Problems with new shower treatment in top decays
As discussed before, the changes in 8c1f96c5 cause significant differences in basic observables in ttbar samples, like leading b-jet pT or b-jet fragmentation function:
(more of these plots in https://wwwpub.zih.tu-dresden.de/~fsiegert/tmp/plots-224-225-had/ (*))
This was noticed in ATLAS both because b-tagging efficiencies are significantly different between <=2.2.4 and >=2.2.5 and because the jet multiplicity distribution and gap fractions are quite different and not necessarily favoured by data:
(more of these plots in https://wwwpub.zih.tu-dresden.de/~fsiegert/tmp/plots-224-225-dileptonic/index.html (*))
(Sidenote: the 2.2.4 samples look very similar to Powheg+Pythia8 in observables like the ones above.)
I know that the changes are theoretically motivated and benchmarked against WbWb NLO QCD in 1709.08615v1 vs. 1709.08615v3, but could there either be a bug in the bugfix, which just didn't show up in that benchmarking, or a need for re-tuning of any ISR or FSR related parameters to counteract these decay-shower changes?
(*) These plots are with a simplified setup with ttbar+0,1j@LO and without UE, but very similar (and partially even stronger) differences are found when comparing ttbar+0,1j@NLO+2,3,4j@LO samples in ATLAS between 2.2.4 and 2.2.5/2.2.6.