Subject: Re: twice the Nasda100 CAGR with simple criteria
Ray: I think we need somebody who has the ability to see which stocks are picked for each screener on a couple of identical dates.

I agree but only P123 gives you a log of the selected stocks buy and sell price for each date. It's not available on GTR1.

Aussi: The GTR1 screen rebalances every 20 days, Portfolio every 4 weeks. . GTR1’s 20 days is 20 trading days, the same as P123’s 4 weeks, so that wouldn’t be the difference if they both compensated for holidays the same way. However P123 always trades on the 1st day of a new week, I’m not positive how GTR1 handles it but I believe they always use exactly 20 trading days in this case. So that seems to be a difference.

I just looked at my pervious backtest posted above for P123 and found the selection of the top 10 used an incorrect selection equation. Rerunning the backtests.

Rerunning using a ranking top 10 100*((Close(0)-Close(251)) / Close(251)), this is price only.
P123 CAGR for 1st buy with rank date 1/04/20 and the buy on 1/06/20 CAGR 23.56, Nasdaq 100 19.09%

Rerunning using a ranking top 10 Ret%Chg(251) which includes dividends
P123 CAGR for 1st buy with rank date 1/04/20 and the buy on 1/06/20 CAGR 23.56 --No Change from price only--

P123 all 100 Nas100 equal weight, CAGR 13.75%

GTR1 CAGR 19.62% for top trp (total return) for start trade date 20200106
https://gtr1.net/2013/?~Naz100...

GTR1 all 100 Nas100 stks equal weight 13.73%

At this point it seems likely that both GTR1 and P123 results are slightly different most likely only because they make their rank and trade decisions on different dates throughout the backtest. As noted previously GTR1’s 20 different start dates have variations between 13.7 to 23.2.

Looking at this reinforces what I’ve come to realize about evaluating a backtest.
1. GTR1’s rolling start date with an average of many starts is much better than P123’s single weekly start date.
2. A 5 or 10 hold portfolio leaves way too much to chance, Use a minimum of 25 for screen evaluation.
3. Backtest over at least 13 years, some research shows around 10 years tends to hold up post discovery.
4. Try several start dates 3+ months apart for any backtest of less than 10 years with different buy and sell criteria. Fundamentals change quarterly and initial by chance selections can distort the results.