What every score meant next: 55 years of history

The score since 2019, under the strictest rules

This chart starts in October 2019. It only draws months where all six parts existed AND each had ten years of its own history to be ranked against. Price jumpiness only starts in 2009, so the window opens in 2019. The longer reconstruction back to 1971 is the chart further down, where the parts that did not exist yet have their weight shared out. This one is stricter, and stricter changes the numbers. The same month can read a few points lower here than in the site's own monthly line. Publication lags and longer minimum histories are enforced here. Both series are published in full, so the gap itself is checkable. Use the buttons to zoom in. Tap any point for its date, score, and zone. The dashed line is the rebuilt history. The solid line is the live daily reading. The solid line gains one new day's reading each day.

7 days30 days90 days1 yearAll

View the underlying data as a table

Date	Score	Zone	Source
2019-10-31	58	Mixed	reconstructed
2019-11-30	60	Favorable	reconstructed
2019-12-31	58	Mixed	reconstructed
2020-01-31	57	Mixed	reconstructed
2020-02-29	52	Mixed	reconstructed
2020-03-31	49	Mixed	reconstructed
2020-04-30	48	Mixed	reconstructed
2020-05-31	52	Mixed	reconstructed
2020-06-30	48	Mixed	reconstructed
2020-07-31	47	Mixed	reconstructed
2020-08-31	49	Mixed	reconstructed
2020-09-30	48	Mixed	reconstructed
2020-10-31	49	Mixed	reconstructed
2020-11-30	52	Mixed	reconstructed
2020-12-31	51	Mixed	reconstructed
2021-01-31	53	Mixed	reconstructed
2021-02-28	55	Mixed	reconstructed
2021-03-31	61	Favorable	reconstructed
2021-04-30	64	Favorable	reconstructed
2021-05-31	58	Mixed	reconstructed
2021-06-30	61	Favorable	reconstructed
2021-07-31	61	Favorable	reconstructed
2021-08-31	63	Favorable	reconstructed
2021-09-30	63	Favorable	reconstructed
2021-10-31	61	Favorable	reconstructed
2021-11-30	56	Mixed	reconstructed
2021-12-31	59	Mixed	reconstructed
2022-01-31	54	Mixed	reconstructed
2022-02-28	50	Mixed	reconstructed
2022-03-31	47	Mixed	reconstructed
2022-04-30	47	Mixed	reconstructed
2022-05-31	52	Mixed	reconstructed
2022-06-30	49	Mixed	reconstructed
2022-07-31	56	Mixed	reconstructed
2022-08-31	55	Mixed	reconstructed
2022-09-30	49	Mixed	reconstructed
2022-10-31	50	Mixed	reconstructed
2022-11-30	53	Mixed	reconstructed
2022-12-31	48	Mixed	reconstructed
2023-01-31	49	Mixed	reconstructed
2023-02-28	55	Mixed	reconstructed
2023-03-31	48	Mixed	reconstructed
2023-04-30	45	Mixed	reconstructed
2023-05-31	42	Mixed	reconstructed
2023-06-30	47	Mixed	reconstructed
2023-07-31	47	Mixed	reconstructed
2023-08-31	50	Mixed	reconstructed
2023-09-30	49	Mixed	reconstructed
2023-10-31	42	Mixed	reconstructed
2023-11-30	44	Mixed	reconstructed
2023-12-31	45	Mixed	reconstructed
2024-01-31	47	Mixed	reconstructed
2024-02-29	48	Mixed	reconstructed
2024-03-31	40	Mixed	reconstructed
2024-04-30	35	Unfavorable	reconstructed
2024-05-31	36	Unfavorable	reconstructed
2024-06-30	35	Unfavorable	reconstructed
2024-07-31	35	Unfavorable	reconstructed
2024-08-31	34	Unfavorable	reconstructed
2024-09-30	35	Unfavorable	reconstructed
2024-10-31	32	Unfavorable	reconstructed
2024-11-30	30	Unfavorable	reconstructed
