# Physics Performance Meeting

Slides: https://indico.belle2.org/event/4302/

## Flat efficiencies in data and bucket 16 results (Daniel Ferlewicz)
* "flat efficiency finder" introduced to get a flat efficiency over pt, theta, etc in data
* proc11+2020a/MC13a vs. bucket 16/MC14ri_a
* large increase of e->pi fake rates in release-05
* Seen in both standard electronID and the chargedBDT
* will study subdetectors next
* e->mu fake rates not changed much
* Jpsi efficiency performance consistent between releases

#### Questions

* When do physics groups need results?
* Marco: It depends on the dataset, and on the deadlines set by the Physics WG (see Ale's intro on 13 May https://indico.belle2.org/event/4394/contributions/21150/attachments/10869/16658/PGM_intro_0513.pdf).
* For analysis of proc11+prompt (ie, rel-4 Moriond 2021 DS), charge-dependent corrections are available for the following "arbitrary" WPs: leptonID > 0.5, 0.9, 0.95.
* For release-5 datasets, we are aiming to provide fixed efficiency WPs. Given the status of proc12 processing, and the fact that some software development is necessary, I foresee them to be ready *not before the end of June*.
* Work is ongoing to prepare workflows using the available rel-5 buckets.
* Which fixed efficiency working points are needed by physics? (outside KS and J/Psi coverage?)
* Marco: We aim at providing standard leptons list in basf2, and the corresponding correction weights (as extraInfo associated to each particle) by reading LID cuts and weights tables from the CDB:
* The flat efficiency working points (tuned in MC) that we wil provide are 50%, 60%, 70%, 80%, 90%, 95%. These are somewhat arbitrary, and analyses that need different requirements should tell us asap.
* We need to possibly adapt existing basf2 modules (eg ParticelWeighting) to produce the lists.
* This approach would allow to seamlessly use the ChargedID BDT for LID, and compare it directly with the likelihood

## Lepton ID (Phill)
* BDT currently works on mdst, we will make it more transparent for the user and include it in the default PID (mostly needed fpr muon ID)

## Upgrade working group (Phill)
* more work on algorithms needed to utilize the proposed upgrades

#### Questions

* How is MC produced for upgrade detectors?
* Not much was evaluated with higher performing readout configurations and dedicated MC for that yet, just higher background levels with default configurations.

## Neutrals (Savino)
* Bias correction included in latest light release, but payload was missing and is being worked on with Frank right now.
* Instructions will be on the neutrals confluence (https://confluence.desy.de/display/BI/Neutrals+Performance) once this is ready

#### Questions for tracking
* Phill: momentum and polar angle dependent corrections?
* Petar: It depends on the correction:
* For momentum scale, we aim to provide a cos(θ) dependent correction instead of the global one.
* For the slow pion efficiency, going from p to (p,θ) correction is tough given the low statistics. The inclusive D0 study is progressing well, so let's see.
* For the high momentum tracking efficiency from taus, as you know, the direction resolution for the missing track is poor due to neutrinos. So although we check the 1-prong (p,θ) dependence (which is correlated to the missing track) we don't bin the correction in 2D. In this regard, for example, D*→[D→KSππ]π with Ks partial reconstruction is superior. Do you know if Daniel plans to continue his very nice work on this? If not, then we need to find some additional personpower within the tracking group.
* For V0 studies with Ks and Λ we check the (p,θ) dependence, although in the end the recommendation is based on the 2D/3D distance.
#### Questions for hadron ID
* semileptonic WG: Is it possible to provide hadron ID correction possible without nCDC hits cut?
* Angelo: the CDC hit requirement is motivated by dE/dx providing very poor separation below that threshold. I would expect that if a track reaches the TOP or the ARICH, it is likely to have >20 CDC hits, so by releasing the requirement the PID performance can only degrade. If you rely on good hadron ID, then it is not recommended to use tracks with <20 CDC hits. If you have a strong case for releasing this requirement, please report it to the Hadron ID meeting.
* Phill: Isolation studies are done for lepton ID and large effects were seen (pions faking electrons), what about e.g. pions faking kaons at low isolation?
* Angelo: we haven't checked this explicitely, but it is certainly something to look into. Thanks.

#### Questions for lepton ID
* Is there any charge asymmetry for high energy leptons? (BELLE2-NOTE-PH-2021-001 (version 3) Table X sees a large charge asymmatry in the highest r-bin which is dominated by high momentum leptopns)
* Marco: charge asymmetries are something that should be studied more in detail, as they are not obvious for lepton ID efficiency. This is for after summer conferences though.

#### AOB
* Performance workshop (like in 2020):may 27th for recorded sessions, 28th for live sessions (at least two to cover different timezones)
* Summary of results on MC13, status of MC14 to get results ready on time
* general topics (like error propagatin etc) can be re-used from last year