EIP-3416: Median gas premium

[mtefagh]

@barnabe

Point 1. I think is fairly uncontroversial, there’s some evidence that the current multiplicative update rule isn’t optimal, though it’s unclear without data in which ways and what the best fix would be.

Of course, it should be discussed. However, just as a shortcut, if people can agree with me that this is the same problem, then we can just borrow the final solution from the economic literature. To me, it seems just like the exact same problem.

@barnabe

This seems incorrect. It’s been shown that unless the effective demand is higher than the block limit

Sure, EIP-1559 is an improvement in this respect. However, we are talking about an additional improvement for the congestion times.

@barnabe

In the “vanilla” 1559 design, BASE_FEE would settle around 6.5, assuming all bidders pay a gas premium equal to 1

I don’t agree with this part as I have shown that there also exist oscillatory equilibriums and I believe that mildly strategic users will converge to those, however, let’s skip this part as I have discussed this before.

@barnabe

Assume basefee is 7.5, then premiums are distributed uniformly in the interval [0, 1.25] and the median premium is 0.625,

At equilibrium, the total fee will sum up to 7.5. Hence, the base fee will be 6.66666 and premiums are uniformly distributed in the interval [0, 1.66666] and the median premium is 0.833333.

@barnabe

In this stationary example, if basefee was lower than 7.5, there would be potentially more users included than the target allows for, and vice versa for basefee higher than 7.5, so 7.5 is indeed the value basefee would tend to.

No, it is not true. The fee cap is on base fee + gas premium. I think the rest of this paragraph depends on this point.

@barnabe

If instead basefee is burned, 0.625 premium isn’t quite enough to cover the miner marginal cost (assuming that cost is 1 Gwei).

If the miner’s marginal cost is 1 Gwei, the block producer won’t include transactions unless the base fee is lower than 6. This can’t converge to equilibrium until the transaction fees are raised from the uniform distribution over [0,10] to uniform over [0,12]. At this price, the equilibrium is reached by the base fee equal to 8 and the median premium equal to 1.

@barnabe

users in the top quartile all declare max_fee = 8.5 such that basefee = 6.5 and users pay basefee + premium = 6.5 + 1 = 7.5.

This will only work under complete information about other transactions’ fees otherwise people in the interval [7.5,8.5] are risking loss. If I ignore the numerical errors that I have mentioned above, I think I get your overall overpayment point. If I am correct, you are trying to say that in any generalization of the second-price auction, if the bids are not sealed, people can cheat by overpayment. Yes, that’s true that the strategy of overbidding is dominated by bidding truthfully in the sealed-bid setting. There are several proofs for it given in game theory. But notice that even if you have some rough estimate about other fees from the mempool, manipulating the median is way harder than the second price and, as in your example, a significant portion of users should simultaneously attempt to manipulate it knowing almost everyone’s else bid.