I've tried using the CVdata "betting errors per hour" it is too different.

My assessment of errors is based on shoe count downs during training. I have a error rate of say +1/-1 of the running count 10% of the time. How can I simulate this to know how this would affect winrate?