A fair upgrade comparison uses cash earned over the same time, under comparable conditions. A damage number, a chest full of unsold items or one unusually lucky run is not an income rate.
Record a sale, not just mining activity
Choose one repeatable route
Use the same crew, zone, temporary boosts and selling policy. Pick a wall-clock duration long enough to include your normal collection and selling trips. Use that exact duration for before and after.
Clear or separate old inventory
Exclude ores mined before the sample. Record starting cash, mine during the window, collect the fresh chest output and select those ores at Shops/Sell items. Record ending cash after selling.
Remove unrelated balance changes
Subtract gifts, code rewards or other unrelated cash inflows. Recurring mining/selling expenses already reduce the cash difference: do not add them back or subtract them a second time. Keep the one-time upgrade cost outside the sample and enter it separately in the calculator.
Repeat, then change one thing
Record at least three baseline runs, keeping the full range. Make one identified upgrade or research change and repeat the same duration and conditions. If the purchase happens in the middle of a sample, restart the measurement.
+ unrelated one-time outflows
Only add back a one-time outflow unrelated to the run. Normal operating expenses stay included. If it changed the mining setup, rerun instead.
SYNTHETIC ARITHMETIC · NOT GAME DATA
Use the median and keep the range.
Each invented example run below lasts 10 minutes. No rewards or expenses occur in these examples. These numbers demonstrate the calculation; they are not My Anime Mine rates or calculator defaults.
| Run | Before: cash start → end | Earned | After: cash start → end | Earned |
|---|---|---|---|---|
| 1 | 1,000 → 1,100 | 100 | 1,000 → 1,150 | 150 |
| 2 | 1,100 → 1,220 | 120 | 1,150 → 1,290 | 140 |
| 3 | 1,220 → 1,300 | 80 | 1,290 → 1,450 | 160 |
| Median | Range 80–120 | 100 | Range 140–160 | 150 |
The median scenario changes from 10 to 15 cash/minute, a 50% arithmetic increase. Enter 100 before and 150 after with a 10-minute window. Enter the upgrade's separate current cost; do not subtract it from the after-run earnings.
If a 100-cash reward arrives during a sample, subtract it from that sample's balance gain. If a recurring run expense of 20 was paid, the ending balance already includes that expense.
Test a less optimistic scenario.
Using the high baseline sample (120) and low after sample (140) changes the comparison to 12 → 14 cash/minute. The extra rate is now 2 rather than 5 cash/minute; the same purchase takes 2.5 times as long to repay from added income. This is a sensitivity check, not a statistical confidence interval.
If you have not bought a candidate yet, its after-income is an estimate. Mark it that way in the calculator. The result then answers “what if this rate occurs?”, not “this upgrade definitely earns this rate”. Do not replace a missing observation with a research health/damage percentage.
Compare your investment →Check a research effect before using it → · Read the evidence boundaries →