Contest Flashback uses the final projections as they stood whenever the last late-swap window was for the games, or in other words, how the projections looked at the last late swap window.
That's deliberate. The point is to measure whether your lineups were a good bet with the information you had. The actual result is one outcome, and a single outcome can't tell you that. So SaberSim runs the slate 100,000 times from the final lock-time simulations against the real field, and the average is what a lineup was worth.
There are a few important implications here:
First, we are using SaberSim simulations for Flashback, which are the same sims that help determine what lineups you should play at lock and when late-swapping. This means there is some inherent bias baked into these numbers.
Second, if you’re on the Ultimate plan and using live sims and live ownership when swapping, sometimes the most profitable swap you can do during the live slate does not line up 1:1 with what the most profitable Flashback ROI would be. As an example, if your lineups are underperforming during the live slate, the most profitable thing to do to earn back min cash equity might be to swap into low-owned, low-projected plays and hope they go off. However, Flashback, grading as if the entire slate was replayed out, might see that as a lower ROI move.
Third, if you make a bet that a player is underprojected and increase your exposure to that player, even if you are right, Flashback ROI won’t see that. It’s grading relative to our sims, so in the same way that your Saber Score or Sim ROI will have gone down in your build, you can expect your Flashback ROI to be a little lower as a result. Flashback uses SaberSim’s default projections, not custom projections or manual projection adjustments.