Practice, play now, dynasty, they all play the same for me. The way I did sliders when this game came out was going into practice first and focusing on certain areas of the game independently. An example being running a set passing concept vs a certain coverage and knowing how it should play out. I'd mess with the sliders and continue running certain play calls over and over until I was happy and would move onto the next area. After getting coverage where I wanted it I'd move onto run blocking as an example and run against certain fronts, heavy boxes, light boxes, all in a controlled environment.
Once I did that for each area of the game for the user and CPU I would then go to play now and play full games 10 games at a time tracking all the stats I felt were relevant to track like yards per play, yards per attempt both rushing and passing, drop rates, fumble rates, explosive play rates, etc.
After getting my samples in and comparing the numbers to real life, I'd make adjustments. After doing a few rounds of samples I eventually settled into a solid set that I would run with in dynasty. If after a season of dynasty I felt any tweaks were necessary I would then make them.
I've never personally thought sliders played differently in practice, play now, or franchise in Madden over the years as that's how I've always built my sliders, so coming into CFB25 I was pretty confident that method would continue to work, and it did.
This method is how I personally enjoy making sliders best. Starting with practice and focusing on each area of the game allows me to have a controlled environment that makes it easier for me to tune different things independently so I can see how everything interacts.
At first it is a method that can take some time to dial things in, but in the long run I believe it helps speed up my slider making because I have a better understanding on how things behave within the game, so, as an example, when I made my first balanced sets, then wanted to try something new like the LAHC stuff, I had a good enough understanding through the initial testing to set the LAHC up quickly and I could run a sample and be confident in the results letting me know what tweaks were needed, if any.
It's like the current set we're using right now. I knew I wanted to try and open up the levels of play for both the user and CPU, so I adjusted the values I wanted to adjust knowing how it all interacted and I was able to craft up a set, test it out, and plug it into dynasty mode with zero tweaks because it gave me what I wanted visually and statistically. I didn't need to tweak any values because all the labbing when the game launched set me up to be better prepared for scenarios like that now.
I hope that helps and answers your questions.