I'm a bit confused how to use WFO Cycles.

In the manual there are some hints to use around 10 Cycles regardless the complete WFO Period in years for Maschine learning.

Some other point out that the WFO Cycles should be customized for the number of years used for the WFO.

What would be the best calculation or number for WFO Cycles used for Maschine learning in Zorro ?

Another question would be the ratio for test and Train. Did a ratio of 20:80 produce a more accurate Result in Live Trading compare to 10:90 ?

Both the number of cycles and the ratio depend on the strategy. The training cycle must not be too long, otherwise it does not properly adapt to the market situation. But it should also not be too short, otherwise you get not enough trades for training. So you have to find out the optimal number of WFO cycles for your system. The ratio of 85:15 is normally ok for most systems.