Strategies

Every rule Archon offers is long or flat: it either holds a market or holds nothing in it. None of them shorts. They trade at 1x unless you choose Turbo, which raises leverage up to the strategy’s measured maximum. They read one thing, the closing prices of 4-hour candles, and they are frozen: the code that decides is the same code that was tested, and it does not change without a new test.

The research desk page shows the numbers and every test behind them. This page explains the rules and how to choose.

The rules

Breadth Trend

Enter a market when its 4-hour close is a new N-candle closing high. Exit when the close falls below the N-candle average.

Run across all 32 markets at once, an equal share each. The idea is that across many markets a few strong trends pay for the many small losses, and no single market decides the result.

Rule	N	In days	Character
Breadth Trend (the frozen rule)	200	about 33	Slowest, fewest trades, shallowest falls of the listed rules
Breadth Trend 150	150	25	Enters a little sooner, exits a little sooner
Breadth Trend 100	100	about 17	The fastest breakout that still cleared the bar

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.

Retired by the gates

The textbook trend filters, on the same 4-hour candles and the same 32 markets, were listed until the research desk’s daily recheck retired them on 28 Oct 2026. Accounts already running one keep it until they switch; nobody new is offered it.

Rule	Holds a market while
Above the 200	its close is above its 200-candle simple moving average
Momentum 150	its close is above the close 150 candles ago
Momentum 200	its close is above the close 200 candles ago
Momentum 320	its close is above the close 320 candles ago, about 53 days
Golden Cross 50/200	its 50-candle average is above its 200-candle average
Slow Cross 100/200	its 100-candle average is above its 200-candle average

Volatility Harvest

Two older recipes on SUI that buy a red candle below a band and sell a green one above it. They do well in markets that chop sideways and badly in strong trends. The app shows their signals for reference; the engine trades the trend rules above.

How a rule gets listed

A listed rule beats holding the same coins, charged the same fees and funding, on return by 5 points, and its worst fall is no deeper than holding’s. It is not carried by its winners: its median market and its average without the top two markets are both positive. The research desk rechecks this on the newest prices every day. A rule that falls short keeps trading, marked as being upgraded, until a stronger rule passes every gate and takes its place.

Rules are ranked by their median market, and by their average with the top two markets removed, never by the plain average, which one runaway market can dominate.

What the evidence says

The fairest test is the walk-forward: rules picked on prices before 1 Mar 2026, then run once on prices from 1 Dec 2025 that they were never tuned on. On that slice the frozen rule made +41.5%, within 3.3 points of holding all 32 markets, and its worst fall was −13.9% against holding’s −41.8%. Most of the upside, a third of the downside.

• Every one of the 38 rules fell less than holding. Worst falls ranged from −25.1% to −12.6%, against −41.8% for holding.
• For every 1% it fell at worst, the frozen rule made 3.0%. Holding made 1.1%.
• In the months the rules were chosen on, holding lost 11.1% with a worst fall of −65.4%, and the frozen rule made +56.8% with a worst fall of −23.3%.

That is what trend rules are built for: be in a market while its trend runs, and be in cash before a collapse takes most of it. The research desk works every day to push returns higher, and its target now is rules that beat holding outright.

The fine print

• The full-history figures on the research desk page come from the history the rules were chosen on, so they flatter them. The walk-forward above is the fairer test. Its two slices overlap until 1 Mar 2026 so every rule is warmed up before it is scored.
• All figures are before funding, with fees of 5 bps a side and no slippage, on the 32 markets liquid at the freeze on 15 Sep 2026, which favours any test.
• A forward window, fixed in advance, closes on 14 Nov 2026, and its result will be published whether it passes or fails.
• Source: docs/evidence/walk-forward-2026-10-23.json, measured 22 Oct 2026.

Choosing a rule

• If you are unsure, choose Breadth Trend. It is the frozen rule, the one the forward test measures, and the one with the most evidence behind it.
• Slower rules trade less. Fewer trades mean lower fees and fewer false starts. They also enter later and exit later.
• Do not pick by the biggest number. The highest full-history returns belong to rules that also fell further. Compare worst falls, and compare the unseen months.
• Sizing matters as much as the rule. One slot per market is the default and the measured construction. Fewer slots raised both returns and falls in testing, and most 4- and 8-slot books would have hit the 30% halt.

Ask your agent

If you self-host, your assistant can compare rules for you from these docs. Try: “which listed rule had the smallest worst fall out of sample, and how did its return compare with holding?”