1: \begin{abstract}

2: There is a $\sim 150\kms$ discrepancy between the measured  motion of the Local Group of galaxies (LG) with respect to the CMB

3: and the linear theory prediction  based on the gravitational force field of the large scale structure in full-sky redshift surveys. 

4: We perform a variety of tests which show that  the LG  motion  cannot be recovered to better than 

5: $150-200\kms$ in amplitude and within a $\approx10^\circ$ in direction. 

6: The tests rely on catalogs of mock galaxies identified in the Millennium simulation using semi-analytic 

7: galaxy formation models. We compare these results to  

8:  the $K_s=11.75$ Two-Mass Galaxy Redshift Survey, which provides the deepest, widest and most complete

9: spatial distribution of galaxies available so far.

10: In our analysis we use a new, concise relation for deriving the LG motion  and 

11: bulk flow from the true distribution of galaxies in redshift space. 

12: 

13: Our results show that the  main source of uncertainty is the small effective depth of  surveys like the 2MRS that

14: prevents  a proper sampling of the large scale structure beyond $\sim100\hmpc$. Deeper redshift surveys are needed to 

15: reach the  ``convergence scale" of  $\approx 250\hmpc$ in a $\Lambda$CDM universe.

16: Deeper survey would also mitigate the impact of the ``Kaiser  rocket" which, in a survey like 2MRS,

17: remains a significant source of uncertainty.

18: Thanks to the quiet and  moderate density environment of the LG, purely dynamical uncertainties of the linear predictions

19: are subdominant at the level of $\sim 90\kms$. 

20: Finally, we show that deviations from linear galaxy biasing and shot noise errors provide a minor contribution to the total error budget.

21: 

22: 

23:  \end{abstract}

24: