Rule check

Plain weekly buys are the default. One other rule spends more when the price is far under a power-law fair value, and less when it is far over. Unspent cash waits. It does not earn interest.

The full backtester, with its own charts, stays on the home page.

Open the backtester

Off. Plain weekly buys stay the default. The lines below are what the other rule did.

Fixed budget, walk-forward fit

Median +3.6%. Ahead in 92% of 464 windows. Worst window −0.4%.

Unspent cash sits idle. Average cost was about −8.0% against the same dollars every week. The 10th percentile was 0.0%.

Same rule, fixed 5.82 exponent

Median +2.6%. Ahead in 70% of the windows. Worst window −11.6%.

That exponent was fit on the whole record, so it can see closes that had not happened yet. The walk-forward fit cannot. The gap between the two is the point.

The windows overlap. They hold about 3 to 4 independent cycles, not 464 separate histories. A past record, not a forecast.

On the file through the latest bundled close, a research count of this same rule was about +4% median, 91% of windows ahead, and −0.4% at the worst, for the walk-forward fit. The fixed exponent was about +2%, 67% ahead, and −12% at the worst. This page recomputes both. A one-day difference in where the weeks start moves the rounding.

Other ideas were measured and left off this site. This rule is a comparison with plain weekly buys, not a recommendation. The power law here is a fit to past closes, not a forecast.

A coin goal, with the same weekly amount if you brought one