01 · Hypothesis
Turn the idea into a rule you could prove wrong
Start with the decision you want to study. A useful hypothesis says what can be owned, how positions are selected, when the portfolio changes, and what success will be compared with.
Universe
The securities eligible for selection, such as large US companies or a fixed ETF list.
Signal
The observable rule used to rank, include, or exclude each security.
Portfolio rule
The number of positions, their target weights, and the rebalancing schedule.
Benchmark
The investable alternative the strategy must justify outperforming.
02 · Eligibility
Decide what the strategy is allowed to own
Use the screener to narrow the market by characteristics such as size, exchange, sector, valuation, or financial quality. The securities that pass become the investment universe supplied to a backtest.
Universe rule
Who is eligible?
Example: US-listed companies above a chosen market-cap threshold.
Selection rule
Who is chosen?
Example: the ten eligible companies with the strongest six-month return.
03 · Method
Choose the builder that matches the question
Each builder represents a different portfolio rule. Choose one because its assumptions fit your hypothesis—not because its example result has the highest return.
Recommended first test
Strategic allocation
Rebalance a defined list of securities to fixed target weights.
- Tests
- Whether a chosen asset mix and rebalancing rule met your objective.
- You choose
- Securities, target weights, dates, and rebalancing frequency.
- Watch for
- A benchmark that does not reflect the portfolio's stock, bond, or cash exposure.
Cross-sectional momentum
Relative strength
Rank an eligible universe by one or more price-strength signals and hold the leaders.
- Tests
- Whether recent relative winners continued to lead after rebalancing.
- You choose
- Signals, signal weights, portfolio size, and optional trend filters.
- Watch for
- High turnover, concentrated positions, and results driven by one market regime.
Company characteristics
Fundamentals
Rank companies by selected financial or valuation factors at each rebalance.
- Tests
- Whether the selected characteristics identified a stronger portfolio.
- You choose
- Factors, factor weights, portfolio size, and rebalancing frequency.
- Watch for
- Combining many factors until the historical result looks unusually smooth.
Trend filter
Moving average
Hold qualifying securities while their prices remain above a selected moving average.
- Tests
- Whether a trend rule reduced exposure during sustained declines.
- You choose
- Moving-average window, ranking signal, universe, and portfolio size.
- Watch for
- Frequent whipsaws when prices repeatedly cross the trend threshold.
Price levels and risk
Breakout trading
Test entries after price moves through a defined level, with explicit exit and risk rules.
- Tests
- Whether moves beyond recent support or resistance showed useful continuation.
- You choose
- Entry level, exit logic, eligible instruments, and test period.
- Watch for
- False breakouts, optimistic execution assumptions, and gap risk.
04 · Test design
Make the comparison difficult to fool
A credible test uses assumptions that could have been followed at the time. Before running it, review the settings that most often make historical results look better than a real decision would have been.
Use a meaningful period
Include more than one market condition where the available history permits. A rule tested only during its ideal regime has not faced much evidence.
Choose the benchmark before the result
Compare with a simpler investable alternative that carries similar exposure—not whichever index is easiest to beat.
Use a realistic rebalancing rule
More frequent decisions can increase turnover, taxes, slippage, and the opportunity for execution to differ from the simulation.
Account for implementation costs
If the result does not deduct every fee or spread you would face, subtract a reasonable estimate when deciding whether the advantage is material.
Reserve an independent check
Develop the rule on one period, then inspect a later period without changing the rule. Treat that later result as new evidence, not another tuning opportunity.
05 · Interpretation
Read beyond the headline return
Open a completed test from Results and inspect the result as a body of evidence. A strategy can finish ahead of its benchmark and still be too fragile, concentrated, or costly to use.
Return
Did the rule add enough?
Compare cumulative and annualized performance with the benchmark, then judge whether the difference is large enough to survive costs and uncertainty.
Drawdown
What had to be endured?
Maximum drawdown shows the deepest historical decline. Also inspect how long recovery took and whether the loss fits the risk you could actually tolerate.
Risk-adjusted return
Was return efficient?
Use volatility and measures such as Sharpe when shown to compare the return earned with the variability accepted. No single ratio is a verdict.
Turnover
How much action was required?
Review position changes and simulated orders. Frequent trading can convert a small paper advantage into a negative real-world result.
Concentration
What produced the result?
Inspect historical holdings and individual positions. A strategy dependent on a few securities, trades, or years deserves a lower level of confidence.
Stability
Did the rule travel?
Compare different periods and reasonable nearby settings. Evidence is stronger when the conclusion does not disappear after a small, defensible change.
06 · Monitoring
Monitoring continues the test; it does not prove execution
Eligible subscribers can select a saved result for ongoing monitoring. PyInvesting reruns the strategy as new daily end-of-day data becomes available and reports the positions or proposed actions produced by the rule.
What monitoring provides
- A repeatable daily evaluation of the saved strategy
- Current model positions and proposed changes
- A record you can compare with the original hypothesis
What monitoring does not prove
- That an order was submitted to your broker
- That a trade received the simulated price
- That future results will resemble the backtest
07 · Evidence checkpoint
A result earns the next test—not automatic trust
Before using your own capital, make sure you can answer each question without relying on the headline chart.
- Did the strategy outperform an appropriate benchmark by a meaningful amount?
- Does the advantage remain after reasonable fees, spreads, slippage, and taxes?
- Could you tolerate the historical drawdown and the time it took to recover?
- Is the result spread across periods and positions rather than one exceptional episode?
- Did an untouched period support the rule without further tuning?
- Can you explain exactly when the strategy should—and should not—work?
Start with the clearest rule
Build a simple allocation, then earn complexity.
Fixed weights and a scheduled rebalance make the first hypothesis easier to explain, challenge, and compare.