Proponents of Fundamental Indexing (FI) suggest that it is more profitable to base portfolio weights on indirectly size-related indicators like accounting data rather than directly on market caps. In noisy marketsà la Roll (1984), it is argued, underpriced stocks (which typically overperform) get underweighted and vice versa. This negative interaction implies a 'drag', which FI claims to avoid.The key question is to what extent the extra return that FI pays really reflects drag avoided rather than style shifts. By way of background check we first look at the proposed changes in the weights across size classes and over time. FI-based weights are biased towards smaller firms; but this size bias is also (i) too large to be just a Bayesian reaction to noisy valuations, and (ii) very unstable over time and across size classes. This means that conventional regression analysis's plagued by unstable exposures. We find direct evidence of this in our own regressions. In addition, the choice of factors and time periods drastically changes the alphas. Regressions, in short, are not helpful To estimate the pure benefits from drag avoidance, purged of style shifts without having to rely on style regressions, we study an investment strategy that should be immune to drag, but without much style shift: we sort stocks into twenty size buckets (vigintiles), and form a portfolio where a stock's weight equals the average value-weight of all stocks in its vigintile. We find no meaningful extra return, whether at the total-portfolio level or per vigintile. Thus, avoiding drag is not why FI does well: drag is empirically unimportant. Most or all of the prima facie benefits must be from time-varying style shifts.