Backtesting software for portfolio validation
Backtesting software replays a portfolio's rules over past market data to see how they would have behaved. Acubic runs walk-forward backtests: at every past rebalance date it rebuilds the portfolio using only the data available then, charges transaction costs on every trade, and reports the result as hypothetical. It exists to test whether a portfolio process survives outside the sample that inspired it, not to manufacture attractive charts.
Why is walk-forward testing a core requirement?
Walk-forward testing repeatedly evaluates a strategy on unseen future windows. That gives investors better evidence than a single train-test split because it shows whether results persist across multiple market regimes.
How does an Acubic backtest run?
- Pick a strategy, how many years to test and how often to rebalance, and keep or change the transaction cost (15 basis points per trade by default).
- At each rebalance date, the engine screens, ranks and allocates using only data available up to that date, on a rolling lookback window.
- It buys the resulting weights and lets each position drift with its own returns until the next rebalance.
- Every buy and every sell pays the transaction cost on the value traded.
- The report compares the run with the S&P 500 and flags any known data limits it hit.
Key terms
- Look-ahead bias: using information in a past decision that was not yet available on that date.
- Survivorship bias: testing only on assets that still exist today, which leaves out the failures.
- Overfitting: tuning rules until they fit past noise, so they look better in the test than they will live.
- Basis point: one hundredth of a percent; 15 basis points on a $1,000 trade is $1.50.
What makes a backtest realistic instead of optimistic?
Realistic backtests include transaction costs, slippage, execution lag, and liquidity constraints. Without those frictions, the gap between paper performance and live performance becomes much wider than most investors expect. Acubic models some of them and states the rest.
| Modelled | Not modelled |
|---|---|
| A flat transaction cost on every buy and sell | Slippage, market impact and bid-ask spreads |
| Positions drifting with their own returns between rebalances | Partial fills and intraday timing: trades fill at historical closing prices |
| Point-in-time ranking inputs, such as financial statements from the date they became public | Taxes, currency conversion fees and each broker's own fees |
| A liquidity floor on every ranking method, for strategies created or regenerated since October 5, 2026 | Companies delisted before today, which are missing from the universe |
Why is bias control so important in backtesting software?
A backtest loses credibility when it leaks future information or ignores dead securities. Acubic's engine uses point-in-time inputs for its decisions and keeps model selection apart from evaluation by rebuilding the portfolio at every rebalance date. It does not remove survivorship bias: its universe is built from instruments listed today, and the methodology says so.
How should investors compare portfolio strategies?
Good comparison is not only about average return. Investors should also review drawdown shape, left-tail outcomes, turnover, cost drag, and how sensitive the strategy is to small parameter changes. Testing many variants against the same history is itself a risk. Reviewing hundreds of published return factors, Harvey, Liu and Zhu concluded that “most claimed research findings in financial economics are likely false” (NBER Working Paper 20592, 2014). A result that survives only one parameter setting deserves suspicion.
What role does backtesting play in final approval?
Backtesting is one decision layer. A strong portfolio process still needs economic rationale, methodology review, implementation planning, and monitoring thresholds before a strategy goes live.
Who should use portfolio-level backtesting workflows?
Portfolio-level backtesting is useful for investors and analysts who want to validate allocation decisions, rebalance rules, and risk assumptions before they move from research into execution.
What are the limits of a backtest?
A backtest is a hypothetical result: no money was traded, and it shows how a set of rules would have behaved, not what they will earn. Correlations and regimes change, and live orders face spreads and timing a daily price series cannot represent. Past performance does not guarantee future results, and nothing on this page is financial advice.