Hi All, today Urs and myself have fixed the last problems with the implementation of the kinematic fit in the official recoil code. If you switch on the abcfit flag you will get the following: - 2C fit ( M(X,L,v)=Mbreco, + 4C ) - different resolution functions for Muon and Leptons - The fit is properly converged when IERR.GE.0 (even if the convergence rate is high please do not forget to ask for this flag!) Attached to this mail you find a plot showing the performance of the cfit as a function of the missing mass cut. - upper plot: MEAN of the distribution MX - MXTRUE (red MX=MXFIT , green MX=MXRECO) - middel plot: RMS of the distribution MX - MXTRUE (red MX=MXFIT , green MX=MXRECO) - lower plot: BIAS=MEAN*RMS of the distribution MX - MXTRUE (red MX=MXFIT , green MX=MXRECO) Over a large range in MMISS the fitted quantities are representing an almost unbiased estimator of the true underlying MX value. The reconstructed Mx values from raw data are only unbiased in a narrow window around zero in MMISS. Hence the the kinematic fit will not only lead to a significant better resolution in Mx but also will allow us to soften the cut in MMISS drastically. Regards, Oliver