2024-12-31	31	Unfavorable	reconstructed
2025-01-31	31	Unfavorable	reconstructed
2025-02-28	33	Unfavorable	reconstructed
2025-03-31	35	Unfavorable	reconstructed
2025-04-30	36	Unfavorable	reconstructed
2025-05-31	35	Unfavorable	reconstructed
2025-06-30	39	Unfavorable	reconstructed
2025-07-31	39	Unfavorable	reconstructed
2025-08-31	40	Mixed	reconstructed
2025-09-30	39	Unfavorable	reconstructed
2025-10-31	39	Unfavorable	reconstructed
2025-11-30	38	Unfavorable	reconstructed
2025-12-31	37	Unfavorable	reconstructed
2026-01-31	38	Unfavorable	reconstructed
2026-02-28	39	Unfavorable	reconstructed
2026-03-31	33	Unfavorable	reconstructed
2026-04-30	35	Unfavorable	reconstructed
2026-05-31	33	Unfavorable	reconstructed
2026-06-30	31	Unfavorable	reconstructed
2026-07-31	32	Unfavorable	reconstructed
2026-08-05	33	Unfavorable	published
2026-08-06	33	Unfavorable	published
2026-08-07	33	Unfavorable	published
2026-08-08	33	Unfavorable	published
2026-08-09	33	Unfavorable	published
2026-08-10	33	Unfavorable	published
2026-08-11	29	Unfavorable	published
2026-08-12	29	Unfavorable	published
2026-08-13	29	Unfavorable	published
2026-08-14	30	Unfavorable	published
2026-08-15	30	Unfavorable	published
2026-08-16	30	Unfavorable	published
2026-08-17	30	Unfavorable	published
2026-08-18	31	Unfavorable	published
2026-08-19	30	Unfavorable	published
2026-08-20	30	Unfavorable	published
2026-08-21	29	Unfavorable	published
2026-08-22	29	Unfavorable	published
2026-08-23	29	Unfavorable	published
2026-08-24	29	Unfavorable	published
2026-08-25	29	Unfavorable	published
2026-08-26	29	Unfavorable	published
2026-08-27	30	Unfavorable	published
2026-08-28	31	Unfavorable	published
2026-08-29	31	Unfavorable	published
2026-08-30	31	Unfavorable	published
2026-08-31	31	Unfavorable	published
2026-09-01	32	Unfavorable	published
2026-09-02	33	Unfavorable	published
2026-09-03	32	Unfavorable	published
2026-09-04	31	Unfavorable	published
2026-09-05	31	Unfavorable	published
2026-09-06	31	Unfavorable	published
2026-09-07	31	Unfavorable	published
2026-09-08	33	Unfavorable	published
2026-09-09	32	Unfavorable	published
2026-09-10	33	Unfavorable	published
2026-09-11	33	Unfavorable	published
2026-09-12	33	Unfavorable	published
2026-09-13	33	Unfavorable	published
2026-09-14	33	Unfavorable	published
2026-09-15	33	Unfavorable	published
2026-09-16	33	Unfavorable	published
2026-09-17	32	Unfavorable	published
2026-09-18	32	Unfavorable	published
2026-09-19	32	Unfavorable	published
2026-09-20	32	Unfavorable	published

What the Barometer said, and what gold did next

The gold line is what one ounce of gold cost, month by month, since 1971. The colour behind it is what the Barometer would have read that month, rebuilt after the fact. It uses the five bands from the dial: red for the worst conditions, green for the best.

Green marks the months it read as good conditions, red the months it read as poor ones. What followed each of them is the line running above.

Two things about how this is drawn. The price axis doubles at each step instead of adding a fixed amount. So a rise from 100 to 200 dollars takes the same space as one from 1,000 to 2,000. It is the only fair way to show 55 years. Without it, the 1980 spike disappears.

