Optimize-all-parameters-vs-some

QUOTE:
... have many parameters, but are not optimizing them all at once. I learned at a workshop that you should optimize ... all parameters at the same time ...

Both approaches are correct but under different circumstances. Let's start with the production strategy that's fully developed.

1) For a production strategy, you should optimize all parameters together but ...
a) Your production strategy shouldn't have more than 10 parameters. I usually try to minimize mine to 5 or 6. You're over fitting if you try to include more, in which case your strategy won't stand the test of time.
b) Each parameter shouldn't have more than 10-15 possible values. I usually try to limit mine between 4 to 6. More than that and the optimization will take forever without much meaningful progress.

2) For a strategy under development, you may have 25 parameters, so you need to divide and conquer to have any hope of optimizing it. And that means optimizing a few related (i.e. interacting) parameters at a time.
a) Your goal should be to figure out which parameters have a low variance across different stocks and can simply be set to a constant for all cases, and which parameters have a high variance across different stocks and may need to be set differently for different stocks or group of stocks (e.g. a dataset with large cap, value stocks).
b) You're also trying to find the correct range limits for each parameter. You want to set those as tight as possible (so the optimizer doesn't get lost in right field), and still cover all the salient values.
c) You want to find the appropriate increment for each parameter. It's probably not time efficient for the optimizer to go over 6 to 8 possible values for a given parameter; 5 possible values is a good choice. Some optimizers like SMAC ignore the specified increment.

I hope that clarifies things better.

Off topic, but the market is constantly changing. So I would pick modern adaptive indicators to use with your strategy over the older non-adaptive indicators. You can learn technical analysis with the older indicators (A good idea), but then choose their modern counterpart that's adaptive. That should stabilize your strategy best with a minimal number of parameters.