Systematic trading makes trading and portfolio decisions through specified, repeatable rules. Given the same available information, holdings and rule version, the process should produce the same decision. This makes it possible to explain why a trade occurred, reproduce it in a backtest, and check whether live trading followed the intended policy.
The rules can be simple. A calendar-based allocation policy is systematic even if a person calculates and enters its orders. Humans still choose the objective, instruments, limits and rules, monitor their operation and decide when to revise or stop them. Full automation is one implementation choice; Robert Carver’s introduction to systematic trading explicitly includes partly systematising investing.
“Keep 60/40” is not yet a trading rule
A target allocation leaves important decisions open. Does “keep” mean continuously, at month-end, or only after a large deviation? What price determines the order? What happens if a required input is missing?
Here is a deliberately small, fictional policy with those choices filled in:
| Policy element | Specified instruction |
|---|---|
| Instruments and target | Hold 60% of current wealth in asset A and 40% in cash; no short selling or borrowing. |
| Decision clock | Rebalance on designated days 0 and 2. Observe a fresh quote and current holdings at 09:00 UTC; submit the calculated trade at 09:01 UTC. Make no allocation trades between those dates. |
| Position calculation | Buy or sell enough A to restore its 60% target, using the observed quote; fund purchases from cash and add sale proceeds to cash. |
| Missing input | If the required quote or holdings record is unavailable, or the quote is not timestamped 09:00 UTC, skip that date, record the problem and alert the operator. Do not substitute a future observation. |
| Example execution | Assume the entire order fills at the unchanged observed price, fractional units are allowed, and there are no transaction costs, cash yield, distributions or external cash flows. |
The last row is a synthetic calculation convention, not a guarantee of an available market price. In a live implementation, order types, price changes, rejected orders, partial fills and cash reconciliation also need explicit treatment. The SEC’s investor guide describes calendar and allocation-deviation approaches to rebalancing; our policy chooses a calendar.
The rule sometimes says “do nothing”
Start with USD100 in cash. On day 0, A’s observed and execution price is USD10 per unit. The policy buys 6 units for USD60, leaving USD40 cash.
On day 1, A’s observed price rises to USD12, a 20% price increase. The 6 units are now worth USD72. With cash still at USD40, wealth is USD112 and A’s portfolio weight is
No trade occurs on day 1. The allocation has drifted, but the calendar condition is false. A systematic process follows its specified decision rule; it need not hold its target weights at every instant.
Suppose the same USD12 price is observed on day 2 and remains unchanged for the assumed execution. The target amount in A is 60% of USD112, or USD67.20. The policy therefore sells USD4.80 of A, equivalent to 0.4 units:
| State | A units | A value | Cash | Total wealth |
|---|---|---|---|---|
| Day 0, after purchase | 6 | USD60.00 | USD40.00 | USD100.00 |
| Day 1, no scheduled trade | 6 | USD72.00 | USD40.00 | USD112.00 |
| Day 2, before scheduled trade | 6 | USD72.00 | USD40.00 | USD112.00 |
| Day 2, after scheduled trade | 5.6 | USD67.20 | USD44.80 | USD112.00 |
The final weights are 60% and 40%. The sale moves value from A to cash; under these assumptions it does not change total wealth at that instant. Resetting A to its original USD60 would be a different instruction, because the target applies to current wealth.
Original synthetic policy ledger. All times are UTC; days 0 and 2 are the designated rebalance dates. Execution assumes unchanged quotes, fractional units and no costs, cash yield or external flows. Blue bars show A’s share of current wealth.
For the derivation behind the changing denominator, Turnover and the Information Horizon develops passive weight drift and the trading needed to change weights. The public page provides a preview; the full lesson requires membership.
Repeatability is a testable property, not a profit forecast
More elaborate systems replace the fixed allocation with a trading signal, then translate that signal into positions subject to risk and trading constraints. A signal alone is not the whole system: “buy when the score is positive” still leaves position size, exit timing and execution unspecified.
A useful replay asks: could this decision have been made with the information available at that moment? Using a later price or revised data that arrived after the decision introduces look-ahead bias. Using the correct information can reproduce a decision without making its forecast correct. Transaction costs, model errors and changing market conditions can still make a consistent strategy lose money.
Monitoring remains part of the process. In its guidance for U.S. member firms using algorithmic strategies, FINRA’s Regulatory Notice 15-09 discusses testing, change control, alerts and post-implementation review. A human intervention can be recorded as a rule-defined response or as an override. Silently changing a rule after a loss destroys the distinction between following a policy and evaluating a new one.
Sources
- Robert Carver — Systematic Trading: the author’s account of full or partial systematisation, trading rules, position sizing and the absence of guaranteed success.
- SEC Investor.gov — Beginners’ Guide to Asset Allocation, Diversification, and Rebalancing, “Rebalancing 101” and “When to Consider Rebalancing”: restoring allocations and choosing calendar or deviation-based rules.
- FINRA — Regulatory Notice 15-09, sections II–IV: development, testing and ongoing supervision of algorithmic strategies; guidance addressed to FINRA member firms.
- Edward E. Qian, Ronald H. Hua and Eric H. Sorensen, Quantitative Equity Portfolio Management: Modern Techniques and Applications, Chapman & Hall/CRC, ©2007, §§8.1–8.2, pp.234–236: passive portfolio drift and periodic restoration of fixed weights. The policy, ledger and diagram here are original.