Back to index

Automatic proxy

How the fund used to extend an ETF's history is chosen.

When an ETF in a portfolio starts after the beginning of the analysis period, Backtes.to can extend its history further back using the real prices of another, very similar fund: the automatic proxy.

How it is chosen

For every ISIN in the catalogue we keep a list of candidates, each with the historical correlation of daily returns against the fund and how many extra days of history they add. The correlation is the Pearson correlation of daily returns over the period where both funds are listed: it measures how closely the two prices move day to day, not how similar they are in level or volatility.

The applied proxy is the first candidate, provided its correlation is above 0.90 (squared, a 0.90 correlation is an R² of 0.81: roughly 81% of the target fund's daily variance would be "explained" by the proxy, a threshold conventionally described as "very strong"). If no candidate clears that bar, the ETF is not extended and the form keeps its real listing range.

The correlation and extra years shown in the prompt that appears before you run a backtest with proxies available are exactly the ones used to make this choice.

Hand-curated composites

For a small number of synthetic or leveraged ETFs, where no single fund is correlated for a long enough stretch, history is extended with a hand-built composite series instead of the automatic correlation-based pick.

Turning the extension on or off

  • In the portfolio form, the "Extend period with automatic proxy" toggle applies the best proxy for each ETF over the chosen period.
  • From account settings you can make this permanent: every new ETF with a matching proxy is extended automatically, without asking for confirmation on every backtest.

Not the same as a reconstructed history

The automatic proxy uses real prices from another already-listed fund. That is a different mechanism from ETFs with reconstructed history, which instead estimate a synthetic series for the period before a fund existed, when no real proxy is correlated closely enough (marked simfactor: in the backtester).