Source-labelled market dataSource, observation time and freshness are shown with every quoteData statusAll pages
GoldObserve

HISTORICAL LOSS FREQUENCY / OVERLAPPING WINDOWS / NOT A FORECAST

Gold Holding-Period Risk

Count completed historical gold windows that ended below their starting monthly average, then separate endpoint loss from drawdown and investor costs.

THE SHORT ANSWER

A longer horizon changed the historical loss frequency, but it never removed investment risk

GoldObserve counts every completed 1-, 3-, 5-, 10- and 20-year monthly-average window that ended below its starting value. This produces a transparent historical loss frequency, not a forecast probability. Adjacent windows overlap, costs are excluded and the sample spans several monetary regimes. Use the result to understand endpoint sensitivity, not to promise that a future holding period will be profitable.

1 YEAR33.4%

263 losing endpoints among 787 completed windows.

3 YEARS28.2%

215 losing endpoints among 763 completed windows.

5 YEARS24.9%

184 losing endpoints among 739 completed windows.

10 YEARS19.4%

132 losing endpoints among 679 completed windows.

FIVE-HORIZON RISK PROFILE

How did the middle range and losing endpoints change with time held?

Compare all five horizons on one zero-centered scale. The chart separates the middle 50%, central 90%, median, observed extremes and historical loss frequency instead of reducing each horizon to one percentage.

SELECTED HORIZON5Y · 739 windows

184 ended below their starting monthly average.

HISTORICAL LOSS FREQUENCY24.9%

Descriptive overlapping-window frequency, not forecast probability.

MEDIAN ANNUALIZED+4.6%

Middle observed endpoint result for this exact horizon.

P5–P95 central 90%Q1–Q3 middle 50%MedianZero annualized return
-20%+0%+20%+40%+60%1Y33.4% lostn=7873Y28.2% lostn=7635Y24.9% lostn=73910Y19.4% lostn=67920Y9.3% lostn=559

How to read it: the thin line contains the empirical P5–P95 range, the thick line contains Q1–Q3 and the dot is the median. The loss label counts total endpoint returns below zero, so it is not read from the annualized median alone. Extremes remain in the table to avoid compressing the central comparison.

Central 90%-6.6% to +29.7%
Middle 50%+0.0% to +16.7%
Worst endpoint-14.8%1980-011985-01
Best endpoint+39.4%1969-121974-12
Annualized endpoint-return summaries. P5–P95 and Q1–Q3 are empirical sample intervals; adjacent windows overlap.
HorizonWindowsLosing endpointsLoss frequencyMinimumP5Q1MedianQ3P95Maximum
1Y78726333.4%-36.4%-18.6%-4.1%+2.9%+19.4%+54.9%+197.4%
3Y76321528.2%-15.8%-9.6%-1.3%+3.2%+14.7%+41.9%+72.3%
5Y73918424.9%-14.8%-6.6%+0.0%+4.6%+16.7%+29.7%+39.4%
10Y67913219.4%-5.6%-3.4%+0.7%+6.4%+13.8%+21.4%+34.4%
20Y559529.3%-4.4%-1.5%+3.9%+7.5%+11.4%+13.6%+15.9%

HOLDING-PERIOD PATH LABORATORY

A profitable endpoint can still contain a difficult path

Compare the worst, median-nearest and best endpoint windows with any completed starting month. Every path begins at 100, while the table keeps the endpoint return and the largest interim drawdown separate. Use pointer, touch or arrow keys to inspect the same elapsed month across all four paths.

SELECTED ENDPOINT+17.6% annualized

2021-072026-07 · +125.3% total.

INTERIM MAXIMUM DRAWDOWN-18.9%

Lowest distance below this path's own prior monthly-average peak, at 2026-07.

INSPECTED MONTH60 of 60

2026-07 · index 225.3.

Worst endpoint: 1980-01Median endpoint: 1985-03Best endpoint: 1969-12Selected window: 2021-07
32641282565121024Start5Y endpointMonth 60

How to read it: every line begins at 100, so the chart compares paths rather than dollar price levels. The vertical index axis is logarithmic: equal vertical distances represent equal proportional changes, which keeps very large long-run gains from flattening the other paths. Endpoint ranking does not rank the loss experienced along the way. The gold line is your selected window; arrow keys, pointer and touch move the shared inspection month.

Representative and selected 5Y paths. Duplicate dates are retained when the selected window is also one of the three endpoint representatives.
PathStartEndTotal returnAnnualizedMaximum drawdownDrawdown monthIndex at inspected month
Worst endpoint1980-011985-01-55.1%-14.8%-55.1%1985-0144.9
Median endpoint1985-031990-03+25.2%+4.6%-25.5%1989-09125.2
Best endpoint1969-121974-12+425.7%+39.4%-20.8%1973-11525.7
Selected window2021-072026-07+125.3%+17.6%-18.9%2026-07225.3

INTERACTIVE OUTCOME PATH

See where positive and negative annualized windows occurred

The zero line separates windows ending above and below their starting monthly average. Change the holding period and inspect the exact endpoint with crosshairs. A point above zero can still contain a substantial interim drawdown.

