Single top tW PDF uncertainties are too large
I noticed the large impacts of the single top PDF uncertainties, and checked one sample (tW top 2018): the PDF uncertainties are very large. Most likely hypothesis is that the weights are buggy (samples with 100% correct weights are the exception rather than the norm, thanks to the marvelous LHE format specifications - rant off).
Or, we assumed the wrong PDF uncertainty method... PDF is 306000/NNPDF31_nnlo_hessian_pdfas, so that can't be it.
I haven't checked other files or uncertainties. It would be useful to have a tool to summarize the pre-fit uncertainties for each file or process (only the yields for a given selection, not necessarily all the shapes). Didn't we have somewhere something of the sort?