Special data series

Also seen as: reference series, macro series, special symbols

Which special data series does Fincanva keep?

Two of the six are worth a note on how they are produced. The Shiller PE is taken exactly as its public reference series publishes it rather than recomputed, so Fincanva's value matches the widely quoted figure instead of a private recalculation. The Buffett Indicator is computed inside Fincanva as the standard ratio above, from market-capitalisation and GDP data. The Buffett Deviation's construction is proprietary and is not published — only the existence of the series is documented here.

What happens when a date falls outside a series?

A requested date outside a series' coverage falls back to a fixed assumed value rather than failing the run, but the condition that triggers the fallback is not the same for all three rate and inflation series. The fallback value itself is not published.

The difference is observable. A date past the end of the data produces the last real inflation reading but a fallback risk-free rate; a date before the beginning of the data puts both series on their fallback. In an ordinary run you will rarely meet either, because the simulation start year defaults to 2000 and these series reach back decades further — it is an edge behavior, not a routine one.

Defaults in Fincanva

• Six special data series exist today: risk-free rate, margin-loan rate, inflation, Shiller PE, Buffett Indicator, and Buffett Deviation.
• The risk-free rate is a live market series matched to your backtest's own dates, not a constant — see risk-free rate.
• The same fallback value is used in every case; only the condition that triggers it differs by series.
• Special data series are references rather than holdings: they are not tradable instruments a strategy buys, so they never appear as a position in a backtest.
• They refresh alongside instrument prices, so a series' last data point moves forward as new data lands — see data freshness and frontier.

Worked example

Suppose a backtest window includes a month for which the risk-free series has no observation. For that month the simulation substitutes the fallback assumption, so the month's excess return is measured against a fixed value rather than against whatever short-term rates actually were. Where the two differ, the month's excess return is distorted by the gap between them — understated when real rates were below the fallback, overstated when they were above it — and any Sharpe ratio computed over a window containing that month inherits the distortion. The longer the window, the smaller the share of it one substituted month can move.

The same month's inflation figure behaves differently: it falls back only if the month predates the start of the inflation series. If it merely sits past the end of the data, the last real reading is reused instead — so one series can be on its fallback while the other is not.