The motivation to revisit Track by Track analysis (TbyT) procedure was observation difference of M² for proton and antiproton in fast production for 9p2GeVc 2020 data sample (see https://drupal.star.bnl.gov/STAR/system/files/kfQA_0.pdf)
    The revision is required because :

• we use  StiCA which allows reuse of hits by multiple tracks (multi map<hit,track>).
• new version of CA is trying to combine loopers in a single track. Thus the single track from one reconstruction can be associated with multiple track from another reconstruction (multi map<track,track>).

    The revision has been released (TFG) and herein I report  the new TbyT results: 
        https://www.star.bnl.gov/~fisyak/star/TbyT/2019_2020/ 
    We have 4 comparisons :

1. 2020 9p2GeVc old = "dev" (fast office), new = ".DEV2" (TFG)
2. 2020 9p2GeVc old = ".DEV2", new = "dev". To test TbyT procedure. We expect "old" <=> "new".
3. 2019 19GeV     old = "SL20c", new = "TFG20c" (The sample has been provided by Guannan Xie)
4. 2020 9p2GeV   old = "TFG20c", new = ".DEV2". To see that we have no differences between the last stable release (TFG20c) and development version (.DEV2) i.e. .DEV2 ~ TFG20c.

   Conclusions:

1. The swap old, new => new, old does work (the first and second columns).
2. There is no difference between .DEV2 and TFG20c (the fourth column).
3. There is practically no difference between dev_.DEV2 and SL20c_TFG20c (the first and third column), only statistics.
4. From the first (and third) column(s) we can see:
   • TFG has higher efficiency for low p_T global tracks (Fig.1),
   • TFG has wider pseudo rapidity (η) acceptance (Fig.2. This was the main reason for upgrading to new  CA (Grigory Kozlov, FIAS)).
   • With new CA overall global track efficiency increased from 94% to 98% (Fig.3)
   • The rate of clones reduced from 8% to 1% (Figs. 4-6).
   • The rate of lost is just complementarily to efficiency (Figs. 7-9).
   • The same we can conclude for primary tracks (Figs.10-18).
5. The comparison of track p_T versus no. of  fit points, p_T, η, φ are show in Figs. 18-30
   
   • There is rather strong dependence of relative difference of  p_T (δp_T) on no. of fit points (Fig.21) which increased with increasing of no. of fit points.
   • There is very strong dependence of δp_T on p_T for dev to .DEV2 comparison ( ±2%/GeV) which is much stronger than SL20c to TFG20c comparison ( ±0.2%/GeV).
   • There is ~2 mm shift between Z of primary vertex for dev to .DEV2 comparison.
   • There is ~1 mm shift between X of primary vertex for dev to .DEV2 comparison.
6. I try to track differences between dev and .DEV2 reconstructions. These are man differences
   • • 0.1% lower MagField (due to switch from scaleFactor to current)
   • • TpcInnerSectorPositionB,
   • • tpcEffectiveGeomB,
   • • tpcSectorT0offset,
   • • tpcOuterHitError,
   • • tpcInnerHitError => iTPCHitError