Strategies
Most of the upside. A third of the downside.
On prices it had never been tuned on, from 1 Dec 2025, the frozen rule, Breadth Trend, made +41.5% against +44.8% for holding all 32 markets. Its worst fall was −13.9%; holding's was −41.8%. The research desk is already hunting for the rule that replaces it.
01On offer today
3 listed rules, long or flat, across 32 markets.
Buy a market when its close is the highest of the last N candles. Sell when the close falls under the average of those N. Run across every market at once, an equal share each.
Being upgraded by the research desk
The desk is hunting a version that beats holding outright. This one keeps trading until a stronger rule passes every gate and replaces it.
Breadth Trend, Breadth Trend 150, Breadth Trend 100.
| Full history, 20 Aug 2024 to 28 Oct 2026 | Unseen months, from 1 Dec 2025 | ||||
|---|---|---|---|---|---|
| Rule | Return | Worst fall | Return | Worst fall | Fall vs holding |
| Breadth Trend Being upgraded by the research desk | +150.0% | −23.0% | +41.5% | −13.9% | 3.0× shallower |
| Breadth Trend 150 Being upgraded by the research desk | +133.6% | −23.7% | +27.0% | −16.6% | 2.5× shallower |
| Breadth Trend 100 Being upgraded by the research desk | +120.1% | −23.2% | +32.5% | −12.6% | 3.3× shallower |
| Holding all 32the line every rule is judged against | +101.7% | −64.8% | +44.8% | −41.8% | |
Full history: the platform catalogue, 32 markets, 4h, equal weight,
5 bps per side; source .context/app/live.json, 28 Oct 2026. Unseen months: walk-forward,
source docs/evidence/walk-forward-2026-10-23.json, 22 Oct 2026. All before funding, no slippage.
Breadth Trend
+150%vs +102%
Breadth Trend 150
+134%vs +102%
Breadth Trend 100
+120%vs +102%
Each rule (teal) against holding the same markets (grey), full history, same scale throughout.
02The hunt
11 strategy families on 1-hour, 2-hour, 4-hour and daily bars. One champion.
The desk declares every family and its settings before it measures anything, searches them on older prices, gives each winner one look at prices it has never seen, and freezes the survivors into paper trials with a fixed verdict day. A challenger that passes every gate is promoted on its own.
- strategy families
- 11
- bar widths, 1-hour to daily
- 4
- variants measured in this hunt
- 102
- challengers in paper trials
- 8
- Funding carry daily
- Ensembles 4-hour
- Multi-timeframe confirmation 4-hour
- Concentrated breakout books 4-hour · daily
- Concentrated momentum books 4-hour · daily
- Volatility targeting 4-hour · daily
- Cross-sectional momentum 4-hour · daily
-
Funding exit
Verdict in 60 days day 1 of 60
Verdict day 27 Dec 2026
-
Multi-timeframe confirmation
Verdict in 61 days starts soon
Verdict day 27 Dec 2026
-
Concentrated breakout books
Verdict in 61 days starts soon
Verdict day 27 Dec 2026
-
Volatility targeting
Verdict in 61 days starts soon
Verdict day 27 Dec 2026
-
Funding carry
Verdict in 61 days starts soon
Verdict day 28 Dec 2026
-
Concentrated breakout books
Verdict in 61 days starts soon
Verdict day 28 Dec 2026
-
Volatility targeting
Verdict in 61 days starts soon
Verdict day 28 Dec 2026
-
Cross-sectional momentum
Verdict in 61 days starts soon
Verdict day 28 Dec 2026
Funding carry
When longs pay heavy funding to stay long the trade is crowded and the carry eats the trend, so a daily breakout that steps aside while the last week cost more than a set yearly rate should keep most of the trend for less cost (the 4h version is the funding_exit template).
Ensembles
Rules that read different things (a channel high, the close n bars ago, two averages) give false exits on different bars, so an equal-weight blend should keep most of the trend while one rule's false exit is covered by the others, for a smoother book than any one of them.
Multi-timeframe confirmation
A 4h breakout against a falling daily trend is more often a bounce than a new trend, so holding 4h breakouts only while the last daily close is above its daily average should skip the bounces and keep the trends.
Concentrated breakout books
Crypto trends are carried by a few markets at a time, so giving the first few markets to break out to a new n-bar high a full slot each, instead of a thin 1/N slice of every market, should keep the big moves at size; the price is a deeper drawdown, which the bar weighs.
Concentrated momentum books
A market whose close is above its close n bars ago is more often in a trend than not, and crypto trends are carried by a few markets at a time, so a full slot each for the first few markets that turn up over n bars, instead of a thin 1/N slice of every market, should keep the big moves at size; the price is a deeper drawdown, which the bar weighs.
Volatility targeting
Crypto markets fall together, so a book of separate breakouts carries the most risk exactly when they all turn at once; scaling the whole book down while what it holds is moving violently should cut the deepest drawdowns by more than the return.
Cross-sectional momentum
Crypto markets that outran the rest over the last weeks tend to keep outrunning them for a while (the cross-sectional momentum documented in crypto returns), so holding the strongest few, and only while they are actually rising, should earn more than holding them all.
Families, bar widths and paper windows: docs/evidence/families-2026-10-28.json, 28 Oct 2026. Paper trials trade no money, and each
verdict is scored once, when its window closes.
03Forward test
The frozen rule's universe and settings were fixed on 15 Sep 2026 and a forward window was set before any of it was collected. It closes on 14 Nov 2026. The tooling refuses to report early, on purpose, and the result is published here when it lands.
The method: every test, the checks we ran against ourselves and the fine print →