The prices are month-end, from the World Bank. It is the only series reaching back to 1971 that we are allowed to republish. We do not hold the rights to a daily series reaching that far back.

Decade by decade

Is 32 out of 100 unusual, or is it an ordinary Tuesday? Here is today set against the whole record, against this decade, and against the last one.

The same reconstruction, averaged. Has the Barometer simply drifted gloomier as gold got dearer? It has not. The gloomiest decade is the 1980s, when savings paid enormously well after inflation. The brightest is the 2010s.

Read this first

• The top zone (80 to 100 "Historically very favorable") has never happened. Across 668 monthly readings since 1971, the score has never crossed 80. We do not publish forward numbers for that zone.
• Over one year, the zones do not line up in the expected order. Median 1-year returns: Favorable 5.8%, Mixed 6.5%, Unfavorable 3.7%. Short-run moves and the score point opposite ways. Read as a 1-year buy signal, the instrument fails.
• Over five years, the long-run thesis does hold. Median 5-year real returns: Favorable 48.5%, Unfavorable -15.8%. Buyers who entered when the reading was Favorable kept and grew their purchasing power. Buyers who entered when it was Unfavorable lost ground.
• The cash-after-inflation part has broken down. We measure the link between the real rate itself and gold's next 12 months, on a scale from -1 to +1. Minus 1 means the two always moved opposite ways. The claim expects a negative link: low rates, better gold ahead. The link was +0.237 for 1971 to 2000, so the claim failed that whole era. It was -0.401 for 2001 to 2021, then -0.851 for 2022 to 2024. The direction the claim expects has only shown up after 2000. The part is still in the score because its economic reason is coherent. See where this framework has been wrong.

1. How the test was built

Monthly readings from January 1971 to August 2026: 668 in total. The price as printed comes from the World Bank Pink Sheet, monthly, in US dollars per troy ounce. The real price is that same price with official US inflation stripped out.

We applied the real-world publication lag to every source. The official inflation figure (CPI) takes 30 days. Trader bets take 3 days. The World Bank Pink Sheet takes 7 days. IMF reserves data takes 60 days.

At each date, every part is ranked in its own history up to that day only. The retail premiums part is absent from the test: we only began collecting on 2026-08-04. The parts available on a date split the full weight between them.

For every date, we then computed what gold did over the next 1, 3, 5, and 10 years. We also computed its worst drop along the way inside each period. And we compared buying a little each month against a lump sum.

2. What each zone meant for a buyer

Middle results for every monthly reading in each zone. The column counting how often it rose over a year asks a different question: how often gold beat its own typical year. That yardstick is 5.9%, the middle one-year result across the whole record since 1971.

Every figure is the middle result, not an average, so one freak year cannot drag it around. Nothing here is a forecast. It is what followed, on the days that already happened.

When the before-inflation figure is positive but the after-inflation one is not, inflation ate the gain.

How much these middle results move

The buyer table above is made of middle results across overlapping stretches. Replaying the record 10,000 times in blocks, as the method page describes, shows how much each figure moves.

The one claim that survives every replay: in 100% of 10,000 replays, the Favorable zone beat the Unfavorable zone at five years after inflation. At one year, 3 of the 4 ranges hold both a loss and a gain, so they point nowhere.

Going in at once, against spreading it over the first year

The last three columns compare two ways of putting the same money in. One buys the whole amount on the reading day. The other buys a twelfth of it each month for a year, then holds what it has bought. Both are valued five years after the reading day.

In 3 of the 4 zones with a record, spreading it out ended slightly behind going in at once. The exception is the unfavorable zone, where spreading ended ahead: prices there kept falling, so the later purchases cost less. Elsewhere, gold ended a year higher more often than lower, so the later purchases usually cost more than the first one.

