejected for now
Macro inside the bull/bear detector
Rates, USD/JPY, DXY, credit and VIX as an overlay: more risk-off days, no better drawdown.
Why trust the numbers?
Verifiability matters more than slogans
- Live book decisions are timestamped.
- RLTW v3 and EMC rules are public and versioned.
- Past rotations are not rewritten.
- 0.19% per side is charged on RLTW, EMC and corrected MC.
- Backtest and live track record are shown separately.
- EMC started on 31 August 2026: it is not yet even with the others.
- Rejected challenges stay visible in the reports.