Muun's estimator allows you to pay more efficient fees for your next-block transactions. What do more efficient fees mean? Not overpaying when it isn’t needed, nor underestimating the fee rate when the network congestion is high.

Fee estimation is hard. Bitcoin's block arrival times follow a Poisson distribution and there is an open market of users competing for block space. Handling the complexity of the fee market is the responsibility of your wallet, not yours.

That was our motivation to built and trained Muun's estimator from scratch. In this post, you can learn how, why, and when you save money using Muun. For real-time savings, visit our website.

What’s different about Muun's estimator

If you aren't using Muun yet, chances are your wallet is getting fee estimations from bitcoind. The main difference with bitcoind, and most fee estimators, is that Muun's is mempool-based: it looks at the current state of the mempool to suggest a next-block fee rate. Other estimators suggest fee rates based on past blocks, transactions, and fee rates.

In all honesty, Muun's estimator does what power users have been doing since the very beginning: checking the state of the mempool to decide the fee for transactions. If this is your case, you can think of the estimator as a super-power version of yourself, trained with lots of data 🦸🏻!

There are cases when both kinds of estimators throw similar suggestions. This happens when the mempool size is relatively constant (fun fact: bitcoind itself was built in a time when the mempool was much more constant than nowadays!). In this scenario, estimators suggest similar next-block-fees and have similar precision^[1].

When the mempool size varies estimators vary a lot as well. Variations in the mempool size happen frequently and may have different causes. Perhaps the simplest one is that a block can take way more or way less than 10 minutes to be mined.

When a block isn’t mined for a long period of time, the mempool gets packed with transactions. If your wallet fails to consider this, it will underestimate the next-block-fee. You will set a fee expecting your transaction to be confirmed in the next block, but it won’t. Spam attacks can lead to the same scenario.

Similarly, if two or more blocks get mined in a shorter period of time, the mempool may lower its size considerably. If your wallet doesn’t take this into account, you will probably end up overpaying fees.

Something similar happens during weekends. Usually, the optimal next-block-fee lowers significantly during weekends due to fewer bitcoin transactions. Most estimators are agnostic to this and they suggest unnecessary high fees. Want to check it yourself? Come visit us during the next weekend!

Benchmarks

In order to benchmark our estimator, we took data from the last year and chose some of the most popular estimators in the industry, including bitcoind and Earn, another mempool-based estimator.

For each estimator, we set the parameters of our estimator to match its precision (higher precision means more chances of confirmation, but more expensive) and calculated the differences of the estimated fees.

  • Bitcoind conservative (precision 82.26%): Muun's fee estimator is 29.89% cheaper, on average, for the same precision.
  • Bitcoind economical (precision 71.89%): Muun's fee estimator is 12.86% cheaper, on average, for the same precision.
  • Earn (precision 87.83%): Muun's fee estimator is 23.75% cheaper, on average, for the same precision.

What’s next

Fees play an important role in a bitcoin wallet, and we believe they will become even more relevant with time. We’ll keep working on improving this estimator making it even more powerfully by leveraging on Replace by Fee (RBF). Stay tuned!

[1] You can think of precision as the probability of getting your transaction confirmed in the next block if you take the suggested fee.