LOCAL LICENSED HISTORY

5-year rolling annualized return

FIRST OBSERVATION1965-01+0.00%
LATEST OBSERVATION2026-07+17.64%
LOWEST-14.80%Inside the selected sample
HIGHEST+39.36%Inside the selected sample
1965-011995-102026-07
Observation2026-07
5-year rolling annualized return+17.64%
Series basisMonthly averages

Crosshairs snap to a calculated observation. Monthly averages smooth intramonth highs, lows and drawdowns; this chart is not a daily close series. Historical windows overlap and are not independent forecasts.

This export contains GoldObserve-derived or source-cleared observations with citation metadata.

Source: World Bank Commodity Price Data (The Pink Sheet), monthly nominal USD per troy ounce, CC BY 4.0. GoldObserve serves the bundled local series; page loads do not download the upstream workbook.

HOLDING-PERIOD RETURN DISTRIBUTION

Compare the shape, middle range and tails across five exact horizons

Select 1, 3, 5, 10 or 20 years, then inspect annualized-return buckets and the empirical box. The loss-frequency label still uses total start-to-end return, while the histogram annualizes each completed window so horizons remain comparable.

739 completed overlapping windows · 24.9% ended below the start
Selected bucket-15% to -10%Observed frequency1.22% (9)

EMPIRICAL BOX AND TAIL MARKERS

P5 -6.6%Median 4.6%P95 29.7%

Each observation annualizes one exact completed holding window. Adjacent windows overlap, so the histogram describes historical endpoints rather than independent trials or a future probability.

Source: World Bank Commodity Price Data (The Pink Sheet), local monthly nominal USD per troy ounce. Source workbook updated 2026-08-04; retrieved 2026-08-07. These charts describe history and are not price forecasts.

LOSS-FREQUENCY TABLE

Count completed windows before interpreting the percentage

Historical nominal USD monthly-average outcomes. Rows overlap and are not statistically independent.
Holding periodCompleted windowsLosing windowsHistorical loss frequencyWorst annualized resultWorst start / end
1 year78726333.4%-36.4%1980-07 to 1981-07
3 years76321528.2%-15.8%1980-10 to 1983-10
5 years73918424.9%-14.8%1980-01 to 1985-01
10 years67913219.4%-5.6%1980-07 to 1990-07
20 years559529.3%-4.4%1980-09 to 2000-09

TWO DIFFERENT RISKS

Ending below the start and suffering a drawdown are not the same event

MeasureQuestion answeredWhat it can miss
Holding-period lossWas the ending monthly average below the starting average?How painful the path was before the endpoint
Maximum drawdownHow far did the series fall from a prior peak?Whether the selected start-to-end result was positive
Investor lossDid actual sale proceeds exceed total acquisition and carrying cost?Nothing, if every cash flow and tax effect is reconciled correctly

Use the drawdown explorer for the second question and the gold investment return calculator for the third.

WHAT CHANGES THE ANSWER

Currency, entry method and product costs can reverse a benchmark result

LOCAL CURRENCYA non-USD investor holds gold and an exchange-rate exposure.

Local-currency appreciation can reduce a positive USD result; depreciation can increase it.

PHYSICAL GOLDPremiums and exit spreads create a higher break-even price.

The shorter the horizon, the more a fixed round-trip spread can dominate the benchmark move.

RECURRING PURCHASESCash-flow timing creates a different average cost.

A monthly contribution plan cannot be represented by one start and one end price.

OPPORTUNITY COSTEnding above the start does not prove gold was the best allocation.

A complete decision compares alternatives, liquidity needs and risk tolerance on matched dates.

Source and study limitsWorld Bank monthly history · CC BY 4.0

SOURCE, LICENSE & LIMITS

The study uses one consistent monthly history

The source is the World Bank Commodity Price Data (The Pink Sheet), licensed CC BY 4.0. It contains 799 monthly gold observations from 1960-01 through 2026-07. The source workbook was published 2026-08-04, retrieved 2026-08-07, and preserved with SHA-256 7902a77505ebdc5d202ce65f666c2ee1b04b626f042d7738ed3e6f7d112c8433.

The values are nominal US-dollar monthly averages per troy ounce. They are not daily closes, dealer quotes, London auction prices or exact transaction prices.

CONTINUE THE RESEARCH

Separate holding-period evidence from a personal investment result

FREQUENTLY ASKED QUESTIONS

Questions about gold holding-period evidence

What does gold holding-period risk mean here?

It means the share and range of completed historical monthly-average windows that ended below their starting value. It is not a forecast probability, value-at-risk model or guarantee.

Why call it historical loss frequency rather than loss probability?

The rolling windows overlap and the market regimes are not identically distributed. Frequency accurately describes this sample; probability would imply a stronger statistical model than the page provides.

Can a positive endpoint hide a severe drawdown?

Yes. A window can finish above its start after spending years below a prior peak. Endpoint loss frequency and maximum drawdown answer different questions.

Does a longer holding period remove gold risk?

No. A longer horizon changes the historical distribution but does not eliminate currency, price, opportunity-cost, liquidity, product-cost or regime risk.

Why can my loss differ from the study?

Your acquisition premium, sale discount, currency, fees, tax and exact dates can turn a positive benchmark window into a negative personal result, or vice versa.