Augur: Block’s Open Source Bitcoin Fee Estimation Library
At Block, we’ve just open-sourced Augur, the Bitcoin fee estimator that powers all of Block’s on-chain transactions. While Block’s engineering blog post covers the broader motivations behind the project, we wanted to share some technical details here - what we built, how it works, how it performs, and what’s next.
What We Built
Augur is a mempool-based Bitcoin fee estimator designed to generate real-time fee recommendations across a range of confirmation targets and confidence levels.
Highlights:
- Mempool-driven: Estimation relies on live mempool data, not historical block data.
- Reference implementation: A starter Kotlin package is available for developers.
- Public API endpoint: You can query live fee estimates, updated every 30 seconds, at
https://pricing.bitcoin.block.xyz/fees
Here’s a sample response from the API:
{
"estimates": {
"6": {
"probabilities": {
"0.50": { "fee_rate": 2.1135 },
"0.80": { "fee_rate": 4.0277 },
"0.95": { "fee_rate": 6.008 }
}
},
"144": {
"probabilities": {
"0.50": { "fee_rate": 1.0101 },
"0.80": { "fee_rate": 1.0202 },
"0.95": { "fee_rate": 1.1052 }
}
}
},
"mempool_update_time": "2025-05-27T16:09:18.461Z"
}
Interpretation Example: If I pay 1.1052 sat/vB in fees, then I can be 95% confident that my transaction will be confirmed within 144 blocks.
How It Works
Augur builds on the mempool-first approach popularized by @FelixWeis in WhatTheFee. Unlike Bitcoin Core that models fee dynamics based on past blocks, Augur considers the current mempool state, transaction arrival rates, and the inherent randomness of Bitcoin mining.
- Current mempool weight: Transactions grouped by fee rate into buckets, tracking total weight at each fee level.
- Transaction inflow rates: Measured over both short (30 min) and long (24 hr) windows to capture recent market dynamics.
- Block production probability: Statistical modeling of how many blocks will likely be found in your target timeframe using a Poisson distribution.
We took the original WhatTheFee source code and re-engineered it into a production-ready Kotlin library—a somewhat unconventional but practical choice, since much of our internal infrastructure runs on Kotlin. In the process, we changed to updating mempool snapshots and estimates continuously as opposed to just after a new block is mined. We also implemented JetBrains’ viktor for faster numerical array operations.
How We Measured It
We developed a custom benchmarking tool to evaluate Augur’s accuracy and cost-effectiveness against other major fee estimators. Tests were conducted under both high and moderate mempool volatility.
To benchmark, we compared estimated fees to actual transaction fees confirmed in each block. We chose to use the following thresholds based on @ismaelsadeeq’s informative post:
- 5th percentile (p5): The confirmation threshold. Transactions below this fee rate likely wouldn’t have made it into the block.
- 75th percentile (p75): The overestimation threshold. Fees above this level were considered unnecessarily high.
Using these thresholds, we calculated these metrics:
- Miss Rate: Percentage of estimates that were too low (i.e., didn’t confirm, below p5)
- Average Overestimate: How much users overpaid when estimates were high (above p75)
- Total Difference: Composite score capturing both under and overestimation
More benchmarking results can be found in Block’s engineering blog post, but below is one example for a 1-3 block target confirmation during a two-week period of normal-to-moderate volatility (May 2-15, 2025).
| Provider | Miss Rate | Avg Overestimate | Total Difference |
|---|---|---|---|
| Augur | 14.1% | 15.9% | 13.6% |
| WhatTheFee | 14.0% | 18.7% | 16.1% |
| Mempool.space | 24.4% | 21.7% | 16.4% |
| Bitcoiner.Live | 3.6% | 65.5% | 63.2% |
| Blockstream | 18.7% | 44.2% | 35.9% |
Bitcoiner.live achieved its low miss rate by massive overpayment—users paid over 65% more than necessary. Augur, by contrast, struck a better balance between reliability and cost, delivering acceptable success rates at roughly 1/4 the cost.
What’s Next
We’re building an open-source version of our benchmarking tool to enable the community to:
- Reproduce our performance metrics
- Backtest other fee estimators using historical fee provider data
- Contribute improvements to Augur or build their own estimators
We’ll share the benchmarking tool here once it’s ready. In the meantime, feel free to check out Augur and its reference implementation. @zpv and I welcome any questions or feedback in our posts and repos.
Addendum
In our original post we discussed our benchmarking results against various providers and we incorrectly used the label “Blockstream (Bitcoin Core)” instead of “Blockstream”. To clarify, we used the publicly available Blockstream API when retrieving fee estimates and we can’t speak to the underlying algorithms or versions that provide these estimates.
We did not explicitly benchmark Augur against Bitcoin Core but we would like to do that, time permitting, in the future, and add it to our open-source benchmarking tool that is now available. Thanks to @ismaelsadeeq for pointing out that if we do, we should use v28.0+ for its default economical mode.