evanmcc
commented
3 years ago @ericjohncarlson it does seem plausible there's some distortion here. I have some theories, but it'll take me a bit to check them out. Please remind me if you don't hear from me on Monday.
In multiple CGs throughout the day, I see a high percentages of validators NEWLY ELECTED into a CG carrying existing penalties. As we're early in the validation process, there are far more validators carrying 0.0 penalties vs. ones that have been selected. As of 5pm EDT, 7/9/21, approximately 620 validators carry a penalty >0.0 with about 1,070 carrying a 0.0. Knowing that carrying an existing penalties should reduce the possibility of election by 10% for each 1.0 of penalty, the chances of the data below occurring are not improbable, but do warrant a look.
This may be nothing, but since we have found other discrepancies, it's maybe worth a post mortem on a few CGs to understand how these validators were elected.
Data: Consensus Election block 912853. CG size of 40, 10 new elected in. Rare-hazel-parakeet elected in with 6.95 penalty Stale-garnet-elk elected in with 6.75 penalty Early-lipstick-spider elected in with 4.90 penalty Proud-viridian-barbel elected in with 4.36 penalty
The next election, block 921921: Steep-stone-octopus elected in with 8.84 penalty. Best-wool-finch with 4.02 Early-fuzzy-gerbil with 2.53 Fluffy-chocolate-owl with 3.37
Looking at each CG after these two today, almost every CG has 30% or more of the new members carrying penalties into the CG.
The data supports that current penalties (as long as they're under 10.0) are not being weighted into the selection criteria properly as the data aligns to a more random selection of validators from the entire pool of validators under 10.0. About 36% of all validators carry a penalty of 0.5 or greater and ~30% of newly elected validators to each CG are carrying a penalty.