MFI Reversion (14, 20/80)
Positive risk-adjusted return in 11 of 18 cells — strongest on SPX500 · 4H (Sharpe 1.33); 7 cells lose after the 0.05% per-side fee. Median cell trails buy-and-hold by 138.1pp; 1 of 18 cells beat it.
§ Verdict rationale
Composite Sharpe 0.26 across 18 cells. The edge concentrates in SPX500 4H (Sharpe 1.33); 7 of 18 cells are negative after the 0.05% per-side fee.
§ Alpha score breakdown
alpha = edge (0-40: clamp(compositeSharpe/2,0,1)*40) + stability (0-25: positiveCells/18*25) + flaw penalty (0 to -25: static findings, 0 across v1 catalog) + reproduction (0-10: engine oracle parity vs pine2py, 251/257 scripts = 9)
§ Equity vs buy-and-hold — fee-adjusted, per cell
§ All 18 cells — including the ugly ones
| Market · TF | Sharpe | Return | vs B&H | Max DD | Win | Trades |
|---|---|---|---|---|---|---|
| BTCUSD 1H | -0.04 | -71.2% | -1452.5pp | -87.1% | 60% | 335 |
| BTCUSD 4H | 0.09 | -45.9% | -1413.2pp | -81.6% | 62% | 84 |
| BTCUSD 1D | -0.12 | -74.1% | -1463.5pp | -82.6% | 46% | 11 |
| ETHUSD 1H | -0.02 | -84.9% | -613.1pp | -95.4% | 61% | 367 |
| ETHUSD 4H | -0.13 | -89.3% | -604.5pp | -96.5% | 54% | 85 |
| ETHUSD 1D | 0.45 | +252.7% | -274.6pp | -88.5% | 92% | 12 |
| SPX500 1H | 0.32 | +9.0% | -63.2pp | -14.5% | 67% | 21 |
| SPX500 4H | 1.33 | +11.9% | -59.7pp | -3.0% | 100% | 4 |
| SPX500 1D | 0.60 | +216.7% | -366.6pp | -28.7% | 93% | 14 |
| QQQ 1H | 0.39 | +15.6% | -77.3pp | -17.2% | 68% | 22 |
| QQQ 4H | 0.94 | +46.9% | -45.0pp | -20.2% | 89% | 9 |
| QQQ 1D | 0.58 | +268.4% | -1141.2pp | -29.9% | 91% | 11 |
| GOLD 1H | 0.80 | +28.0% | -101.0pp | -19.7% | 75% | 20 |
| GOLD 4H | 0.19 | +4.3% | -124.1pp | -18.4% | 75% | 4 |
| GOLD 1D | 0.51 | +72.8% | -152.1pp | -18.2% | 75% | 12 |
| EURUSD 1H | 0.00 | +0.0% | -9.1pp | 0.0% | 0% | 0 |
| EURUSD 4H | 0.00 | +0.0% | -9.1pp | 0.0% | 0% | 0 |
| EURUSD 1D | 0.00 | +0.0% | +18.6pp | 0.0% | 0% | 0 |
11 of 18 cells positive · best SPX500 · 4H · cells are not equal length — the composite is a median
§ The exact source that ran — Pine v5
//@version=5strategy("MFI Reversion (14, 20/80)", initial_capital=100000, default_qty_type=strategy.percent_of_equity, default_qty_value=100)m = ta.mfi(hlc3, 14)longSig = ta.crossover(m, 20)exitSig = ta.crossover(m, 80)if longSig strategy.entry("L", strategy.long)if exitSig strategy.close("L")plot(strategy.equity, "equity")plot(strategy.position_size, "pos")plot(strategy.closedtrades, "closed")plot(strategy.wintrades, "wins")plot(strategy.grossprofit, "gp")plot(strategy.grossloss, "gl")This source was executed verbatim by the wavealgo JS engine — there is no port step to drift. The engine is oracle-verified bar-by-bar (1e-9 tolerance) against an independent Python implementation on 251/257 corpus scripts.
§ Flaw checklist
Verdict rules: pass: composite Sharpe >= 0.9 and >= 12/18 cells positive · cond: composite Sharpe > 0 and best cell Sharpe >= 0.9 · fail: otherwise · rep: static repaint/look-ahead finding (none in the authored v1 catalog: confirmed-bar signals, next-bar-open fills)
§ What the numbers mean
- Sharpe ratio
- Risk-adjusted return: mean daily return divided by its standard deviation, annualized ×√252. The composite figure is the median of the 18 cell Sharpes, not the mean — cells differ in length by an order of magnitude, and one long cell should not carry the other seventeen.
- Max drawdown
- The worst peak-to-trough equity loss over the window, measured on the fee-adjusted curve. What you would have sat through, not what you would have ended with. The headline figure is the deepest of the 18 cells, not their average.
- Profit factor (PF)
- Gross profit divided by gross loss, with fees charged to the loss side. Above 1.0 means gross gains exceeded gross losses; 1.0–1.2 is usually noise.
- Win rate
- Share of closed trades that closed profitable. A high win rate with a profit factor near 1 means many small wins and a few large losses — common in mean-reversion.
- vs buy-and-hold
- Strategy total return minus the return of simply holding the asset over the same window — the benchmark any active strategy must beat to justify existing. Shown as the median of the 18 cells plus how many cells beat it; a mean would let one long cell stand in for the rest.
- Alpha score
- The desk’s 0–100 composite: out-of-sample edge (0–40; the methodology page defines what out of sample means for the catalog), regime stability across cells (0–25), flaw penalty (0 to −25), reproduction accuracy (0–10). The exact formula ships with the data.
Bring your own. We grade it the same.
Engine: resin 0.2.2 (next-bar-open fills, TV rule) · scored UTC 2026-11-07 10:58 · deterministic rerun — same data, same numbers. Backtested results are measurements of the past, not investment advice; nothing here is a recommendation to trade. wavealgo is operated by Streamize LLC.