Research · Dataset
Gold price history, 1960–2026
A complete monthly gold price series with the drawdown and volatility structure computed from it. Every figure on this page is calculated by our published script from the raw data, not typed in by hand.
Last updated: 2026-10 (data through 2026-10) · Source data updated November 03, 2026
Source and method
Monthly averages, not daily closes. A monthly average smooths intramonth volatility, so drawdown depths computed from this series are slightly shallower than the same drawdown measured on daily closes. That bias is why the numbers here are used as a structural reference, not as a worst-case bound.
Long-run picture
The multiple is computed as the product of every monthly ratio from the first observation to the last: ∏(Pi/Pi−1) = 117.29. It is a price change, not a return: it excludes storage cost, the spread you pay on entry and exit, and any financing on margin. A strategy that traded this series had to out-earn that friction, not just finish positive.
Drawdown episodes
26 distinct peak-to-trough episodes occurred in 802 months of monthly average data. Each row is a peak, the lowest monthly average before the prior peak was regained, and the depth between them. All 25 but one were eventually recovered on monthly averages.
| Peak | Peak $ | Trough | Trough $ | Depth | Months to trough | Months to recover | Status |
|---|---|---|---|---|---|---|---|
| 1980-01 | 675 | 1999-07 | 256 | -62.1% | 234 | 82 | Recovered |
| 1974-12 | 184 | 1976-08 | 110 | -40.2% | 20 | 19 | Recovered |
| 2011-09 | 1,772 | 2015-12 | 1,076 | -39.3% | 51 | 55 | Recovered |
| 2008-03 | 968 | 2008-11 | 761 | -21.4% | 8 | 10 | Recovered |
| 1973-07 | 120 | 1973-11 | 95 | -20.8% | 4 | 2 | Recovered |
| 2026-02 | 5,020 | 2026-07 | 4,073 | -18.9% | 5 | — | Open |
| 1969-05 | 43 | 1969-12 | 35 | -18.6% | 7 | 20 | Recovered |
| 1974-04 | 172 | 1974-07 | 143 | -16.9% | 3 | 4 | Recovered |
Two figures matter more than the depth itself. First, the duration: the deepest episode ran 234 months from peak to trough and a further 82 months back to the prior peak. Second, the recovery asymmetry — a 62.1% loss needs a 163.6% gain to undo. Any position sized so that the account survives the drawdown but not the recovery duration is a position that was never actually survivable.
Monthly loss distribution
How often a month was this bad, measured on every one of the 801 month-to-month changes in the series.
Read this as a sizing input, not a forecast. A month of −5% or worse happened in roughly 6% of months in the sample — so a rule like "one −5% month is acceptable" is true roughly 6% of the time, and false the rest.
Worst single months
| Month | Monthly change |
|---|---|
| 1980-03 | -16.69% |
| 1983-03 | -14.46% |
| 1980-12 | -13.92% |
| 1982-03 | -11.76% |
| 2006-06 | -11.70% |
The five largest monthly declines in the whole series. Notice the clustering: 1980-03 and 1980-12 and 1982-03 all fall inside the post-1980 monetary tightening cycle. Gold's worst months have historically arrived in runs, not as isolated events — which is what makes stop-loss distance on a monthly horizon much harder to size than on a daily one.
By decade
| Decade | Years | Avg annualised vol | Best year | Worst year | Down years | Decade compound |
|---|---|---|---|---|---|---|
| 1960s | 10 | 2.2% | +13.9% | -16.7% | 1/10 (10%) | -5.1% |
| 1970s | 10 | 17.3% | +100.4% | -21.0% | 1/10 (10%) | +822.2% |
| 1980s | 10 | 16.2% | +19.1% | -26.4% | 5/10 (50%) | -40.0% |
| 1990s | 10 | 9.1% | +16.4% | -18.6% | 7/10 (70%) | -28.4% |
| 2000s | 10 | 12.2% | +32.1% | -8.3% | 2/10 (20%) | +202.5% |
| 2010s | 10 | 10.3% | +24.4% | -26.9% | 4/10 (40%) | +11.4% |
| 2020s | 7 | 11.5% | +59.0% | -13.6% | 3/7 (43%) | +115.6% |
Annualised volatility is the standard deviation of that year's monthly returns multiplied by √12. The 1960s figure is not a usable baseline: gold was still pegged to official prices and traded in a closed institutional market, so it reflects policy rather than price discovery. The 1970s onward is the honest comparison set. Notice also the down-year share: the 1990s were negative in 70% of years, which is the base rate a long-only gold trend system has to survive.
