Initial strategy testing should go without transaction costs
First step - testing without transaction cost. If results are disappointing, you throw away this strategy, nothing can be done here. If results show something promising, go to step two and include your transaction costs. Some people perform initial testing with costs included and throw away anything that doesn't show profit. I don't think this is right. When testing, you want to answer two questions. First one: "is there any edge around this strategy?" If answer is "yes", you go to second one: "can I exploit this edge with my transaction cost structure?". If answer is "no", you may want to change strategy rules somehow to make turnover lower, or use different instrument with different transaction cost structure, or something else.
When you initial test with costs, you saving you time by throwing away potentially profitable strategies,