xs bias when generating small number of events

Hi Andrew,

For a vbf process this is actually not surprising since our phase space integration is not optimal for this kind of topology.

It is actually never advise to ask less than 10k events

Cheers

Olivier

> On 18 Apr 2014, at 19:11, Andrew Levin <email address hidden> wrote:
>
> New question #247212 on MadGraph5_aMC@NLO:
> https://answers.launchpad.net/mg5amcnlo/+question/247212
>
> When I generate 5000 events of a certain process 20 times with different random seeds I get these cross sections:
>
> .15294E+00
> .15200E+00
> .15281E+00
> .15235E+00
> .15267E+00
> .15244E+00
> .15260E+00
> .15171E+00
> .15202E+00
> .15180E+00
> .15184E+00
> .15042E+00
> .15237E+00
> .15250E+00
> .15254E+00
> .15306E+00
> .15226E+00
> .15146E+00
> .15277E+00
> .15179E+00
>
> When I generate 1000 events of a certain process 20 times with different random seeds I get these cross sections:
>
> .14930E+00
> .14674E+00
> .14699E+00
> .14870E+00
> .14798E+00
> .14982E+00
> .14802E+00
> .14868E+00
> .14823E+00
> .14754E+00
> .14894E+00
> .14652E+00
> .14767E+00
> .14823E+00
> .14852E+00
> .14759E+00
> .14916E+00
> .14872E+00
> .14927E+00
> .14994E+00
>
> Clearly, the average is not the same. Is this expected? I would expect that the cross sections I get when generating fewer events have the same average but a larger standard deviation.
>
> The process I am generating is:
>
> generate p p > w+ w+ p p QED=4 QCD=99, w+ > l+ vl
> add process p p > w- w- p p QED=4 QCD=99, w- > l- vl~
>
> Thanks.
>
> Andrew
>
>
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

If I ask for less than 10000 events, are the unweighted events wrong or is only the total cross section wrong?