Most recent 12 years
| Year | Months | Avg price $ | Dec close $ | Return | Ann. vol | Best month | Worst month |
|---|---|---|---|---|---|---|---|
| 2026 | 10 | 4,507 | 4,105 | -13.6% | 16.7% | +8.3% | -7.8% |
| 2025 | 12 | 3,442 | 4,309 | +59.0% | 12.4% | +10.6% | -0.4% |
| 2024 | 12 | 2,388 | 2,648 | +30.2% | 10.6% | +8.0% | -1.5% |
| 2023 | 12 | 1,943 | 2,026 | +6.7% | 7.9% | +4.6% | -2.5% |
| 2022 | 12 | 1,801 | 1,798 | -1.0% | 12.4% | +5.0% | -5.7% |
| 2021 | 12 | 1,800 | 1,790 | -4.1% | 9.3% | +5.1% | -5.0% |
| 2020 | 12 | 1,770 | 1,858 | +19.0% | 11.0% | +6.6% | -2.4% |
| 2019 | 12 | 1,393 | 1,479 | +14.5% | 9.5% | +6.2% | -1.6% |
| 2018 | 12 | 1,269 | 1,250 | -6.1% | 6.1% | +2.4% | -3.4% |
| 2017 | 12 | 1,258 | 1,264 | +6.0% | 7.6% | +3.7% | -2.6% |
| 2016 | 12 | 1,249 | 1,157 | +5.4% | 14.5% | +9.3% | -6.5% |
| 2015 | 12 | 1,161 | 1,076 | -14.0% | 9.1% | +3.0% | -6.3% |
Return is December over January within the year; it excludes the change from the prior December, so it does not tie exactly to the compound column above. Rows with fewer than 12 months are partial years.
How to use this — and how not to
- This is a price history, not a strategy backtest. It contains no position sizing, no spread, no commission and no slippage.
- Drawdown depths on daily closes are worse than the numbers here. Sizing against monthly averages leaves you under-sized for a real intraday move.
- The 1980 peak-to-trough episode took 234 months. If your risk model cannot hold through that, the position size is wrong regardless of the entry thesis.
- Past price structure does not establish a forward edge. Nothing on this page is a forecast or a performance claim.
Source: World Bank Commodity Price Data (The Pink Sheet). Reuse under CC BY 4.0 with attribution. Generated
by scripts/gold_history_analysis.py. Last updated 2026-10.
For how this data is used in practice, see Reading an Equity Curve and Gold Position Sizing, or run the arithmetic yourself in the drawdown and Kelly calculators.
For the same history in inflation-adjusted terms — where the 1980 peak looks far worse than the nominal chart suggests — see Inflation-adjusted gold, 1960-2026.
For the same monthly series broken down by calendar month, with the bootstrap, permutation and out-of-sample tests that show how much of the seasonal story survives scrutiny, see Gold and the calendar.
Frequently asked questions
What data is this page built on, and over what period?
Monthly average gold prices in nominal US dollars per troy ounce, 802 months from 1960-01 to 2026-10, taken from the World Bank Pink Sheet monthly commodity price file. The series starts at $35 per ounce and ends at $4105, a 117.3 times multiple over 67 calendar years. The all-time high monthly average in this sample is $5020 in 2026-02. Prices are monthly averages, not closes and not intraday, so every return on this page is a change in the monthly average.
How deep do gold drawdowns actually get?
The deepest in this sample ran from the $675 peak in 1980-01 to $256 in 1999-07, a decline of -62.1% that took 234 months to reach the bottom and a further 82 months to recover — it did not get back to its old high until 2006-05. The table lists 26 drawdown episodes in total. The current one, from $5020 in 2026-02 to $4073 in 2026-07, is -18.9% over 5 months and is the only episode in the sample that has not yet recovered.
How often does a monthly average gold price fall by 5% or more?
Far less often than the headlines suggest. Across 802 monthly observations, 244 months (30.5%) fell by at least 1%, 50 months (6.2%) fell by at least 5%, and only 10 months (1.2%) fell by 10% or more. The worst month in the whole sample was 1980-03 at -16.7%, and the best was 1980-01 at 48.4% — a reminder that the tails are genuinely two-sided on a monthly-average series.
Were there whole decades where owning gold cost money?
Yes, two. In the 1980s gold compounded at -40% with 50% of years negative, and in the 1990s it compounded at -28.4% with 70% of years negative. Compare the 1960s, where gold was an administered $35/oz price and compounded at -5.1% — the volatility numbers from that decade (2.1% average annual) reflect a fixed price, not a market. A strategy that could survive the 1980s and 1990s needed a drawdown budget measured in decades, not quarters.
How is gold doing right now?
The last month in the series, 2026-10, averaged $4105 per ounce. The 2026 year is partial (10 months) and stands at -13.6% with annualised volatility of 16.7%. The all-time monthly high was $5020 in 2026-02, so the last observation sits 18.2% below it. Current price levels are a statement about today, not about the risk in holding the asset through the 234-month decline that this history actually contains.
Can I use this price history to backtest a trading strategy?
Not directly. This is a price series with no spread, no commission, no slippage, no financing and no position sizing, and monthly averages are not prices you could have executed at. For a monthly return distribution, drawdown structure and decade behaviour it is the right source; for anything that assumes you traded, use the strategy pages, where the grid and trend rules are published line by line and the costs are counted. Note also that daily-close drawdowns are deeper than the ones here — sizing against monthly averages leaves you under-sized for a real intraday move.