Backtests are usually limited by their shortest history. A fund launched in 2021, or a model whose first version is recent, can cap the whole comparison. Two stand-in mechanisms extend it, and both say so on screen:
- Reference symbol (set per fund, in a slice's Details… drawer on Target Models): the symbol the slice is modeled on, often an advisor-only or institutional fund you hold a public equivalent of. Its price history also stands in, and it does so all or nothing: a fund with no stored price history is backtested entirely on its reference symbol; a fund with any stored history is backtested on its own, and the reference is never consulted. There is no splice at the inception date. A young fund with a short history still caps the window, and Extend back below is the mechanism that reaches past it. A ladder slice has no price series of its own, so it may name a backtest proxy (a short-Treasury ETF, say); with none named the slice is carried flat at $1 with zero return, neither helping nor hurting the curve.
- Extend back (on Model Backtest): before a model's first version, its earliest weights are held constant and extrapolated backward as far as you allow, or (at Max) as far as every holding has price data. This is on by default at Max, because a model's first version dates from when you recorded the allocation, not from when you started holding it. Honouring that date by default would hand a model you built today a one-day chart. The extended stretch draws dashed and ends at a marker where recorded history begins; set Extend back to Off to limit the chart to that history.
What a stand-in can't tell you
An extended curve answers "how would this allocation have behaved?" It is not the track record of the actual funds (the real fund didn't exist) or of your actual decisions (you hadn't made them yet). Fee and tracking differences between a fund and its reference symbol are inherited from whichever series is in force. Treat extended history as context for the allocation, and weight recent, real history more heavily.
Live rebalancing never uses any of this. Stand-ins exist only inside the backtest.