Rather than chemically (i.e. relying too much on a glaze calculator)
Some beginner glaze mixers over rely on glaze calculator
Use calculated thermal expansion as a crutch to fit glazes to clay bodies
Give up on the first try if an online glaze doesn't meet expectations
There are things a calculator can't tell you
Materials that deviate from ideal (whiting is typically "contaminated" with dolomite)
Poorly melting glazes are highly dependent on physical properties of glaze
Because hobbyist ceramics are highly variable and volatile
Ingredients can get contaminated
Some people measure quite fast and loose
Kilns can fire unevenly - cold spots that are 1 cone colder are not uncommon, kilns may over fire, natural cooling is highly dependent on kiln packing, etc
Test empirically
Can produce extremely simple glazes
My oribe glaze is conceptually a blend of only 3 ingredients (feldspar, wood ash, silica) in simple ratios, fits a stoneware, and is pretty
Very useful if going for aesthetics and not functionality
Limits and food safety need to be checked empirically anyways
No guarantee of food safety unless you send a pot off for lab testing
Knowing how to adjust a glaze is far more useful than just going blindly off of a printout
Scientific method establishes a cause-effect between ingrendiets and glaze qualities
How
Change the percentages of different ingredients
Perform line blends to see if an ingredient / ratio needs to go up or down