One row looks like it contradicts that, and it is worth saying why it does not. In the historically very unfavorable zone the all-at-once column reads 9.2 and the spread column reads 13.5, yet the gap column reads -3.1. The first two are each the middle of their own column, taken across every month in the zone. The last one compares the two inside each single month first, and then takes the middle of those comparisons. They are two different questions and they can point different ways.

Two things that gap does not contain. These are middle results across five-year stretches that overlap heavily, so they are not independent of each other. And spreading purchases is not done to end up with more. It is done so that no single day decides the outcome, and that is a kind of safety these columns measure nothing about.

3. What each zone meant for a seller

The instrument reads both ways from one score. Two seller numbers come out of each zone. First, the opportunity cost of selling: the median return you would have missed. Second, the drop avoided by selling: the median fall you would have skipped. When history showed no gain to give up, or no fall to escape, the cell reads 0.0%.

These read the same instrument from the other side. A seller in a favorable zone gave up gains. A seller in an unfavorable zone escaped falls. Both are middle results since 1971. The seller's question has its own page.

4. Where the framework failed

This section is on the page for a reason. The three parts below put numbers on the failures rather than paper over them.

4.1 The cash-after-inflation link, era by era

How closely the cash-after-inflation part tracked gold's next 12 months, split into three eras. Positive means the part pointed at higher returns.

We do not drop the part. The economic reason is coherent, and two decades of it worked. But the disclosure is numeric, not adjectival.

4.2 The 2022 to 2024 break, month by month

Nine sample months from that stretch. The full monthly record is in the public archive: see the data page.

4.3 Do the zones line up in the right order?

If the score is well-oriented, medians should rise from the lowest zone to the highest. They do not always. The gaps are listed here.

• 1-year: Historically very unfavorable 9.67% beat Unfavorable 3.66%. Mixed 6.51% beat Favorable 5.8%. Two mis-orderings.
• 3-year: Historically very unfavorable -1.01% beat Unfavorable -5.12%. One mis-ordering.
• 5-year: Historically very unfavorable 9.21% beat Unfavorable -0.25%. One mis-ordering.
• 10-year: No mis-orderings.

The 27 months in the lowest zone are all clustered in specific historical episodes. What followed those months tells you about those few episodes, not about the next month that lands in the zone. That is why we publish the count of months per zone.

5. Does the score hold up under different choices?

5.1 Move the zone edges by 5 points

The official edges are 80/60/40/20. We recomputed with edges +5 (85/65/45/25) and -5 (75/55/35/15). The check: does the story survive? A higher zone should still mean a higher five-year result. It mostly does, except the bottom zone, which sits out of order here too.

5.2 Give every part an equal weight

The design weights are 25/20/15/10/10/10/10. We reran the test with an equal weight for each active part, leaving retail premiums out. The order across zones should look about the same if the score does not rely on any one weight.

6. When each part first became usable

Each part joins the score as soon as its data is usable. In each era, the active parts split the full weight between them. The retail premiums part is missing from the whole test because collection only started on August 4, 2026. Below is the first month each part was active in the test.

7. The most recent 24 readings

Last 24 monthly readings from the rebuilt history and the live engine. The full archive lives at /data/history.csv and is mirrored at github.com/thegoldbarometer/data.

8. What this test is not

It is not a trading strategy. There are no entry rules and no exit rules. There are no rules for how much to put in, no trading costs, and no schedule for adjusting along the way. It is a plain record of what gold did after every historical reading. Every number is a median across the months in that bucket, from a single 1971-to-today sample.

It is not fully tested on years the rules never saw, either. Each date's ranks come from data known that day only. But the direction of each part was set by us. Low cash after inflation is better. Weak dollar is better. Cheap price is better. We set those directions based on published economic evidence that itself built up over the same period. Any test with only one history to check against shares this limit.

Source note. Nominal gold price is the World Bank Pink Sheet (CC BY 4.0). Inflation is stripped out with the main US consumer price index (CPI-U). Exact series codes are on the provenance page. We do not use LBMA or IBA gold price series, per our data-rights rules. See provenance.