Is it best to optimize with slippage, or without slippage, and test for slippage after you have your desired data?
Depends on what you understand as "optimize". If you're doing a search or data mining, you may want to remove slippage at first and add it later as the search progresses. It also depends a lot on what time scales you're using and your assumptions as to what "slippage" really is and how it affects market and stop orders. On 1 minute bars it might be important, whereas on 30 minute bars you can tone down the pessimism or even switch it off. Accurate slippage models require a considerable amount of data, but you'll still be left with assumptions. In the end, it's a matter of taste for risk.
IMHO, if you're optimizing, you're already cheating yourself. And if you're doing it without slippage you're cheating yourself even more.
Optimizing can be used in a number of different ways.....it all comes down to what u are doing and how.