Portfolio
A model book. It starts from a balanced mix, leans into what is working on the table, and steps aside from what is breaking. Since 2015 it made 23.0% a year with a worst fall of 18.2%, against 22.9% and 23.4% for the same mix left alone.
Rule-based, validation pending. Not advice.
Line by line · 2026-W43
| Line | Base | Timing | Stage | Keeps | Weight |
|---|---|---|---|---|---|
| Crypto | |||||
| Bitcoin | 15 | Hold | Normal | 75% | 11.2% |
| Large-cap crypto | 2 | Hold | Recovering | 75% | 1.5% |
| Small-cap crypto | 1.5 | Hold | Recovering | 75% | 1.1% |
| Equities | |||||
| Large caps | 29.5 | Add | Normal | 131% | 38.5% |
| Energy and power | 5 | 88% | 4.4% | ||
| Energy companies (XLE) | 2.5 | Add | Normal | 125% | 3.1% |
| Utilities (XLU) | 2.5 | Wait | Normal | 50% | 1.3% |
| Small caps (always 100%) | 3.5 | Hold | Normal | 100% | 3.5% |
| Hard assets | |||||
| Gold and silver | 14 | Reduce | Normal | 61% | 8.5% |
| Gold (GLD) | 11.5 | Reduce | Normal | 25% | 7.0% |
| Silver (SLV) | 2.5 | Reduce | Normal | 25% | 1.5% |
| Royalties and land | 3 | 30% | 0.9% | ||
| Franco-Nevada (FNV) | 0.8 | Wait | Normal | 50% | 0.4% |
| Wheaton (WPM) | 0.8 | Reduce | Normal | 25% | 0.2% |
| Royal Gold (RGLD) | 0.8 | Reduce | Normal | 25% | 0.2% |
| Texas Pacific Land (TPL) | 0.8 | Reduce | Post-peak | 25% | 0.2% |
| Collectibles (always held at 100%) | |||||
| Cards | 1.5 | Add | Normal | 100% | 1.5% |
| NFTs and digital art | 0.5 | Reduce | Post-peak | 100% | 0.5% |
| AI collectibles | 0.5 | 100% | 0.5% | ||
| Not on the table | |||||
| Trend funds (always 100%) | 6.5 | Add | n/a | 100% | 6.5% |
| Commodities | 2.5 | Add | above its 200-day | 100% | 2.5% |
| Defense | 15 | 18.9% | |||
Bold: the column that set Keeps. Faded: the other one, or a line the rule never cuts (small caps, collectibles, trend funds; commodities are in or out at the 200-day, their Timing shown for reference). How each word turns into a share is under How each weight is set.
How each weight is setTiming and Stage on the table; the more cautious wins.
How each weight is set
Every line starts at a base weight. Each holding then keeps a share of it, read from two columns of the table, Timing (the trend over months) and Stage (whether its multi-year run shows excess or a crash). The more cautious of the two wins. So a holding goes above its base, to 125%, only when both are strong. What a weak holding gives up goes to the lines that are working, up to each line's ceiling; the rest goes to defense, which never drops below 5%.
Bitcoin: base 15 x 75% (Timing Hold 75%, Stage Normal 125%) = 11.2
Not gated: US small caps and the trend funds. Collectibles are hold-only: never sold on the rule, since selling a card costs 10 to 20%; a weak read means no new money. Commodities are in above their 200-day and out below it. In the backtest since 2015, taking the more cautious of the two gave the steadiest ride of the versions tested.
If the path changesA · Muddle-through, base weights, no tilt.
The five paths and this book
Each column is one way things could go (the paths). The top half shows how much of this week's book is in lines that tend to do well or badly there, and how the book did in the closest past period. The bottom half is how the book would lean if we moved there, in points of a 100-point book.
| A Muddle-through we are here | B AI boom | C AI bust | D Fast takeoff | E War shock | |
|---|---|---|---|---|---|
| What it would do to this book | |||||
| In lines that tend to do well | 64% | 52% | 37% | 26% | 42% |
| In lines that tend to get hurt | 15% | 9% | 58% | 0% | 39% |
| Closest past period | none | none | none | none | none |
| The book then | |||||
| S&P 500 then | |||||
| How the book would lean | |||||
| Large caps | · | · | −8 | −4 | −6 |
| Energy and power | · | +3 | · | +3 | +3 |
| AI and chips | · | +3 | · | +3 | · |
| Defense stocks | · | · | · | · | +2 |
| Small caps | · | · | −2 | · | · |
| Gold and silver | · | · | +4 | +2 | +2 |
| Trend funds | · | · | +2 | · | · |
| Defense | · | −6 | +4 | −4 | −1 |
| base book | tilt | tilt | tilt | tilt | |
Which lines do well where is a judgment from the research, not a test. The past periods are stand-ins with the book's rules applied after the fact; the boom and the fast takeoff have none. A tilt applies half once a path arrives and in full at three-quarters of its signs, on its own at the monthly rebalance. The tilts are small because a backtest found bigger ones gained a little when the call was right and lost several times that when it was wrong. The boom and takeoff tilts add an AI and chips line (SMH, gated by the semis theme's Timing and Stage) and energy and power. The first boom tilt cut gold and lost even in 2023 to 2025; this one gained 7 points there and would have cost 2.5 in the 2000 bust. No period was a fast takeoff, so its tilt leans on the 2023 to 2025 AI rally as the nearest stand-in, which flatters any AI tilt.
When we'd sell the coreNone has fired.
The Timing rule trims a line when its trend breaks. These are different: the few things that would break the reason for holding gold or bitcoin at all. If one fires, the letter says so and the line is decided at review. Nothing sells on its own.
InflationThe rule is on, half of defense is in short TIPS.
Inflation
CPI, year on year. Above 3% and higher than six months earlier: half of defense goes to short TIPS. At 5% and rising, 3 points also move from US large caps to commodities and gold.
Ring: six months earlier. Dot: the latest reading.
Backtest and when it tradesSince 2015: 23.0% a year, Sharpe 1.37.
2015 to 2026
| Strategy | A year | Volatility | Sharpe | Worst fall |
|---|---|---|---|---|
| The book | 23.0% | 14.4% | 1.37 | -18.2% |
| Base weights, no rules | 22.9% | 16.1% | 1.24 | -23.4% |
| 60/40 | 8.8% | 10.3% | 0.67 | -20.6% |
| S&P 500 | 13.4% | 16.5% | 0.72 | -31.8% |
Weekly, the live rules on history rebuilt from public data, run 2026-10-28. Proxies stand in for lines without long histories, and small crypto and collectibles earn T-bills. One path through history, not proof.
Rebalanced on the first Sunday of the month. The inflation rule is read every week, and the book moves the week it changes. A line trades only when its target moves a point (half a point for crypto). Base weights and bands change only in January and July.