How the premises are linked
Each site checks one premise: gold keeps rising, the weak yen lasts, AI spending keeps going, and so on. If two premises have always risen and fallen together, believing both is really one bet placed twice. Here each premise is replaced by the market move that pays off when it is right, and past co-movement shows which are the same bet, what happened to the others when one broke, and how many separate bets there really are.
In three lines
- BetsThe 7 premises add up to 5 separate bets (4–5 depending on how you look). Doubled-up bets: Gold + Miners and AI + Power.
- Hedge“Gold keeps rising” and “The weak yen lasts” move in opposite directions (−0.35). In the months when “The weak yen lasts” broke down worst, Gold moved +3.7% on average.
- CrashesIn the worst months for stocks (S&P 500 −8.0% on average), AI −11%, Power −11%, Commodities −4.3% and Miners −4.1% fell together; only Gold (+1.3%) rose. When it matters, separate bets tend to break together.
The bets
Premises whose monthly moves line up at 0.4 or more (in either direction) are counted as one bet. Today's call from each site is shown alongside.
Bet 1: Gold + Miners
- Gold keeps rising ▲ Watch
- Miners beat gold ● Case holds
Month-to-month alignment: Gold–Miners 0.55.
Gold is on watch today, and Miners in the same bet is no separate safety net.
Bet 2: AI + Power
- AI spending keeps going ▲ Watch
- Power is the bottleneck ● Case holds
Month-to-month alignment: AI–Power 0.66.
AI is on watch today, and Power in the same bet is no separate safety net.
Bet 3: Commodities
- A commodity supercycle ▲ Watch
No other premise reaches 0.4 (closest: Power, 0.21).
Bet 4: Weak yen
- The weak yen lasts ▲ Watch
No other premise reaches 0.4 (closest: Gold, −0.35).
Bet 5: US–China
- US leads chips, China leads power ● Case holds
No other premise reaches 0.4 (closest: AI, 0.24). Only 57 months of data, from Feb 2022.
Every pair, side by side
1 means they always move the same way together, −1 always opposite, 0 unrelated. Blue = same direction, orange = opposite.
| Gold | Miners | AI | Power | Cmdty | JPY | US/CN | |
|---|---|---|---|---|---|---|---|
| Gold | — | 0.55 | 0.01 | 0.07 | 0.03 | −0.35 | 0.04 |
| Miners | 0.55 | — | 0.24 | 0.18 | 0.07 | −0.24 | −0.19 |
| AI | 0.01 | 0.24 | — | 0.66 | 0.16 | 0.09 | 0.24 |
| Power | 0.07 | 0.18 | 0.66 | — | 0.21 | 0.12 | 0.06 |
| Cmdty | 0.03 | 0.07 | 0.16 | 0.21 | — | 0.12 | 0.12 |
| JPY | −0.35 | −0.24 | 0.09 | 0.12 | 0.12 | — | −0.13 |
| US/CN | 0.04 | −0.19 | 0.24 | 0.06 | 0.12 | −0.13 | — |
If one breaks, what happens to the others?
Each row takes the worst 10% of months for one premise's yardstick and shows how the others moved on average in those same months. Figures are average monthly moves in %; plus = a tailwind for that premise, minus = a headwind. Each row uses 25 months (US–China: 6).
| Broke | Gold | Miners | AI | Power | Cmdty | JPY | US/CN |
|---|---|---|---|---|---|---|---|
| Gold | −7.9 | −7.1 | +1.3 | −1.0 | 0.0 | +1.3 | +2.3 |
| Miners | −3.0 | −12 | −3.4 | −2.9 | −2.5 | +1.2 | +4.3 |
| AI | +1.5 | −4.8 | −13 | −8.0 | −3.2 | −0.8 | −2.6 |
| Power | −0.1 | −3.7 | −7.4 | −12 | −3.5 | −0.6 | +1.3 |
| Cmdty | +0.6 | −1.2 | −1.9 | −3.9 | −12 | −0.5 | · |
| JPY | +3.7 | +2.9 | −2.0 | −2.1 | −1.6 | −4.2 | +0.6 |
| US/CN | +2.9 | +5.2 | +0.3 | +2.0 | −1.2 | +0.9 | −16 |
The biggest fall in each yardstick since Jun 2006, and what the others did meanwhile (the miners figure is miners against gold; the weak-yen figure is USD/JPY):
- Gold keeps rising: from Aug 2011 to Dec 2015 its yardstick fell −42%. Over the same 52 months: Miners −62%, AI +87%, Power +23%, Commodities −56%, Weak yen +57%. Since the low it has recovered to 3.7x.
