Unsurprisingly for you @fradamt there is some good data and analysis in here and I respect the time and effort you have put into this. You know I have no time for MEVA, especially in the base layer, but as well as that I don’t see that smoothing achieves much here.
“…with the ultimate goal of getting the distribution of rewards for each validator to be as close as possible to uniform: a staker would then get a share of rewards proportional to their stake”
The staker already gets a share of rewards proportional to their stake. Smoothing out returns makes no difference to their profits over time because the expected value does not change however much you smooth. I know first hand because my income over the last 19 years has literally been based on this fact. The ev of a staking pool will always be lower (and therefore less attractive to a rational actor) than the ev of running your own validator, and smoothing does not impact this.
If there was a risk element to validators extracting MEV (ie: if they could lose MEV as well as win it) then you would have some argument because there would be a risk of ruin from the variance. To mitigate this optimally requires the application of the Kelly criterion which benefits those with a bigger bank.
But there is zero risk of loss with MEV extraction, so this is not a factor. MEV extraction is a “bet to nothing” which is one of the most prized outcomes for traders and professional bettors.
So while there is no rational basis for finding a smoothed return to be desirable, you could argue that psychologically stakers prefer smoother returns and a steady income- but that is a subjective opinion.
I could just as easily argue that validators who are already getting a steady income from block rewards and gas might find the lottery aspect of massive outlier wins from MEV extraction more exciting/attractive, especially when the odds are stacked in their favour. The popularity of the national lottery, premium bonds and Las Vegas supports my thesis. Degen culture suggests this is likely also true in the Ethereum community. Anyway, my point is that it is cultural not rational or mathematical.
At the end of the day (and from previous conversations with you I think you get this actually) the MEV problem is not one of how to distribute extracted value, the problem is that value can be extracted from users in the first place.
If we are going to address MEV in the base layer and modify attestations (and I absolutely think that we should) I’d like us to be reducing the MEV that can be extracted (like… ahem… in a content layer).
I would love to see you direct your obviously wrinkly brain away from MEV redistribution and towards mitigation.