The flex position models are meant to improve upon making lineup decisions according to “fantasy-points-allowed to this position by the opposing defense”. Instead of that weakly-predictive number, these models are more complex and more accurate. So really that’s it! If you previously used “fantasy points allowed” indicators, then think of using these in a similar way.

If you need more than that, then here are some other FAQs:

• Are the numbers for PPR or standard?

  • So far, the models are only developed for standard scoring.

• Can you compare players within the same week?

  • You can, but you need to process the numbers with some interpretation, since the projections represent the total-team fantasy points, for the given position.

    • To gauge your player’s likely scoring, you need to guess if he will get 20% or 80% or X% of those points. (Some of the low-point teams will load the opportunity onto 1 player, e.g. a single RB.

  • Some rare leagues use rosters with “Team-RB”, “Team-WR”, and “Team-TE”. If that is your case, then just proceed and use them as usual!

  • For WRs, I recommend factoring in the opposing CB, for the player you are considering.

• How do you recommend using the charts? This is what I do:

  • Compare the current week to all the other weeks of the season. (Does this week give more opportunity? Or less?)

  • Compare how your player performed in previous games, relative to expectations.

  • Observe a player’s outlook for the rest of the season. (Does scoring opportunity appear to decrease in the second half of the season?)

• Comparison to “points-allowed”: I have written previously about how fantasy-points-allowed is a weak predictive indicator, concluding the following:

  • For most positions, Points-allowed is a poor measure, with correlation less than 0.10. (Therefore little improvement over a coin-flip.) In comparison, the offensive fantasy "points-for" are more predictive, correlations in the range of 0.15-0.20. Finally, by combining raw data, points-gained, and points-allowed into a single cross-validated prediction model, correlations can reach 0.35 for most positions.

• What situation is considered for projecting Future weeks?

  • The number projections for future weeks reflect how all scores would project if they were played within the current week. This means they account for current-week injuries/absences etc. Maybe not ideal, but it would get very complicated otherwise.