Fee Estimation Accuracy Compared
Wallets estimate what a transaction will cost. We recorded estimates against actual costs for a month and the variation between wallets was substantial.
Every wallet shows an estimated fee before you confirm. The estimates are not equally good and the differences show up as stuck transactions and overpayment.
What is being estimated
Two things.
Gas used, which depends on what the transaction does. For a simple transfer this is fixed. For a contract interaction the wallet simulates it, and simulation can be wrong if conditions change between estimation and execution.
Gas price, which depends on network conditions. Since the base fee for the next block is deterministic, this should be accurate for the immediate next block and becomes a forecast beyond that.
Where estimates go wrong
Underestimating gas used. Produces a failed transaction that consumes the full limit. This happens when conditions change between simulation and execution, and it is the expensive failure.
Overestimating gas price. You pay more than necessary. Invisible, because the transaction succeeds.
Stale base fee. A wallet using a cached value during a rapidly rising fee environment produces transactions that do not confirm.
Poor speed tiers. Some wallets offer fast, average and slow without meaningful differentiation, particularly since the priority fee is a small part of the total in quiet conditions.
The method for testing your own
For each transaction over a month, record the estimate shown and the actual cost from the explorer.
Compute the ratio. A wallet consistently estimating above actual is costing you a small amount on every transaction. One estimating below is producing failures.
Also record whether transactions confirmed within the expected time for the tier selected.
What we found
Across the wallets we tested, estimates for simple transfers were broadly accurate everywhere, because the calculation is straightforward.
Contract interactions were where wallets diverged. Some added a substantial margin to the gas limit, which is safe and occasionally results in a higher-than-necessary limit being authorised, which costs nothing unless the transaction fails.
Others estimated tightly, which produced failures during volatile conditions.
Given that a failed transaction costs the full gas consumed and delivers nothing, a wallet that overestimates the limit is strictly preferable to one that underestimates it.
The feature that matters more
Transaction simulation showing expected balance changes, which several wallets now offer.
That is not a fee feature and it catches a more expensive category of error than any fee estimate improvement, which is why we would choose a wallet on that before on estimation quality.
The practical rules
Do not reduce the gas limit manually. There is no saving and there is a failure mode.
Check the estimate against a gas tracker when it looks unusual.
For withdrawals from a venue, the network fee is handled by the platform. Venues that show it separately from their own charge, such as venues that quote the network fee separately, let you see whether their estimate is reasonable before confirming.
Fees move with network demand. Any figure here reflects the date shown above and should be re-checked before you act on it. Where a measurement was taken by hand, the article says so.
Related reports
- How to Read a Fee Estimate Before You Confirmtransaction troubleshooting
- The Cost of Dusttransaction troubleshooting
- Failed Transactions: Why You Still Pay for Themtransaction troubleshooting