- Miners beat gold: from Aug 2006 to Sep 2015 its yardstick fell −81%. Over the same 109 months: Gold +78%, AI +34%, Power −25%, Commodities −26%, Weak yen +2.4%. Since the low it has recovered to 1.6x.
- AI spending keeps going: from Jun 2007 to Nov 2008 its yardstick fell −60%. Over the same 17 months: Gold +26%, Miners −44%, Power −50%, Commodities −14%, Weak yen −23%. Since the low it has recovered to 53.9x.
- Power is the bottleneck: from May 2006 to Feb 2009 its yardstick fell −67%. Over the same 33 months: Gold +50%, Miners −43%, AI −57%, Commodities −23%, Weak yen −13%. Since the low it has recovered to 26.0x.
- A commodity supercycle: from Jul 2008 to Apr 2020 its yardstick fell −70%. Over the same 141 months: Gold +86%, Miners −60%, AI +408%, Power +8.7%, Weak yen −0.5%. Since the low it has recovered to 3.2x.
- The weak yen lasts: from Jun 2007 to Oct 2011 its yardstick fell −39%. Over the same 52 months: Gold +164%, Miners −41%, AI −23%, Power −30%, Commodities +38%. Since the low it has recovered to 2.1x.
- US leads chips, China leads power: from Jan 2022 to Aug 2025 its yardstick fell −76%. Over the same 43 months: Gold +96%, Miners +6.7%, AI +63%, Power +357%, Commodities −21%, Weak yen +28%. Since the low it has recovered to 1.9x.
When stocks crash
The worst 10% of months for the S&P 500 (25 months, −8.0% on average).
| Premise | Average move | Share of those months it fell |
|---|---|---|
| Gold keeps rising | +1.3% | 40% |
| Miners beat gold | −4.1% | 76% |
| AI spending keeps going | −11% | 100% |
| Power is the bottleneck | −11% | 96% |
| A commodity supercycle | −4.3% | 72% |
| The weak yen lasts | −0.6% | 64% |
| US leads chips, China leads power | −1.5% | 67%6 months only |
Does the answer change with the view?
The same grouping over different periods and at quarterly frequency. The effective number counts how many independent moves there are once overlaps are netted out: it would be 7 if every premise moved on its own and 1 if all moved as one.
| View | Bets | Effective |
|---|---|---|
| Monthly, full period | Gold + MinersAI + PowerCommoditiesWeak yenUS–China | 5.2 |
| Last 10 years (monthly) | Gold + MinersAI + PowerCommoditiesWeak yenUS–China | 5.1 |
| 2006–2015 (monthly) | Gold + MinersAI + PowerCommoditiesWeak yenUS–China (too little data) | 4.1 |
| Quarterly moves | Gold + Miners + Weak yenAI + PowerCommoditiesUS–China (too little data) | 3.9 |
From the record of calls
No two calls have changed on the same day since recording began (2026-09-29). The record of calls is still short, so the links on this page are measured with market moves instead.
Yardsticks and method
| Premise | Yardstick (up = the premise is winning) |
|---|---|
| Gold keeps rising | Gold futures (USD)From Jun 2006, 245 months |
| Miners beat gold | Gold-miner ETF (GDX) ÷ goldFrom Jun 2006, 245 months |
| AI spending keeps going | US semiconductor index (SOX)From Jun 2006, 245 months |
| Power is the bottleneck | Three power-equipment stocks (Eaton, Hitachi, Furukawa Electric; USD)From Jun 2006, 245 months |
| A commodity supercycle | World Bank commodity price index (all, monthly average)From Jun 2006, 245 months |
| The weak yen lasts | USD/JPY (up = weaker yen)From Jun 2006, 245 months |
| US leads chips, China leads power | Half US chip stocks (NVIDIA, AMD, TSMC) minus Chinese ones (SMIC, Cambricon), half Chinese power-equipment stocks (NARI, Dongfang Electric, TBEA) minus US power stocks (Vistra, Constellation)From Feb 2022, 57 months |
- Monthly change in each yardstick from month-end values (the current month is left out until it ends). Data up to Oct 2026, recalculated every morning from the data each site already holds. No AI is used.
- Alignment is the correlation of monthly moves, shown only for pairs with at least 36 overlapping months. Bets are formed by merging the premises whose average alignment (ignoring direction) is highest, and stopping below 0.4.
- The effective number of bets is the participation ratio of the eigenvalues of the correlation matrix: (sum of eigenvalues)² ÷ sum of squared eigenvalues.
- The World Bank commodity index is a monthly average, which understates how it lines up month to month with the others; the quarterly view corrects for that. The US–China yardstick starts in 2022, so its figures are indicative only.
- The yardsticks stand in for the premises; each site judges its premise on its own fundamentals (see each site's method page).