SPX Statistical Mean-Reversion — Independent Replication

Strategy rules (from the article)

Buy the index at the close of any day whose open-to-close move is ≤ −threshold (2% primary); sell after a fixed holding period of N bars. Rationale: big intraday down-moves mean unexpected forced selling; market makers absorb and hedge, driving short-term mean reversion (a statistical cousin of the "VIX Rule of 16").

Replication results — 5 article configurations

$100 start · 10% sizing per signal · pyramiding on · zero costs (as the article assumed). Equity marked-to-market daily.

Article vs replication

Side-by-side for the five configurations. Win rates / PF / returns come in lower than the article; max drawdowns line up well once the equity simulation is correct.

Equity curves — strategy vs buy & hold

The edge is real but modest: at 10% sizing the strategy earns single-digit CAGR while deploying little capital most of the time. Buy & hold dwarfs it — as expected for a diversifier, not a wealth builder.

Robustness checks (not in the article)

Cost sensitivity — 2% / 7d

Edge survives 10 bps/side, erodes at 20 bps/side. Typical retail round-trip ≈ 5–10 bps — fine for a daily strategy.

Per-decade breakdown — 2% / 7d

Edge is not consistent across decades: 1980s lost money, 1970s was a coin flip, and the bulk of gains concentrated in the 2000s (dot-com crash + GFC volatility).

Alternative execution assumptions — 2% / 7d

Result is robust to execution assumptions. Notably, no-pyramiding gives the best risk-adjusted profile (PF 1.66, DD −3.4%) — pyramiding is not the source of the edge.

Walk-forward — IS 70% → OOS 30%

Enhancement analysis — can we squeeze more out?

Battery of enhancements tested on the base 2%/7d and 2%/24d configs (56y, zero costs). Full detail in results/enhancement_analysis.csv, combos in enhancement_combos.csv, out-of-sample validation in enhancement_isoos.csv.

1. Freshness — entry lag (2%/7d)

The edge decays fast as the signal ages. Buy the signal-day close; waiting kills the return.

2. Full enhancement battery (2%/7d)

Key: 3% profit target is the single best, robust enhancement on the 7d hold. VIX filter rows use the 1990+ overlap window; the Rule-of-16 rationale does not help. Stops (trailing/ATR) hurt; vol-sizing halves DD at a proportional return cost; cooldown is a risk lever.

3. Combos — 7d & 24d

The 24d trend-filtered config (close > 200d SMA) is the risk-adjusted star — at the cost of trade count and crisis-year returns. An 8% profit target on 24d improves the unfiltered long-hold profile.

4. Out-of-sample validation (IS 70% → OOS 30%)

The top enhancers are not overfits — they hold or improve out of sample. See the table for IS vs OOS profit factor, win rate, and drawdown.

Verdict

Turn the tables into a live sleeve: analysis details and actionable playbooks.