A "validated" pair's profit factor already fell below 1.0 in its last quarter

Out of 1,400 pairs we've run through our full 2-year validation pipeline, exactly 3 survive with a GRALIS_VALIDATED_2YR stamp: MEGAUSDT, EWJUSDT, and the newest one, NAORISUSDT, added after a broader MAP scan flagged it. 2,000 trades, 87% win rate, range_bounce family. On paper it looks like the strongest of the three by win rate alone. Then I opened the quarter-by-quarter profit factor our own pipeline logged for it: 1.14 → 1.33 → 1.25 → 0.98. The fourth quarter we measured is already underwater.

87% win rate, profit factor barely above 1

This is the gap that matters and it's easy to miss if you only look at the headline win rate. NAORISUSDT's overall profit factor across all 2,000 trades is 1.17 — not fake, not cherry-picked, but thin. Profit factor is gross profit divided by gross loss; 1.17 means for every dollar lost, $1.17 came back. That's a real edge, but it's a small one sitting on top of an 87% win rate, which tells you the losing 13% of trades are large relative to the winning 87%. Compare that to MEGAUSDT's validated config on the same list: 98% win rate and PF 3.31. Same family, same validation bar, completely different margin of safety.

All three carry the same GRALIS_VALIDATED_2YR verdict in our override file. Same bar cleared, three very different safety margins.

The quarterly trend is the real story

An aggregate PF of 1.17 is already thin. What our pipeline's own stability note adds is worse: it logs NAORISUSDT's edge quarter by quarter as q1=1.14, q2=1.33, q3=1.25, q4=0.98. Read left to right, that's not noise around a stable number — it's a rise through q2 and then two straight quarters of decline, ending the stretch we measured below break-even. The internal note calls this "edge slowly declining... but still positive," which is technically true of the trailing average and not true of the most recent quarter on its own. A PF of 0.98 means that specific three-month window lost money, full stop, regardless of what the first three quarters did.

EWJUSDT, by contrast, is logged as "stable across all 4 quarters" — a thinner edge (PF 1.41 overall) but one that didn't erode over time. NAORISUSDT has the bigger number up front and the worse trend line. If I had to rank these three pairs by how much I'd trust the next quarter to look like the last one, NAORIS would be last, not first, despite having the highest raw trade count tied with MEGA and a win rate that beats EWJ by 31 points.

Why "validated" doesn't mean "finished"

This pair earned its stamp through a real process — a broader MAP scan flagged it, it ran 2,000 trades across roughly two years of data, and it cleared whatever bar separates GRALIS_VALIDATED_2YR_NEW from the thousands of pairs that never made the shortlist in the first place. None of that is fake. The 269 viable records out of 152,484 we audited back in May prove the filter is strict enough to reject almost everything. But "passed validation on the full two-year window" and "still has an edge in the most recent quarter" are two different claims, and only the second one tells you anything about tomorrow. A pair can be net positive over two years and net negative over the last three months at the same time — that's exactly what the q4=0.98 number is showing for NAORIS, and a wide enough backtest window can hide that kind of recent decay behind a healthy-looking aggregate.

What I actually do with this

I'm not pulling NAORIS off the list — one sub-1.0 quarter out of four, on a "medium" trust-level, tier-2 pair, isn't proof the edge is dead, just proof it's thinner than the headline 87% suggests and trending the wrong way. What changes is sizing: this is not a pair I'd run at the same conviction as MEGA's PF-3.31 config, and it's exactly the kind of setup where risk of ruin math matters more than win rate, because a PF near 1.0 means your bust probability is far more sensitive to bet size than a PF of 3+ ever would be. Before adding size to anything sitting on a thin and declining profit factor, I run it through required win rate against the actual payoff ratio, not the headline win rate the backtest reports.

Method: numbers pulled directly from our own validated-pairs override file (stage9/MAP_map_v1, 280-pair 2-year time-stability import, imported 2026-05-30), cross-checked against the May 2026 results audit (152,484 backtest records scanned, 269 viable, 9 unique pairs surviving before the stricter 2-year filter cut that to 3). Quarterly PF figures are the pipeline's own internal stability tags, not recomputed by hand. Past backtest performance, even validated, is not a guarantee of future results. Not financial advice.

• Risk of ruin calculator — how bust probability scales with a thin edge
• Required win rate calculator — the win rate your payoff ratio actually needs
• Win rate calculator — track your own numbers trade by trade
• Break-even win rate calculator — where fees push the line