delvingbitcoin

Mempool Based Fee Estimation on Bitcoin Core

Mempool Based Fee Estimation on Bitcoin Core

Original Postby ismaelsadeeq

Posted on: April 1, 2024 20:35 UTC

The analysis and adjustments made to the format for fee estimates have led to more visually appealing and informative data representation.

Thanks to additional data provided by ClaraSh, a comprehensive range between the 5th percentile fee rate to the 75th percentile fee rate has been mapped out effectively on a graph. This newly formatted data reveals that the initial assumption regarding the underestimation of threshold-based estimates may not be as significant as previously thought. In most cases, these estimates accurately fall within the target block fee rate range specified.

A detailed observation summary highlights the performance of the CBlockPolicyEstimator::estimatesmartfee in its economic mode, with a focus on confirmation target 1. The findings show that 5,643 estimates, or 29.46% of the total, were overpaid; 2,114 estimates, accounting for 11.04%, were underpaid; and the majority, 11,397 estimates (59.50%), accurately fell within the designated range. When comparing these statistics to mempool-based estimates without a threshold, one notes a slight difference in accuracy. Specifically, without the threshold, 0.05% of estimates were overpaid, 18.99% underpaid, and a significant 80.96% fell within the range.

Further examination of mempool estimates with the implementation of a threshold for CBlockPolicyEstimator::estimatesmartfee in economic mode confirms a nuanced shift in accuracy: only 0.03% were overpaid, 26.47% underpaid, and 73.50% fell within the desired range. This data underscores the efficacy of using the 5th percentile as the low end and the 75th percentile as the high end for fee rate estimations.

For those interested in exploring or modifying the data further, access is provided through a collaborative platform with a link to Colab. Additionally, the original dataset and analysis can be found and run through a notebook available in the repository at GitHub, facilitating broader engagement with the findings. This approach not only enhances transparency but also invites further scrutiny and potential improvements to the fee estimation process.