Chio VPVMA Daily ETF Panel — Literal Published Formula

Hypothesis

Derived by the Research Lead from Pat Tong Chio's paper: test the volume-, typical-price- and volatility-weighted MACD derivative on DIA, QQQ and SPY daily bars. Preserve the published 12/26/9 periods, 0.1 bandwidth and both printed signal inequalities. Equation 4.2-6 uses SVWMA rather than LVWMA; this experiment deliberately implements that literal expression, without claiming to resolve the author's intent. Population rather than sample standard deviation is an explicitly chosen convention. Replace the unspecified constituent universe with the three ETFs explicitly named in Section 3.2.2, using split-adjusted platform OHLCV. Allocate independent, non-transferable sleeves of one-third initial capital, with 95% of each sleeve's available equity invested on entry. Whole shares, no shorts, leverage 1. Orders decided at session close fill at the next session open. No added stop or target: risk is bounded by sleeve allocation rather than a guaranteed loss limit. Source performance claims remain unverified.

The concentration brief is empty, so no instrument-share inference is possible. USEQ avoids the overrepresented Binance bucket and the ETF panel reduces single-symbol concentration. Faithfulness to this long-only US-equity source takes precedence over adding shorts or options solely to meet quotas. This is distinct from the listed EMA/OBV, Supertrend and cointegration failures.

Implementation

Literal daily VPVMA on QQQ, DIA and SPY, long-only whole shares, independent one-third initial-capital sleeves invested at 95%, next-session-open execution and no stop or target.

Preserves Chio equations and both printed inequalities, including SVWMA in the slow EMA (https://arxiv.org/pdf/2206.12282, pp. 32-33). Uses population OHLC standard deviation, SMA-seeded 12/26 EMAs, a 9-bar SMA signal and bandwidth 0.1. Each ETF observes 260 warmup bars. Scalar sleeve ledgers debit purchases and credit sales at actual fills including commissions; gains cannot fund another sleeve. Extra bars contribute only at the matching primary close timestamp. The research parameters remain frozen; sleeve allocation limits exposure without guaranteeing the drawdown criterion. Source performance claims remain unverified.

Verification Results

Document this next-open sizing limitation. If strict non-transferable funding is required at execution, enforce a per-sleeve affordability check in the execution layer without reading the future open in strategy code.

Share quantity uses the signal close, while execution occurs at the next session open. An upward gap exceeding approximately 5.26% can exhaust the sleeve's 5% cash reserve and make its ledger negative. The ledger separates realized proceeds but does not enforce isolated funding at execution.

The smoke test reports Sharpe 0.131 and stressed-regime return -29.19%. These are advisory observations; full-history evaluation must assess the frozen strategy.

Backtest Review

Sharpe

0.33

Total return

148.33%

Max drawdown

39.66%

Trades

1316

Win rate

59.9%

Profit factor

1.27

Numerically viable baseline. 1,316 closed long trades from 2000 to 2024 are a reliable sample. avg_trade_return_pct is 0.204% against the USEQ floor of 0.05% (4.1x), PF is 1.27, Sharpe is 0.33 and accounting reconciles.

The trades match the literal Chio specification. All trades are long and split across the three sleeves (QQQ 468, DIA 437, SPY 413). Each position is about 0.32x leverage, roughly one-third sleeve times 95%. The average hold is about 11 days and fills occur at the next open, so the mechanism is implemented as written.

Fees are not a factor: commission is zero and impact is 9.4% of gross.

There is no measurable forecasting information. The executed-signal IC is -0.003 (t=-0.24, n=5,991), alpha is 0.5%/yr and the information ratio is -0.16.

The result looks like beta rather than edge. The equal-weight QQQ/DIA/SPY buy-and-hold basket returned about +252% with a 55.2% drawdown. The strategy returned +148% with a 39.7% drawdown, which is 0.72x the benchmark drawdown at beta 0.58 (lesson 170 pattern).

Max drawdown of 39.7% breaches the contract's 30% risk constraint. The strategy loses in the stressed-volatility tercile (-30.2%, Sharpe -0.10) and in 2001, 2008, 2020 and 2022. The longest drawdown lasted 2,390 days.

The Sharpe 95% CI (-0.08 to 0.70) includes zero. The parameter set is fully frozen with nothing tunable, so the optimization phase is only a frozen OOS/holdout evaluation and cannot improve these numbers.

Analysis

Accounting is valid. The input audit is clean on all 3 legs (6252 bars each, 0 missing). Leverage peaks at 0.96, inside the cap.

Per-trade economics clear the USEQ floor: avg_trade_return_pct is 0.204% against a 0.05% floor. Commission is $0 and impact is 9.4% of gross, so fees are not the problem.

The sample is large: 1316 closed trades over 24 years, with 18 of 24 years positive.

The hypothesis's own risk constraint fails. Max drawdown is 39.66% against the pre-registered 30%. The optimizer rejected the frozen candidate on max_drawdown_pct in the inner training of all 59 walk-forward windows (37.6% to 39.7%), so there is no risk-admissible candidate, no OOS and no final parameters (insufficient_candidate_coverage, category 'rejected').

It has no edge over the benchmark. Information ratio is -0.156 and alpha is 0.005. The equal-weight DIA/QQQ/SPY basket made +278% with a 55.2% drawdown; the strategy made +148% with a 39.5% drawdown. That drawdown is 0.72x buy-and-hold's at about 55% time in market per sleeve, so the profile is scaled beta, not timing skill (lesson 170).

The signal carries no information. The executed-signal IC is -0.003 (t=-0.24, n=5991, verdict rejected_no_information).

It loses money in the stressed volatility third (Sharpe -0.10, -30.2%, drawdown 42%) and in every listed crisis window: COVID -21.5%, the 2022 rate shock -18.7%, and -24.3% for 2022 as a whole. Win rate is 0.60 with a payoff of 0.85, so PF is only 1.27 (lesson 201).

Pre-registered predictions missed: the drawdown limit (39.7 vs 30) and per-trade return (0.20% vs 0.95%). The trade count and return predictions were met.

> 0, significant

pre-registered research constraint; enforced in every WF window

Outcome Summary

Clearing the per-trade fee floor is not evidence of edge: a long-only timing overlay on index ETFs with zero IC and beta of about 0.58 just reproduces scaled beta and, with no stop, can break a pre-registered drawdown limit in crises.

The backtest review sent it to optimization. Walk-forward then rejected the frozen candidate on the hypothesis's own 30% drawdown constraint in all 59 windows, so no out-of-sample result or final parameters were produced. The analyst abandoned it as premise_falsified: the risk limit was breached, it had no edge over the benchmark (scaled beta), the signal carried no information, and it lost money in the stressed-volatility tercile and in the 2008, 2020 and 2022 drawdowns.

A literal implementation of Pat Tong Chio's volume-, typical-price- and volatility-weighted MACD (VPVMA, eq. 4.2-6 with SVWMA as the slow input, 12/26/9 periods, 0.1 bandwidth), trading long-only one-third sleeves on DIA, QQQ and SPY daily bars with next-open fills.

From 2000 to 2024 it returned +148.3% (CAGR 3.89%) with Sharpe 0.33, 1,316 closed long trades, a 0.204% average trade return, profit factor 1.27 and a 39.66% max drawdown. The executed-signal IC was -0.003 (t=-0.24) and the information ratio was -0.156 against an equal-weight ETF basket.

Analysis

A comparative study of the MACD-base trading strategies: evidence from the US stock market