Step 1: Build a baseline
Before forcing takes into the optimizer, see what the current projections and strategy assumptions produce. A baseline gives you something to compare against and makes the cost of every later restriction visible.
Step 2: Review the important disagreements
Which players are you much higher or lower on than the field?
Which games or stacks dominate the build?
Which assumptions are sensitive to injury, lineup, weather, or role news?
Where is duplication or portfolio overlap higher than expected?
Step 3: Revise selectively
Change the inputs or constraints only where you have a real reason. Then rebuild and inspect the impact. If a uniqueness rule destroys the viable lineup pool or a fade removes too much expected value, that is useful information — not a reason to hide the tradeoff.
How JARVIS supports the loop
Model first, opinions second
The JARVIS strategy philosophy is to surface important disagreements, apply targeted source or human takes, re-evaluate the result, build a diversified portfolio, and adapt to new information. It should not turn every analyst opinion into a hard optimizer rule.