Understanding the clustering of dark matter halos (halo bias) is crucial
to successfully using large scale structure as a cosmological tool.
First I will describe how a scale-dependence of Lagrangian halo bias in
Fourier-space can be captured by an appropriate *real*-space
measurement. I will discuss recent results showing that the excursion
set approach, which works with random walks in the smoothed real-space
initial density field, naturally predicts the form of the Fourier-space
bias coefficients.
When combined with the idea that, in order to be physically meaningful,
one must only count walks centered at peaks in the density field, this
leads to predictions for the large scale (scale-independent) linear bias
that are in good agreement with results from N-body simulations. I will
also show preliminary measurements indicating that the scale-dependence
at smaller scales is also captured well by this excursion set peaks
model.