Daedalus stake pool ranking

Hi All,

I thought I understood how desirability and ranking worked for stake pools, but I was surprised at the turn of the epoch. From reading the design spec for staking https://hydra.iohk.io/build/3156413/download/1/delegation_design_spec.pdf specifically paragraph 5.6.1 (pages 38/39). I came to the conclusion that desirability is a function of:

pledge: higher = better
apparent performance: higher = better
a0: higher is more influence of pledge (same for everyone)
cost: lower = better
margin: lower = better

Nowhere in these functions is the current active stake a factor. And that makes sense because the ranking is supposed to be non-myopic so considering long term rewards, when the pool is saturated.

Currently the ranking seems to place large pools at the top. Making it seem like active stake is indeed a factor. Where did I misinterpreted the design document or has it not been implemented according to (this version of the) design?

3 Likes

Interesting insight, thank you for sharing

1 Like

The ranking system is an absolute mess right now. Probably give it a few more epochs to gain the correct traction as envisioned by the research.

1 Like

Sure, I understand it’s early days and there is heavy developments going on. Just until epoch 212 the ranking was more or less in line with what I expected with the exception that apparent performance was artificially kept at 100% for every pool due to the high d parameter. This was also announced.

With epoch 213 it just completely flipped for some pools going from 400+ to top 10 and other way around.

1 Like

Yeah, I thought the rankings had stabilised, too—my high-availability CRAB pool was ranking 1xx on Daedalus, indicated in green, and then it suddenly plummeted to 8xx. Admittedly during that time some stake was undelegated—but equally, I put a fair amount extra in myself. Hopefully, things will stabilise soon. I think I read previously that pool rankings were randomised, but I’m not sure if this is now the case—or indeed, whether it ever was, since unstable initial values could of course give the impression of a random sort.