Wrong fee estimation in bitcoinfees

[Little Mouse]

Fees in bitcoinfees.earn.com (2 sats/byte)



Fees in electrum (Much higher)



I think bitcoinfees estimation is wrong, but how can this be such huge different? By this wrong estimation, anyone who have little idea about the fee system can send a fee with low tx and stuck (can be replaced the fee although).
How bitcoinfees estimate the fee? Anyone got an idea?

[bob123]

bitcoinfees.earn.com is known to have a horrible fee estimation.
Usually they overestimate the fee (e.g. recommending 100 sat/vB when 4 would be enough to get a confirmation in the next block).

The fact that they are currently recommending 2 sat/vB, for a "fast and cheap" transaction, probably just means that this is a temporary bug.

However, i'd always advise to not use that site to choose a fee. The estimation almost always is extremely bad.
You are better off checking https://core.jochen-hoenicke.de/queue/#0,2h and using the 3rd graph to check the mempool. Then choose a fee at roughly 0.5MB to 1MB from the tip to get a (very) fast confirmation.

I do not get your point, if bitcoin fee is wrong, then what is the fee displayed on earn.com that is very low which is 2 satoshi/vbyte above? Is it not still bitcoin fee?

OP just meant that they estimate the fee wrong. 2 sat/vB won't get you confirmed within a few blocks, which "fast" would indicate.

[BitMaxz]

bitcoinfees.earn.com is good at giving estimation fees before but right now it's not actually giving you a good estimation. It seems they are delayed compared to other fee estimators.

Much better use other fees estimator here's my list below.

- https://coinb.in/#fees
- https://bitcoinfees.net/
- https://btc.com/stats/unconfirmed-tx
- https://txstats.com/dashboard/db/fee-estimation?orgId=1&var-source=blockchain

You can also use this tool below to manually estimate the best fee at the current network condition(Already suggested above).
- https://jochen-hoenicke.de/queue/#0,24h

[gentlemand]

I gave up on that earn site ages ago. It hardly ever bore any relation to reality. No idea why they don't fix it or take it down. It must've caused an awful lot of inconvenience over the years.

I can't be fucking arsed to figure out that Jochen site. Bitcoinfees.net does a good enough job for me.

[Bitcoin_Arena]

I prefer using mempool observers like;
- https://mempool.space
- https://mempool.observer

These are live and help me know the best fee rate to consider and the position my unconfirmed transaction could get stacked from the tip,

bitcoinfees.earn.com and https://coinb.in/#fees are probably the worst fee estimators around and I hope no one ever thinks of using them. Their estimations are usually way off.

bitcoinfees.earn.com- https://bitcointalk.org/index.php?topic=5183269.msg54563475#msg54563475
https://coinb.in/#fees - https://bitcointalk.org/index.php?topic=5183269.msg54570419#msg54570419

[joniboini]

Did they change the code somehow? I'm sure they were always giving overestimation before (something like 20x more than necessary at worst case scenario).

[o_e_l_e_o]

https://coinb.in/#fees - https://bitcointalk.org/index.php?topic=5183269.msg54570419#msg54570419

This post of mine you have linked to is from when coinb.in/#fees was experiencing a bug, and was showing a fee based on data from old blocks. I'm not entirely sure why the bug happened, but it was fixed a few weeks later and has been working normally again since.

Now that it is working, it remains one of the better sites for suggesting a fee, certainly far superior to bitcoinfees.earn. I think the best site at the moment for suggesting a fee is mempool.space, although the best option remains to judge the mempool for yourself with https://jochen-hoenicke.de/queue/#0,8h.

[NeuroticFish]

Although it's "work in progress" and could show a little better the first 3 blocks' delimitation even when the mempool is full, I'll add this to the list:
https://bitcoindata.science/bitcoin-fee-size-calculator.html

The ANN is here, in case fixes and improvements are needed (I'll just go now there and notify that the blocks are not displayed correctly).

[stompix]

Did they change the code somehow? I'm sure they were always giving overestimation before (something like 20x more than necessary at worst case scenario).

They probably made a mistake in the 0-2 fee calculations.
They have a 0-2 sat/b part and then 1-2 sat/b , the first one shows the estimated time of being Inf which I guess it's infinity.

But it's been nearly 2 days since the OP spotted it, maybe even more and they don't seem in a hurry to fix it, probably because the page was getting less and less traffic and it's not worth the time? Anyhow, not a loss, their algorithm was useless if you wanted to save some $.

I think the best site at the moment for suggesting a fee is mempool.space, although the best option remains to judge the mempool for yourself with https://jochen-hoenicke.de/queue/#0,8h.

I've gone though a lot of those sites and although I keep coinb.in in the bookmarks for its size calculator,  if it's a simple tx mempool.space is probably the best, not only showing all the data you need but I really like the design of the website, especially the preview of the next 5 blocks.

[joniboini]

if it's a simple tx mempool.space is probably the best, not only showing all the data you need but I really like the design of the website, especially the preview of the next 5 blocks.

Yea been using mempool.space for a while now. The website was updated at least three times since I used it afaik, with a focus on its UI. Unfortunately, if a user searches "bitcoin fee estimate" or something similar website like bitcoinfees is usually easier to spot compared to mempool.space or others reliable fee estimators.

[o_e_l_e_o]

But it's been nearly 2 days since the OP spotted it, maybe even more and they don't seem in a hurry to fix it

Looks like it is fixed now. They are back to their stupidly high recommendations. At time of writing they suggest 316 sats/vbyte, which would put your transaction within about 0.005 MB from the tip, which is utterly ridiculous. For comparison, coinb.in/#fees is suggesting 103 sats/vbyte which would put you around 5 MB from the tip, and mempool.space is suggesting between 151 sats/vbyte (low priority) to 246 sats/vbyte (high priority). 246 sats/vbyte would still put you within 0.06 MB from the tip.

not only showing all the data you need but I really like the design of the website, especially the preview of the next 5 blocks.

Yeah. The only thing that I would like them to add is the "mouse over" feature from Jochen Hoenicke's site which tells you the exact numbers on the graph. If they added that, I would use them all time.

[pooya87]

being wrong is when you report 5 satoshi/byte as 7 or 3 instead. and i could see it be "wrong" if they did it for a short time (a month or two) and then fixed their tool. but when for many years they are reporting ridiculously high fees that is when they report 5 as 20 instead that is no longer "wrong" but malicious.

for example right now even with the congested mempool and high fees, the worst case scenario is 30 satoshi/vbyte while the malicious https://bitcoinfees.earn.com/ is reporting 346 satoshi/byte to be the "fastest and cheapest transaction fee". that is more than 10x higher.

[o_e_l_e_o]

while the malicious https://bitcoinfees.earn.com/ is reporting 346 satoshi/byte to be the "fastest and cheapest transaction fee".

As my post above, this fees in current conditions would put you within about 0.005 MB of the tip - almost guaranteed to get in the next block, but still needlessly high. If I was desperate to get in the next block right now, I would probably pay around 170 sats/vbyte, which would still put me less than 0.1 MB from the tip at half the cost.

I'm not sure they are being purposefully malicious, more just misguided. The fee they suggest is certainly the "fastest", in that that it will almost certainly get in to the next block, and could possibly be argued at it is the "cheapest" fee to achieve this with a high degree of certainty. The issue is that most people don't need to be in the very next block, and they do not make it clear on their website that that is what they are aiming for. As you say, a fee over 10 times lower will still likely get you confirmed with 5 blocks or so, which is more than enough for most purposes.