Minor update on transaction fees: users still don’t care.

I ran some quick numbers on the last retargeting period (blocks 415296 through 416346 inclusive) which is roughly a week’s worth.

Blocks were full: median 998k mean 818k (some miners blind mining on top of unknown blocks). Yet of the 1,618,170 non-coinbase transactions, 48% were still paying dumb, round fees (like 5000 satoshis). Another 5% were paying dumbround-numbered per-byte fees (like 80 satoshi per byte).

The mean fee was 24051 satoshi (~16c), the mean fee rate 60 satoshi per byte. But if we look at the amount you needed to pay to get into a block (using the second cheapest tx which got in), the mean was 16.81 satoshis per byte, or about 5c.

tl;dr: It’s like a tollbridge charging vehicles 7c per ton, but half the drivers are just throwing a quarter as they drive past and hoping it’s enough. It really shows fees aren’t high enough to notice, and transactions don’t get stuck often enough to notice. That’s surprising; at what level will they notice? What wallets or services are they using?

8 replies on “Minor update on transaction fees: users still don’t care.”

  1. One reason for paying higher fees than the “mean fee needed to get into a block” is to account for variance.

    On average, the mean fee needed might be 16 sats/byte, but I might want to be reasonably certain that I get into the next block. In that case, the mean won’t be sufficient.

    For example, there’s a 10% chance that the next block will take more than 23 minutes to arrive. What’s the fee that I need to pay to get into the next block with 90% probability?

    The Bitcoin Core fee estimation is based on this criteria, using a 95% success probability. As of this moment, the estimate for confirmation in 2 blocks is 51 sats/byte on my node.

    Of course, I’m not claiming that users are indeed following this logic, just that the fees are not really excessive, at least according to this criteria.

    1. Indeed, bitcoin core is very conservative in its estimates; I can’t get mine to estimate below 18.5 satoshi per kb (at 25 blocks). Part of this may well be because of all the overly generous fees being set by at least 55%.

      Looking at the last 25 blocks (ending 416359), 18.5 satoshi fee rate would get me into 14 of them (though it seems 6 of those allow lower fee txs due to priority, so say 9 of 25).

      Asking to get into the next block no matter what should not be the default behavior for a wallet: users care about time, not blocks, and there’s diminishing returns in raising fees.

      1. I think the issue is that luck is major factor – the fee needed to get into the next block depends a lot on how lucky block solvers have been, or whether there are currently any transaction surges or lulls.

        For example, starting where you left off at block 416359, according to my data, 18.5 sats/byte would not have been sufficient to get into any of the subsequent 25 blocks.

        Similarly, going backward from block 416334, 18.5 sats/byte would not have been enough for the previous 16 blocks.

        (I also obtained 9/25 for blocks 416335-416359.)

        So paying the mean doesn’t seem viable to me because of the strong correlation in the required fee rate from block to block; there can be long stretches where the required fee is consistently high, and vice versa for low fees during lulls.

        Ideally, users should use wallets that can adjust rapidly to transient mempool surges, but I don’t know any that really do that at this time.

        1. Indeed, you really need rbf to do this without a very nasty penalty for cutting it too fine. But if people are still using fixed fees today, it’s clear that kind of rollout well be very, very slow…

    2. RBF and nLockTime will help with this. You can set your fees aggressively, but also ramp up quickly, with tx variants paying higher fees setting higher nLockTimes.

  2. Would you mind sharing the code used for this analysis? It would be a great bootstrapping tool for a newbie like me.

  3. Did anybody model what is likely to happen if amount of bitcoin users doubles during a year?

    Do you think there is a risk that the fees will quickly become the main limiting factor for Bitcoin’s growth and the growing demand for permission-less transactions will be satisfied by some other coin?

Comments are closed.