A long backtest is only useful if the underlying series mean the same thing throughout it. The KISS history rebuild therefore changed the data methodology and documented the boundaries before extending the performance record.
The production reconstruction begins on the first trading day of 2020. A separate 2018–2025 diagnostic provides an earlier comparison. Neither should be mistaken for signals published contemporaneously across its full history.
Joining GLD and GLDM without inventing a return
GLDM did not trade throughout the desired indicator warmup. The reconstruction uses GLD before the June 26, 2018 overlap, then switches to GLDM. At the overlap, the GLDM price scale is multiplied by approximately 4.73705 to match the earlier series.
That scaling prevents a mechanical price discontinuity at the handoff. It does not remove the two funds’ different expenses, tracking or trading histories. The transition return remains an observed market return rather than an artificial jump between share-price levels.
A portfolio study that actually holds GLDM is a different question. The five-fund comparison starts at GLDM’s first trading close and does not use GLD to manufacture an earlier investable record.
Replacing an unavailable credit history
The rebuild replaced the previous ICE high-yield spread input with the difference between 30-day A2/P2 and AA nonfinancial commercial-paper rates. The source series are FRED’s A2/P2 rate and AA rate.
This is a short-term funding-quality premium, not a high-yield corporate-bond option-adjusted spread. The project reconstructed the full history using the replacement rather than switching definitions partway through the sample. That methodology change altered some historical regimes and allocations.
Source coverage and observation dates remain part of the model’s provenance. More history does not solve release delays or later revisions: current-vintage macro reconstruction still contains information that may differ from what an investor originally saw.
Participation is not constituent breadth
Equal-weight versus capitalization-weighted ETFs, sector ETFs and other relative-price baskets can describe participation or risk appetite. They do not count how many historical index constituents were above a moving average.
The participation source evaluation required point-in-time membership, delisted-stock prices, adequate history, source dates and usable data rights before promoting true breadth. A price feed alone did not satisfy those requirements.
Missing optional evidence must remain visible. It should not silently become a neutral or reassuring signal. The KISS 2.4 corrections applied that principle to sector coverage and macro domains.
When comparing two backtests, check the series definition, earliest usable observation, release convention and model version before interpreting a difference in returns. Those details can move an allocation even when the trading rule’s name stays the same.