Strategy families

Level 6. Assumes risk and sizing.

There are not hundreds of strategies. There are a handful of claims about how prices behave, and thousands of implementations of them. Learn the claims and the implementations stop being mysterious.

Six families have a lesson here, because six are what Arvo ships a rule for. If a family is not listed, Arvo cannot test it, and by the rule of this section it does not get a page.

The six

Family	The claim, in one line	Arvo's rule	Needs
Trend following	what has gone up keeps going up long enough to pay for the false starts	momentum_breakout	a trend
Volatility breakout	an unusually large move is the start of something, not the end	volatility_breakout	expansion
Mean reversion	price stretched away from a fair value comes back to it	vwap_reversion	a range
Relative strength	the strongest few keep outrunning the rest	cross_sectional_momentum	dispersion, and many names
Opening range	where a session goes in its first minutes sets where it goes after	opening_range	a session
Selling options	you get paid to take on risk other people want to shed	put_spread	recorded chains

sma_cross is in Arvo as the control — a rule nobody disputes, not a good idea. Use it the way you would use a placebo: if your clever rule cannot beat it, you have learned something.

The two that matter most are opposites

Trend following and mean reversion are each other's failure mode. Given the same instrument and the same year, they disagree on every trade:

flowchart LR
    P["A new high"] --> T["Trend follower:<br/>buy, it continues"]
    P --> M["Mean reverter:<br/>sell, it is stretched"]
    T --> R{"Which was right?"}
    M --> R
    R --> A["A fact about the regime,<br/>not about the rules"]

    style A fill:#1f6feb,stroke:#1f6feb,color:#fff

Both claims are true, in different conditions, and neither tells you which condition you are in — see market types. This is why "which strategy is best?" has no answer, and why the useful question is "what does this one need, and can I tell when it has it?"

How to use these six lessons

Read them in any order; they do not depend on each other. Each one ends in a ruleset you can run, and each one tells you the verdict to expect — which on a single instrument over a couple of years is usually Inconclusive, for reasons level 8 covers.

Run the control first

Before testing any family on an instrument, run sma_cross on it. You learn what that instrument's history does to a rule that has no edge, and every result afterwards has something honest to sit against.

What is not here

Named so you know they were considered rather than forgotten:

• Chart patterns read by eye — head and shoulders, flags, wedges. Not testable without turning "I see a flag" into a condition, at which point it is a rule and belongs in rules as data.
• Fundamental and sentiment strategies — Arvo keeps bars, dividends, quotes and chains. A rule needing earnings transcripts or positioning data has no data layer here.
• Pairs and statistical arbitrage — needs two instruments traded against each other as one position, which nothing in Arvo currently books.
• Scalping — the edge lives below the bar, in the order book, and Arvo studies bars.

Next

Start with trend following, the best-documented edge in public markets and the one whose failure mode teaches the most.