Improve Initialization of OnchainFeeMonitor with Conservative Synthetic History #210
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
IMHO we should remote the all code because it is not safe anymore do something like this with the current fee env |
|
@vincenzopalazzo do you mean "remove the old code"? |
|
Here is some data from one of my testnet nodes which ends roughly 12 hours after starting w/ fresh history: Here is data for a second testnet node with similar fresh start and time-frame: They both recognize a coherent low-fee env around the same time, looks pretty good to me. |
ksedgwic
force-pushed
the
2024-06-current-fee-samples
branch
9 times, most recently
from
9615654
to
7da2f08
Compare
|
I've modified this PR to use a "hybrid" approach with a little synthetic data. It's ready for review |
ksedgwic
changed the title
ksedgwic
force-pushed
the
2024-06-current-fee-samples
branch
from
7da2f08
to
c991a5b
Compare
|
I pushed an additional commit which logs the h2l, mid, l2h and last_feerate in addition to the decision: |
…ic History Previously, CLBOSS initialized the OnchainFeeMonitor with 2 weeks of synthetic data collected at an arbitrary time on an unknown system. This historical data often failed to accurately determine low/high fee conditions until 2 weeks had passed. This update changes the initialization to use a smaller amount of deliberately conservative history. This approach discourages CLBOSS from prematurely declaring a low-fee environment, while still allowing it to recognize low fees after a few days. The new size is designed to have 50% influence on the lower 20th percentile after 24 hours (24 * 6 * 20% * 0.5 = 14.4). This influence decreases over time: to 25% after two days, 10% after five days, and continues to decay until it has no effect after two weeks. It's important to note that CLBOSS will still function in high fee environments to manage initial liquidity, so this change does not impact its ability to operate effectively.
ksedgwic
force-pushed
the
2024-06-current-fee-samples
branch
from
f7114d1
to
1bb6156
Compare
vincenzopalazzo
approved these changes
Jul 14, 2024
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
CLBOSS preloads the fee history with 2 weeks of some particular history. This is almost incorrect for whatever current feerate environment exists.
By altering the
get_feerate_at_percentileroutine to use the percentile values of whatever data exists (doesn't presume a full history table w/ 2016 values) we can skip preloading any sample data and instead use current data.This means that CLBOSS will learn about the current fee rate environment gradually; it's early impressions will be based on a small amount of data. However, I don't see how basing the history on arbitrary past data is better, it will make good decisions sooner w/ all real data ...