How do you test whether an upgrade helped?
Record the rate and label before the upgrade, change one thing, then record the same fields after it. Keep the boost, pet, route and session as close to unchanged as possible. If several things changed, the result is a description of the new setup, not evidence about the upgrade.
This method works for a mutation, a pet replacement or any other change that claims to improve income. It does not require knowing the internal formula. It requires a baseline, one change and a record that another person could inspect.
Run a controlled A/B test
Record the before-state
Copy the displayed rate, pet label, boost state and date. If the interface shows a final boosted rate, note that the value already includes the boost.
Make one change
Apply the upgrade without adding a new pet, a new boost or a different route. If the game forces another change, record it as a confounder rather than ignoring it.
Record the after-state in the same view
Use the same interface and, if possible, the same location. A value read from a different panel may be measured differently. Copy the label as well as the rate.
Repeat or label the result provisional
One pair can be affected by a rounding change, an event or a UI update. Repeat the pair later if the decision is expensive, and keep both results instead of replacing the first one.
Upgrade test record
| Target | Field | Before | After | Note |
|---|---|---|---|---|
| Boost state | Boost state | None or multiplier | Same if possible | A changed boost is a confounder |
| Displayed rate | Displayed rate | $/s | $/s | Same interface |
| Pet label | Pet label | Exact label | Exact label | A prefix may change |
| Rate change | Rate change | Baseline | After minus before | Keep the sign |
| Upgrade cost | Upgrade cost | — | Displayed cost | Needed for payback |
How do you read the result?
If the rate changed and nothing else did, you have an association between the upgrade and the change. If the boost, pet or route also changed, the result is inconclusive. A negative or zero change is still useful: it tells you not to repeat the purchase under those conditions.
A percentage change uses the before-rate as its denominator. A change from $4/s to $5/s is a 25% increase, not a 20% increase. State the denominator so the result is not misread, and keep the raw rates alongside the percentage.
What counts as a confounder?
A confounder is anything that changed alongside the upgrade: a boost, an event, a different pet, a new route, a UI update or a server change. If a confounder is present, the before-and-after difference cannot be assigned to the upgrade alone.
Record the confounder instead of deleting the observation. A labelled uncertain result is more useful than a clean-looking conclusion with a hidden variable. If the decision is expensive, rerun the pair when the conditions are stable.
When to stop testing
- The decision no longer depends on the difference.
- The cost of the next test exceeds the value of the answer.
- The game changes between the before and after records.
- You cannot keep the other conditions stable.
- The result is already consistent across two clean pairs.
Put the result to work
Calculate the payback
Use the observed rate change and the upgrade cost.
Open the calculatorCheck mutation evidence
Separate an observed label from a verified multiplier.
Read the mutation guideUpgrade testing questions
Can one before-and-after pair prove an upgrade works?
It shows a change under the recorded conditions, but one pair is weak evidence. Repeat it if the decision is expensive, and check whether a boost, event or UI change could explain the difference.
What if I cannot keep the boost constant?
Record the boost state for both sides and label the result as confounded. Do not compare a boosted after-rate with an unboosted before-rate and call the difference an upgrade effect.
How do I calculate the percentage change?
Subtract the before-rate from the after-rate, divide by the before-rate, then multiply by 100. The before-rate is the denominator. Keep the raw rates so the calculation can be checked.
Sources and evidence boundaries
- Fake an Egg — official Roblox experience
The public experience description identifies Fake an Egg by Slime Time Studios and its egg-disguise and pet-stealing premise. It does not supply a pet database, current codes, hatch odds, income multipliers or a carrying-speed formula.
Accessed - Official Fake an Egg experience icon
The icon contains an offline-mutation promotion. It does not establish mutation names, odds, formulas, offline duration or fuse effects.
Accessed
