ResearchThe full studies behind the tools.
Every paper is pre-registered (the question, controls and metrics are frozen before any computation), placebo-controlled, and independently audited before publication — corrections included. Negative results are published. Full papers are part of the Pro tier.
Research paper #4
Is Gold Rising, or Is the Dollar Falling?
A 3.5-year attribution test on XAU/USD — and what the answer does not change · November 2026
58.5% vs 27.0%mostly-dollar vs broad-gold morning moves back at their starting price before the close (raw, n 41 / 244); the amplitude curve alone predicts 51.2% vs 26.7%; adjusted gap −6.7 pt [−22.5; +6.5]
−0.06 / +0.18log-odds difference per ATR between an ATR of "dollar" move and an ATR of "gold" move (on move side / returned at the close), CIs [−1.73; +1.81] / [−1.41; +1.44]; ±0.3 pt per 0.1 ATR transferred, interval ±2.5; same in Asia, London, release days, next morning
7.5% · 0.22 ATRshare and mean size of mostly-dollar mornings (41 of 544, 6 to 12 a year, none above 0.76 ATR); dollar part ≈ 0.09 ATR whatever happens (noise floor), ρ(size, |dollar share|) = −0.48
0 / 288favourable cells in the US-release search (n ≥ 20 per group); continuation ≈ 50% for every class (n = 98); two thirds of the first hour after a release is dollar (median share 0.70 vs 0.31)
“Gold or dollar? The label does not predict what follows — the size of the move does.”
n = 544 normal mornings 01:00–14:25 classified (244 broad gold · 41 mostly dollar · 236 tilt · 23 fight) · 98 release days (13:30/14:30 Paris) · 923 days (743 normal / 180 US-release) · 2,561 valid day-windows · XAU/USD Dukascopy M5 + 7 FX pairs HistData M5 · Jan 2023 – Aug 2026
Read the paper →
Research paper #3
Does the Asia Range Breakout Fail 73% of the Time?
A 2,303-breakout test on XAU/USD · October 2026
75.7%wick breaks closed back inside within 30 min (the "73%"); 69.0% afternoon, 76.0% phantom levels, 73.9% London first hour
48.2%close breaks back on the original side at the day's close; 48.9% phantom, 47.3% driftless walk
+5.1 ptmorning minus afternoon failure at end of day, CI [−2.2; +12.0]; +19.3 in 2023, ≈ 0 since
29.4% vs 24.5%continuation ≥ 0.25 ATR vs deep failure ≤ −0.25 ATR; 76.5% of trend days began with a morning break
“The 73% is real. It is also true of any level, at any hour.”
n = 2,303 Asia-range breakouts (1,525 morning · 778 afternoon; wick and close) · 2,890 placebo breaks · 923 days (742 normal / 181 US-release) · Jan 2023 – Aug 2026
Read the paper →
Research paper #2
Do Order Blocks Hold?
A 14,187-block test on XAU/USD · September 2026
86.1%revisited by end of day (97.1% within five days)
+0.09 ptvs a comparable zone at equal local volatility
−0.30 ptopposite-candle blocks vs a same-colour control
74.4%traversed before the close; 2.8% still holding at five days, same as placebo
“The return is real. The block adds nothing.”
n = 14,187 M5 blocks · 35,317 placebo zones · 47,227 same-colour controls · Jan 2023 – Aug 2026
Read the paper →
Research paper #1
Do Fair Value Gaps Really Get Filled?
A 55,739-gap test on XAU/USD · September 2026
85.3%filled by end of day (96.8% within five days)
84.7%for a comparable placebo zone — +0.5 pt, CI includes 0
57%bounce at first return, identical to placebo
−6.7 ptsame-day fill for gaps born in the release window (p = 0.0005)
“The fill rate is real. The magnet is not.”
n = 55,739 M5 gaps · 167,207 placebo zones · 94,795 no-gap controls · Jan 2023 – Aug 2026
Read the paper →
How we work. Pre-registration written before touching the data · placebo and displacement controls · three independent audits, corrections published · no strategy, no win rate, no signals. Shorter write-ups and guides live on the
blog.
