aprox.
APROX / A little perspective

Try it. Change one assumption.

A place to experiment before the daily challenge. No account, no competitive attempt and no need to keep the tab open.

How close are your estimates?

Aprox compares numbers multiplicatively. Guessing twice the reference and guessing half of it both produce 2×. An exact match gives 1×. Enter three pairs to see why the combined result uses the geometric mean. All values must be finite and greater than zero.

Estimate 11,2×
Estimate 22×
Estimate 310×
Combined multiplier2,88×

Geometric mean of 1.2×, 2× and 10×. Lower is closer; the best possible value is 1×.

factor = max(guess ÷ reference, reference ÷ guess) combined = exp((ln(f₁) + ln(f₂) + ln(f₃)) ÷ 3)

A reference of 100 and a guess of 200 give 2×; a guess of 50 also gives 2×. The tool calculates in logarithms to avoid multiplying three potentially large factors. It is a learning tool, not a score submission.

How much do your photo assumptions matter?

Change resolution, image count or bit depth one at a time. This model counts uncompressed RGB pixel data: three channels, no transparency, layers or metadata. It does not predict the size of JPEG, HEIC or camera RAW files.

Uncompressed RGB data360 GB

About 335.3 GiB. Decimal GB and binary GiB describe the same bytes in different units.

Read the full photo example to understand the assumptions. For a real backup, measure a sample of your actual files. Adobe’s bit-depth documentation explains RGB channels and bits per pixel.

Take the method with you

Keep the input fixed when the question fixes it. Change one uncertain assumption at a time, check units and explain the resulting scale. The estimation guide gives a practical sequence; the common mistakes guide helps check whether the model still answers the question.