What we measured, what we removed, and what we got wrong. Mostly the last two.
Real yields and the dollar still move gold month to month, but neither predicts the next month, and since 2022 gold has left their old relationship far behind: $4,411 in August 2026 against $1,655. They explain about half of the fall from the January record. Every first Fed cut and hike since 1994, with what gold did next.
Read →Every month since August 1971 as a purchase: up after one year 61% of the time, after twenty years 88%, ahead of inflation far less often. The longest wait to break even was 26 years. Which buying conditions held up against random price paths, and which did not.
Read →Gold has fallen 20% or more 16 times since 1971. Falls that stopped short of 30% were back at their peak within 3.4 years; the deeper ones took up to 28. Every year since 1925 after inflation, what buying at 20% down did a year later, and the 1980 peak.
Read →Since 1971, $10,000 became $3.48 million in the S&P 500 with dividends reinvested and $1.03 million in gold; without the dividends, stocks trailed gold. Gold beat stocks in 27% of ten-year periods, rose in 5 of 7 big stock declines, and won the last five years.
Read →Above 80, silver beat gold over the next three years 90% of the time since 1971, but the ratio has crossed 80 only five times, and in four of them it kept rising first, by up to 28% against a silver buyer. The 80/50 swap rule, recessions, and the ratio by year.
Read →Gold beat US inflation in only about 55% of one- to ten-year holding periods since 1971, and lost ground in every high-inflation year since 1981, including 2021-22, when it fell 15% after inflation. Bought at the 1980 average, it took 45 years to break even. Plus gold's price adjusted for inflation, every year since 1972.
Read →In 2024 gold and the dollar both rose, and it was widely reported that their inverse link had broken. The daily correlation that year was -0.465 - fully intact. Gold moves about 1% against every 1% of dollar, every year since 2014. But the dollar cannot predict it.
Read →The first indicator in this series to survive the statistics - 31 of 75 tests beat chance where 3.8 were expected. But 7 of the 10 strongest results point to momentum, meaning higher RSI preceded higher returns. That is the opposite of what overbought is supposed to mean.
Read →Gold’s ATR runs from $0.90 to $7.04. Against a fixed stop risking the same average distance, the adaptive stop matched it through eight of ten market conditions - then held at 50.0% where the fixed stop fell to 44.5%. One configuration contradicts it, and that is in the article too.
Read →Gold tripled over twelve years, so every buy signal looks profitable. Measured against the return from a randomly chosen day, the 50/200 golden cross had a negative edge at every horizon - and 2 of 36 tests beat chance, where 1.8 were expected.
Read →We measured where 32,592 gold pullbacks actually stopped. No Fibonacci level was more popular than the levels beside it, and the golden ratio scored exactly 1.00. It feels like it works because five zones cover 46% of the range.
Read →January averages +3.61% and rose in nine of the last twelve years. It still fails once you account for having tested all twelve months at once. Volatility seasonality, measured the same way, is real - March runs 42% wider than September.
Read →Every ratio from 1:0.5 to 1:5, across 66,572 gold trades and 5.6 years. The win rate fell to offset each increase in target size to within 0.021 percentage points. The ratio is not an edge - what is left after it cancels is the spread.
Read →The median release minute moved gold $4.81, only 2.7x normal. The largest moved $40.48 in sixty seconds. Direction continued an hour later 81.8% of the time against a 54.4% control — on a sample of eleven.
Read →Every gold analysis page publishes a list of levels clustered on round numbers. Across 35,002 approaches, round levels turned price back 56.6% of the time and deliberately unremarkable ones 56.1%. The gap contains zero.
Read →The spread is not the problem: it is $0.16 and identical in 23 of 24 hours. The problem is that a $2 stop is hit 81% of the time within an hour by ordinary movement, and a scalp needs a stop that tight.
Read →After ten trades a strategy with no edge finishes profitable 44.1% of the time, and a genuinely good one looks like a failure 43.5% of the time. More trades fix the second error and never fix the first.
Read →772 signals across 5,884 hours of XAUUSD. The oversold buy beats doing nothing for about an hour, then loses to it. At twelve hours the median is +$2.10 and the mean is −$6.36 — which is the whole story.
Read →Every session guide says skip Asia and wait for London. Across 258 sessions the Asian block set the day’s low 48.1% of the time, and London set either extreme about one day in ten — the least likely block of the day, hour for hour.
Read →We simulated a strategy with no edge at all and counted how often it finished in profit. 44.1% at ten trades, 50.4% at five hundred — it never falls. Which makes “it made money” almost useless as evidence.
Read →The limit that follows your peak equity up and never falls back — which is why traders fail challenges while still up on the month. Gold closed below its own running high on 84.5% of sessions across a year.
Read →Median gap $7.60, 48% exceed $10, largest $97.98 — and a $10 stop is a reasonable intraday distance. So a sensibly-sized stop is jumped by the weekend about half the time.
Read →Nine questions to put to any vendor, including us. Drawdown, out-of-sample testing, what the strategy count actually means, and the claims that should make you walk away.
Read →117 trades, a 28% win rate and a net loss. Why the parameters that work on gold fail on bitcoin, and why we removed the symbol instead of leaving it in the settings menu.
Read →The software is legal. The brokers mostly cannot accept you, and spot gold CFDs are off the table entirely. What actually applies to a US trader, and the one route that works.
Read →