I use historical stats as a guideline in some instances, but when getting into specific player projections and or tiering I don't put much stock into it. My analysis is much more detail oriented.
??I do projections and let them go from there with no thought about historical data, but historical data is MUCH more than a guideline for predicting what the actual EOY numbers will be. Can't use it to plug specific players' names in there or to make projections - but you can generally tell what this year's EOY variance between players will look like at certain levels - ESPECIALLY WR 4 down to WR20, for example.
The reason historical variances are not useful in making projections is that historical variances are pure numbers with no thoughts to what names are there, or what special circumstances (such as CuMar leading the league in rushing after age 30, and Priest going bye-bye) occur to each player - heck, Mike Anderson has one appearance at RB4 and no other appearances anywhere - historical data is not a predictor of projections, but it is an excellent predictor of how the EOY numbers will actually look come January 2006.
My point is that it is much less important to nail the #4-#8 RBs in order than it is to nail which RBs finish 1-3, and in order. Do your projections as you wish, but it is unlikley this year wil be radically different historically - in all likelihood, WHOever you end up projecting at RB4, with whatever numbers, will probably have numbers much closer to whoever ends up at RB8 than whoever ends up at RB1 - assuming you correctly guess who is RB4. therefore, as I said, it is quibbling over very little - in any year including last year, at most 3 FF points a game - to debate who ends up RB4-8. However, the RB1, whoever it is, consistently outscores the RB4 by something like 4-6 FF points a game - and, consequently, outscores the RB8 by somewhere in the neighbnorhood of 6-10 FF points a game.
The more important debate is who are the #1, #2 and #3 backs - not who is #4 and who is #5.