Holy- but not every player has same chance of playing in week 17. And Saying someone who simply rested in a week 17 the year before is worth 1/15 less or over 6% less in your base to start with is just starting out 6% wrong on that base the next season.
I'm not even sure I completely understand what the problem is.If you don't use week 17 in your league how does it even matter? Who cares if Manning plays or not in week 17 if no one's stats from week 17 are used? all you really care about is relative production. So if ALL of the numbers are (consistently) "inflated" a smidge, it doesn't effect any rankings or anything that matter at all.
As for baselines, a good projection will take those sorts of things into consideration. Last year's stats are obviously not the only thing that's thrown into the mix. If Manning didn't play week 17 last year, and you don't think he'll play in week 17 this year either, it's still a decent starting point for a projection. If you DO think he will absolutely play in week 17, you bump up the projection a little bit. If you think he MIGHT (or might not) play in week 17, you bump it up a little bit less. For my own personal projections, the chance of correctly predicting a week 17 no-show is small enough where I don't think I would use it very often and the vast majority of time would base a projection on a full 16 games anyway.
That approach WILL lead to slightly "off" projections if you don't use week 17 because Mannings projections in terms of points/game might be a little artificially low through the first 15 games (because all of his total production will come in those 15 games vs the typical 16). But frankly that's well within our margin of error on these things and I like the alternative (14 or 15 game projections) a lot less than dealing with this small potential error for a handful of players that you could with ANY degree of confidence assume that they won't be playing in week 17.
Honestly, it would make the most sense to base predictions completely on average points/game, but I don't see that happening for a lot of reasons. The big one is that a casual review won't lead you to compare those numbers to where a guy finished in years past.
We just have to keep these things in mind:
-We always have to adjust projections based on our particular league setup.
-Our goal is figure out RELATIVE per-game production for players, so the net numbers really don't matter all.
But heck, maybe the "projectors" already use a full 16 week production base regardless of whether or not they think the player will actually play in week 17 games.
We should ask them. Any FBG projectors want to chime in on the approach you use?
If you already project based on the assumption that all player play a full 16 games, this whole discussion is moot. The 15/16 rule would apply for all projections without any additional error factors (though again, I'm not sure that the totals really matter that much anyway).
If you factor in the potential for a lack of a week 17 games in your "total" numbers, some folks may need to make some very minor adjustments to their PERSONAL rankings.
ignoring a week based on a perceived change in production that is out of the norm from history is ridiculous.
