I did some testing with floating point calculations to minimize the precision loss. I stumbled across a phenomen I want to show here and hopefully get an explanation.
Wh
Yes, you frequently have to cast them to float get better precision. My take on it:
For better precision cast decimals before calculations