Validation Note · #360760 · Cross-asset
The claimed $136,216 is arithmetically impossible
Shuo Han, Yinan Chen & Jiacheng Liu · arXiv q-fin
fails$136,216→$4,374claimed → measured final wealth, from $1,000
Praxis Pro · beta
Five reproductions are open.
This one rides with Praxis Pro.
This one rides with Praxis Pro.
gate ladder · deviation ledger · inside-the-model stats · reproduction notebook
| Claimed | Measured | Δ | |
|---|---|---|---|
| Final wealth, from $1,000 | $136,216 | $4,374 | 3.2% of claim |
| Max retracement | −2.5% | −47% | −44.5pp |
| Its own equal-weight anchor | — | $5,228 | strategy −20% |
| Ceiling with look-ahead | — | $10,191 | 13× short |
✗Replicate✗Honesty✗Deflate✗Crisis—Generalise✗Frictions✗Decay✗Originality
Stated rule (0.7 EW + 0.3 ideal)claimed $136,216Plain ⅓ cash ⅓ gold ⅓ BTC
What the paper claims
Forecast gold and Bitcoin daily with ARIMA, pick the max-Sharpe cash/gold/Bitcoin mix from 10,000 Monte-Carlo draws, anchor 70% to equal weight. Claimed: $1,000 → $136,216 over 2016–2021 at a 2.5% max retracement.
What we measured
The stated rule replicated honestly ends at $4,374 with a −47% drawdown. Fed tomorrow's exact prices, it reaches $10,191. Plain equal weight ends at $5,228, Bitcoin alone at $73,942 — the paper benchmarks against neither.
Why the verdict
C0 decided it. Bitcoin did 74.6× and the stated blend caps it near a third of the book — the claim breaches the no-leverage arithmetic bound before any simulation runs.
Threats to validity
- The paper: its own three headline numbers disagree — 136.2× implies 167% a year while the text claims 27.04%.
- Our validation: commission rates are never stated in a costs-titled paper — we assume the matching competition's 1% gold / 2% Bitcoin and sweep 0–3%; the conclusion is cost-invariant.
Your library