Estimators like EthGasStation and Etherscan are important tools for Ethereum users due to the uncertainty in transaction processing times.
The accuracy of transaction processing time estimates by these platforms had not been empirically studied prior.
Data from EthGasStation (Gas Price API and Prediction Table API) and Etherscan (Pending Transaction page and Gas Tracker) was used to evaluate prediction accuracy.
Accuracy was determined by comparing actual processing times with predictions, calculating absolute error for each studied transaction.
Models were ranked based on AE distributions using statistical tests and effect size calculations.
The Alpha centrality method was used to assign ranks to models based on their performance in predicting transaction processing times.
At a global level, all four models were found to have equivalent prediction accuracy with negligible effect sizes.
Stratifying the analysis by gas price categories showed varying performance, with specific models excelling in different price categories.
Developers aiming for high prediction accuracy may need to utilize multiple models from different estimation services based on gas price categories.
Etherscan Gas Tracker performed better for cheaper transactions, while EthGasStation Gas Price API excelled in other gas price categories.