I am running tests to try to understand how Zorro works. I found something weird when doing backtest using Z12 system.

The Zorro version is 2.30 (H 5 B 0 V 3). The history data I am using are the available t6 files from 2010-01-01 to 2020-02-29.

1. First I run the backtest and get the equity curve file "Z12_pnl.csv".
2. Then I modify the "Z12_17.par" file (change the US30:VO line from "US30:VO 2.17 36.6 1.817 228 0.198 17.07=> 1.313" to "US30:VO 3.03 63.4 2.01 44.2 0.401 1.910=> 1.905"), and run the backtest again and get another equity curve file "Z12_pnl_2.csv".
3. Now I find that "Z12_pnl_2.csv" starts to be different from "Z12_pnl.csv" in the very early stage (since 2013-10-10).

Maybe other Z systems have the same issue, but I haven't tested yet.

My question is, how come the system parameters from the later stage can affect the equity curve of the earlier stage?

Please find attached the 2 equity curve files, as well as the .par files used for the 2 tests.