The short answer
Forecast the evidence, not the vibe. A deal belongs in commit only if you can name the economic buyer, point at a written next step with a date, and say what the customer has done, not said, in the last week. Best case is for deals with a credible path that are missing one of those. Everything else is pipeline, however good the last call felt. Then call your number early in the quarter and keep a running note of why anything moved, because managers override reps whose forecast changes without explanation, not reps whose forecast is occasionally wrong. Being consistently honest about a smaller number earns you far more room than being optimistic and right once.