Building a systematic strategy

A systematic strategy is a set of rules precise enough that a computer — or a sufficiently obedient human — can execute it without judgment calls: what to trade, when to enter, how much to size, when to exit, with no clause that begins “unless it feels like.” The case for trading this way is not that rules are smarter than people. It is that rules are testable, consistent, and honest: they can be evaluated on history, they do not revenge-trade after a loss, and when they stop working, the failure is visible in numbers rather than buried in selective memory.

What follows is the full lifecycle — hypothesis, rules, capacity, monitoring, and the stage most traders refuse to plan for: retirement. Strategies are not heirlooms. They are products with shelf lives.

Start with a hypothesis, not a pattern

The defining question of strategy design is why does this edge exist — and who is paying for it? Markets are competitive; persistent profit requires a counterparty who systematically accepts the other side, and a reason they keep doing so. Sturdy answers tend to fall into three families: someone is paid for bearing a risk others shed (and occasionally gets hurt doing it), someone exploits a behavioral regularity (herding, anchoring, panic) that persists because humans persist, or someone is compensated for providing a service like liquidity or capital at inconvenient moments. A backtest discovery with no such story — “this pattern worked for six years” — is a correlation in search of a coincidence. The hypothesis also earns its keep later: when performance degrades, a strategy with a stated mechanism lets you check whether the mechanism still operates, while a pattern-mined strategy offers no way to distinguish a rough patch from death.

The classic edge families

Three sources of return recur across decades and asset classes — worth knowing as categories, not recipes:

• Momentum: assets that have performed well or poorly over recent months tend, on average, to continue for a while. The standard stories are behavioral — investors underreact to news, then herd into the move. Momentum’s signature failure is the violent reversal at turning points, where the strategy is by construction positioned with the old trend.
• Mean reversion: prices stretched far from a reference tend to snap back, particularly at short horizons where liquidity provision dominates. It typically wins small and often, then loses large and rarely — the stretched price sometimes keeps stretching, because this time the move was information, not noise.
• Carry: holding the higher-yielding asset against the lower-yielding one earns the differential while you wait — FX rate differentials are the classic case. Carry collects steadily in calm regimes and gives much of it back abruptly in stressed ones; the yield is partly payment for being short exactly that abruptness.
• Notice the shared fine print: each edge is compensation for a discomfort — momentum’s reversals, reversion’s tail losses, carry’s crash risk. Edges that pay without discomfort attract capital until they stop paying.

From hypothesis to rules

Translating an idea into a tradable system means making every vague word operational. “Buy strength” becomes: universe (which of the instruments you can actually trade — on Obsidiate, a definable subset of 99 across four asset classes), signal (return over a defined lookback, ranked how, refreshed when), entry and exit (rebalance schedule, exit conditions), sizing (fixed fraction, volatility-scaled, position caps), and cost assumptions (your real fee tier, realistic slippage, maker or taker execution). Two disciplines keep the translation honest. First, prefer the simplest rule that expresses the hypothesis — every additional filter is another knob that can overfit, and momentum does not require eleven parameters. Second, write the rules down before testing at scale, including what you will not vary; the document is your defense against the quiet slide from testing a hypothesis to mining for whatever works.

Capacity: every edge has a weight limit

Capacity is the amount of capital a strategy can run before its own trading destroys the returns. The constraint is market impact: your orders consume liquidity and move prices against you, and the cost scales with your size relative to what the market absorbs. A signal that earns 0.3% per trade at small size can earn nothing at twenty times that size, because the act of entering moves the price most of that 0.3%. Capacity scales with the liquidity of the instruments and the holding period of the strategy — fast strategies in thin instruments have tiny capacity; slow strategies in deep markets have enormous capacity. This is also why genuinely high-return strategies are rarely sold or shared at any price: they tend to be small-capacity, and every dollar of someone else’s money running the signal crowds out the originator’s own returns. An edge that fits in a small account is a real and respectable thing; pretending it scales is how small wins become large losses.

Monitoring decay

Most strategies decay. Markets adapt, the inefficiency gets crowded, the regime that fed the edge ends. The operational problem is brutal: distinguishing normal variance from death, using only a noisy returns stream, while emotionally invested in the answer. Improve your odds by deciding in advance what failure looks like:

1. Define expected behavior from the backtest before going live — typical drawdown depth and duration, hit rate, the distribution of bad months. Decay is judged against this baseline, not against hope.
2. Set tripwires, not vibes: a drawdown materially beyond the worst in-sample, a hit rate degraded for a defined span, live slippage persistently exceeding modeled slippage. Crossing one triggers review; crossing several triggers de-risking. Decided now, while you are calm.
3. Monitor the mechanism, not just the P&L. If the hypothesis was a behavioral regularity, is the regularity still measurable in the data? Mechanism intact plus poor returns can be variance. Mechanism gone is the answer, regardless of last month’s gain.
4. Watch execution quality separately — deteriorating fills and rising impact can mean the trade is crowded, which is its own form of decay and often precedes the visible kind.

Retirement: the exit nobody designs

The hardest discipline in systematic trading is killing your own strategy. Every incentive argues for one more month: sunk development time, the memory of good years, the genuine statistical possibility that this is merely a drawdown. The cleanest mitigation is deciding the retirement criteria at launch — written next to the entry rules, by the version of you who had no position to defend. Retire, or at least cut to token size, when the tripwires are crossed and the mechanism check fails; scale back in only if the mechanism demonstrably returns. And expect, as the base rate, that any honest edge has a finite life — the practitioners who last decades are not the ones who found an eternal strategy, but the ones who ran a pipeline: researching the next edge while the current one pays, and retiring each without sentiment when its time came. The strategy is not the asset. The process that produces strategies is.

Key takeaways

• A strategy starts with a hypothesis about why the edge exists and who pays for it — patterns without mechanisms are coincidences with parameters.
• Momentum, mean reversion and carry are the classic edge families, and each pays as compensation for a specific, recurring discomfort.
• Translate the hypothesis into the simplest rules that express it, written down — universe, signal, sizing, costs — before testing at scale.
• Capacity is real: your own market impact caps every edge, and fast strategies in thin instruments cap out quickly.
• Define decay tripwires and retirement criteria at launch, while you have nothing to defend; monitor the mechanism, not just the P&L.
• Strategies are products with shelf lives. The durable asset is your research process, not any single set of rules.