2026-06-24 · Brent Akamine
Edge vs. noise: what eight retired strategies have in common
Eight archetypes have come off the board. Lined up, they tell a clearer story than any single report:
| Strategy | Verdict | The tell |
|---|---|---|
| Basket-grid | PF 0.51 | Coin flip before costs; martingale can’t manufacture alpha |
| Forex Fury / CORE | net PF 0.73–0.90 | Gross PF ≈ 1.0; no edge in the parameters |
| D1 trend (50/200) | PF 0.94 | Only gold trended; the FX majors ranged |
| NightMeanRevert | PF 0.66 | Reverts 21%, stops 66%; no parameter island |
| TrendPullback | PF 0.80 | Right skew, wrong win rate (41.7% vs 47% breakeven) |
| Carry (hardened) | PF 0.84 | Lost to a random-sign placebo |
| Crabel NR7 | PF 0.64 | Signal beats random, but loses net of costs |
| ETH/BTC pairs | PF 0.69 | 58% of random permutations beat it |
Pattern 1: the edge was never there (most of them)
For the majority — basket-grid, Forex Fury, D1 trend on FX, night mean-reversion — the decisive diagnostic was a zero-cost or placebo control showing a coin flip. Strip the costs and these systems sit at profit factor ≈ 1.0. There is no directional edge for the risk machinery, the grid, or the kill-switches to compound. Costs then turn a zero-edge bet into a steady bleed.
The lesson the basket-grid report put most sharply: a spec can fully define the risk machinery — kill switches, recovery, two-strike overrides, regime vetoes — and still never define an entry edge. Risk management cannot manufacture alpha. If the signal is a coin flip, the most disciplined risk overlay in the world only changes how fast you lose.
Pattern 2: the right shape, the wrong calibration (the near-misses)
A smaller group failed differently — and more instructively. TrendPullback had the genuine positive skew a trend system should have (winners 1.22× losers) but a 41.7% win rate where ~47% is breakeven: right-shaped, wrong-calibrated. Carry earned a real but tiny premium that the FX leg dwarfed. NR7 produced a signal that beat its placebo — real information — but still lost money net of costs on gold.
These are the valuable failures. They map the boundary between “no edge” and “edge too thin to monetise,” and they tell you where the remaining research problems actually are (closing 5 points of win rate; finding regimes where the premium is large enough; lower-cost execution).
What separates the survivors
The strategies still on the board — Turtle, VRP, HLP, the funding harvest — share two things the retired eight lacked:
- They beat their placebo decisively. Not marginally — at the 99th percentile or better. The signal carries information that random labels do not.
- They survive the multiple-testing correction. A positive deflated Sharpe means the result isn’t an artifact of trying many configurations and reporting the winner.
What the survivors don’t automatically have is a deployable risk profile — Turtle’s −49% drawdown and VRP’s fat left tail are real constraints. But a real edge with a risk problem is a fundamentally different object than a coin flip with a good story. The first earns a Round 2. The second gets retired.
That distinction — drawn honestly, in public, on every strategy — is the whole point.
See the live board: scoreboard · how the gates work: framework.
By Brent Akamine (Founder, Vinovest). Part of The Validation Gauntlet. Backtests are not investment advice.