Skip to main content

Building a Swing Trading System

Turning tactics into a repeatable process.

Updated Jun 23, 2026 at 8:47pm

  • 18374b0a563a Defining Your Edge (Rules) 1 1,342
  • 1838397f6259 Backtesting Swing Setups 1 1,162
  • 1839c2055d38 Forward / Paper Testing 1 1,253
  • 1836f9416c74 Metrics (Win Rate, Avg R, Expectancy, Drawdown, Profit Factor) 1 1,225
  • 18353b5cca34 The Trade Journal 1 1,255
  • 184017ad556a Periodic Review & Iteration 1 1,341
Tree Key
Expandable — has sub-topics
475Local Id for node
a1b2c3d4Click to see full UUID
6Sub-topics
7Documents
8.7k wordsResearch depth
5Open node
Research Draft Medium 1,133 words

Most traders collect setups; few build systems. A system is not a single setup — it is the full apparatus that turns a tactic into a repeatable, measurable process: objective rules, an honest validation procedure, a small set of metrics that describe the rules' behavior, and a journal that keeps the live record honest against expectations. Van Tharp's central insight is that you do not trade the market, you trade your beliefs about the market, encoded as a written system and executed with discipline. This document is the meta-process that sits above every individual swing setup in this branch.

Defining the edge (rules)

An edge is a positive expectancy — over many trades the average outcome is greater than zero after costs. To know whether you have one, the rules must be objective and testable: another person reading them should take the same trades you would. That means a written specification of the entry (the exact conditions that trigger a position), the exit (both the protective stop that defines your initial risk and the profit target or trailing logic), and the position-sizing rule (how much of the account is risked per trade). Tharp frames the unit of risk as R — the distance from entry to initial stop — so every trade's result can be expressed as an R-multiple (a +2R win risked 1R to make 2R). Pairing R with a fixed-fractional sizing rule (e.g. risking a fixed small percentage of equity per trade) keeps risk proportional as the account grows or shrinks, and lets you separate the question "is the setup any good?" from "how much do I bet on it?" Vague discretion ("buy when it looks strong") cannot be tested, journaled, or improved — it is the enemy of a system.

Testing

A backtest's job is to estimate whether the rules have edge before real money is at stake — but most backtests lie, and knowing how they lie is the skill. Overfitting / curve-fitting is the core failure: with enough tunable parameters you can fit historical noise rather than durable behavior, producing a beautiful equity curve that dissolves live. Look-ahead bias uses information not actually available at decision time — the classic case is filling at the open using the same bar's close. Data-snooping is the slow-motion version of overfitting: repeatedly testing and tweaking on the same dataset until something "works," which guarantees a fit to that specific history. The defenses are procedural, not clever: reserve out-of-sample data the rules never touched during design, and prefer walk-forward analysis, which optimizes on a rolling window and tests on the next unseen window repeatedly, so the strategy must re-prove itself across regimes rather than once. Finally, forward / paper testing in real time catches what historical data cannot — slippage, fills, and your own behavior — before capital is committed.

The metrics

A handful of metrics describe a system's R-distribution; none is sufficient alone, and each trades off against the others.

  • Win rate — the share of trades that profit. High alone tells you nothing: it can hide a tiny-win/huge-loss profile. A low win rate is normal and tradeable if winners are large (the trend-following shape).
  • Average R / win-loss ratio — average reward relative to the 1R risked. This is the lever a low win rate leans on.
  • Expectancy — the mean R-multiple per trade: roughly (win% × avg win in R) − (loss% × avg loss in R). This is the single number that answers "does this system make money over many trades?" Positive expectancy is the definition of an edge.
  • Profit factor — gross profit divided by gross loss. Below 1 loses money; modestly above 1 is marginal. Suspiciously high values can signal overfitting rather than brilliance.
  • Maximum drawdown — the largest peak-to-trough equity decline. This is the survivability metric: it decides whether you can actually hold the system through its bad stretch. A technically profitable system with a drawdown deeper than your tolerance is untradeable, because you will abandon it at the worst moment.

The tradeoff to internalize: expectancy and profit factor tell you if there is edge; win rate and avg-R tell you how the edge is shaped; drawdown tells you whether you can stand it. A smoother, lower-expectancy system often beats a higher-expectancy one with violent swings, because the first one you will still be trading next year.

The journal & iteration

The backtest produces expected metrics; the journal produces actual ones, and comparing them is the only way to know if the system is working live or quietly degrading. Keep the journal contemporaneous — entry rationale, the exact rule that triggered, execution, exit, and emotional state, logged at the time, not reconstructed afterward. Crucially, score process separately from outcome: did you take every valid signal and reject every invalid one? Rule-adherence matters more than any single trade's P&L, because a good process can lose and a broken process can get lucky. Review periodically against the backtest baseline: if the setup metrics hold but your executed metrics are worse, the problem is behavioral and the fix is a protocol (checklists, daily loss limits, a losing-streak plan), not a new indicator.

Strengths & limitations

A system's real value is twofold: it manages discipline (it tells you what to do so emotion does not), and it makes performance measurable (you can improve what you can count). Its honest limitation is backtest-to-live degradation — live results routinely come in worse than historical simulation, and live drawdowns tend to run deeper than the backtest suggested, because of slippage, costs, regime change, and the difficulty of executing rules under pressure. Treat the backtest as an optimistic upper bound, not a promise. A system does not predict the market; it disciplines you and tells you, with evidence, when your edge has stopped working.

System relevance

This is the meta-process the rest of the swing branch feeds into. Individual setups (trend-continuation, breakouts, reversals) are candidate rules; this document is how those rules are validated, sized, measured, and iterated. In this platform, Cairn's post-mortems are the live journal layer — the contemporaneous, process-vs-outcome record that flows back into the metrics here and drives the periodic review that decides whether a rule keeps its place in the system.

Sources