Interesting screen. The backtest could easily be extended to 1987 and brought forward using P123, but first some steps could be cleaned up some.

Define {JoinS5roeMo_20260316_musselmant}
step0: [S&P 1500 Member] != null
step1: [S&P 500 Member] == null
step2: [MktCap] Top 75
step3: [Average dollar-volume over 63 days] > 2e+07
step4: [SI Return on Equity Y1] > 20
step5: [Actual closing Price] > 15
step6: [Total Return Multiplier over 60 days] > 0
step7: [SI Return on Equity Y1] < 45
step8: [[Total Return Multiplier over 200 days]/[Total Return Multiplier over 5 days]] Top 10
step9: [SI Return on Equity Y1] Top 5

step6 does nothing, as the Total Return Multiplier is always positive.

step3 and step5 have magic numbers that do not age well (market dollar-volume was less in 1987). And the screen does about the same without these steps.

Results from 19970902 to 20251128:

    Screen      CAGR  SAWR  GSD  Sharpe
JoinS5roeMo26a  21.6  10.2  31    0.79
JoinS5roeMo26b  20.7  9.1   30    0.79
JoinS5roeMo26c  12.8  5.1   32    0.51
JoinS5roeMo26d  19.5  8.8   29    0.75
JoinS5roeMo26e  20.4  9.4   30    0.77

JoinS5roeMo26a is the OP screen.
JoinS5roeMo26b drop step3, step5, and step6, and add a few standard steps.
JoinS5roeMo26c uses ROE TTM. This decreases CAGR.
JoinS5roeMo26d uses ROE = (netincy1.s / equityy1.s).
JoinS5roeMo26e uses Equity = average(equityy1.s, equityy2.s).

Why does ROE TTM result in a much lower CAGR?

Using 2 years to calculate equity (version e) is slightly better than just using one year (version d), maybe awarding growth in equity.

gtr1.net: GTR1 Backtester - JoinS5roeMo26e
gtr1.net: GTR1 Backtester - JoinS5roeMo26d
gtr1.net: GTR1 Backtester - JoinS5roeMo26c
gtr1.net: GTR1 Backtester - JoinS5roeMo26b
gtr1.net: GTR1 Backtester - JoinS5roeMo26a_20260316_musselmant