What are your real recovery odds?

The measured speeds behind this calculator

Every rate in the dropdown comes from the official btcrecover performance benchmarks — the numbers the open-source recovery community actually measures, not advertises. “CPU” is a Ryzen 9 9950X (16 cores); “GPU” is an RTX 5090. Where a format is memory-hard (scrypt), GPUs gain little, which is why some rows show the CPU rate for both.

wallet.dat rows use the measured btcrecover reference values (67,908 and 267,488 rounds on the reference wallets). All rows are published measurements from the btcrecover benchmark suite (Ryzen 9 9950X / RTX 5090, July 2026). The MetaMask row is the legacy 10,000-iteration vault — newer vaults store their own KDF parameters in the vault data (read the actual count; 600,000 and 900,000 are both documented), which can run up to roughly 90× slower per guess. These are point estimates — real hardware varies with driver, tool version and your file’s KDF parameters — so treat every result as an order of magnitude, not a promise. Hardware, method and dates are on the benchmarks & methodology page.

How the math works

• Entropy = length × log₂(character-set size). Each bit doubles the work: a 10-character lowercase password is ~47 bits; a 16-character mixed password is ~100+ bits. The jump from “hard for a week” to “hard for the universe” happens around 80 bits.
• Possible passwords = character-set^length. This is the search space — the number of guesses a blind attack would have to try.
• Time = search space ÷ guesses per second. A low-round wallet.dat checked at 472 guesses/sec needs ~2×10²¹ seconds for a 12-character random password — compare that with the ~4.4×10¹⁷ seconds the universe has existed.
• Clues shrink the space, not the speed. Remembering that the password began with your initials doesn’t make the GPU faster — it removes all passwords that don’t fit, which is worth orders of magnitude more. This is why memory reconstruction beats hardware.

What the calculator cannot tell you

• Uncertainty. The output is one number from one benchmark run; the realistic range around it is usually “an order of magnitude either way”. Use it to separate minutes from months — not to bet on 4 hours vs 5.
• Whether your memory is right. It models the password you give it — not the one you actually used. If you feed it a 16-character guess and your real password was 11, the verdict is wrong in the optimistic direction.
• Hardware you don’t have. The rates assume hardware you can actually run offline. Cloud GPU rentals multiply the check rate — at a price.
• Pattern-based shortcuts. Real attacks use wordlists, keyboard-layout masks and year lists that are far smaller than the full space. If your password is patterned (a name, a date, a habit), the estimate sits closer to the “with your clues” column than the brute-force one — read how offline tools actually work.
• The difference between guessing and unlocking. A MetaMask password or a PIN gate may reset without any guessing — the calculator only models encryption that has to be broken offline.

Next steps with your number

• Fast to crack? Fix it now. If your verdict is minutes or hours, the priority is securing the wallet file and doing the 68-place search — your password is likely findable, and you want to be the one who finds it.
• Reachable with clues? Invest in memory reconstruction, then run the numbers again with the better clue level.
• Impossible? You now know the realistic range. Trust the math — anyone who claims otherwise is not doing real work.

Sources & references

1. btcrecover — performance benchmarks
2. BIP39 — mnemonic code for generating deterministic keys
3. ethereum.org — Web3 Secret Storage (keystore specification)
4. BIP38 — passphrase-protected private keys