I ran the historical-friction test on one of my 1 stock strategies, using its preserved audited trade history and changing only the transaction-cost assumption.

I had AI use the schedule we agreed on after examining evidence on historical spreads (until this I had used a flat .1%): 0.75% full round trip for 1993–96, 0.25% for 1997–2000, 0.05% for 2001–10, and 0.02% thereafter. This is intentionally conservative in the early period. The historical breakpoints are well supported: Nasdaq's 1997 reforms substantially narrowed spreads, and the SEC reported that Nasdaq quotation spreads fell about 50% with decimalization in 2001.

The result was reassuring:

strategy using Friday closing based on Thurs. closes' pricess SumRank 5 stocks equal — SPY 325 timing SumRank 5 Yes Existing 36.30% 1.16 2.45 30.86% 0.95
previously looked at constant 0.10% friction then tested on what the evidence shows was the historical friction Change
59.99% CAGR v. 59.42% CAGR i.e. only lost −0.57 pct.

So the answer to the question we were testing is quite clear: the extraordinary cagr result is not an artifact of our having assumed today's relatively low trading friction back in the 1990s.

The reason the damage is only about half a CAGR point is that the punitive costs apply only to the early portion of a roughly 32-year test. Once the Nasdaq reforms occur—and especially after decimalization—the historically reasonable friction becomes lower than our old constant 0.10% assumption. The SEC evidence supports precisely that structural change.

There is also a useful methodological point. I did not estimate this by simply subtracting some number from the reported 53.8% CAGR. I used the preserved trade history data, which records every change of position, the number of transaction sides and equity after the trade.

I.e. a change in spreads does not defeat the evidence.

I also tested my highest 5-stock strategy and the result was a loss of no more than .5% from a 36% plus cagr.