Small Improvements Only

This is a stated project rule, not a fallback for teams that ran out of time. The point isn't a bigger redesign — it's tightening your model where it was weakest, the same "Iterate ≠ starting over" idea from Session 1.

If your Session 21–23 model is already solid, "improvement" can mean tightening precision at the edges of your usable range instead of changing anything mechanical.

Bungee safety still applies.

Build

Step 1: Identify Your Weakest Point

Look at your Session 21–23 calibration data. Where was the model weakest or least precise — widest scatter, lowest confidence at one edge of your usable range, a setting your conditional formatting flagged?

Step 2: Make One Targeted Change

Change one specific thing aimed at that weakness. Not a redesign — a targeted adjustment, the same distinction Session 1 made between Iterate and starting over.

Step 3: Re-Test and Re-Plot

Re-run trials after your change, and update your calibration curve: new plot, new trendline, new R², updated usable range.

Step 4: Log the Change

In your notebook, record:

• What changed
• Why (tied to the specific weakness you targeted)
• How the calibration curve shifted as a result

If your re-test shows no improvement, or a slightly worse fit, that's still valid data — document it honestly rather than quietly discarding it.

Checkpoint

Your notebook entry must show all of the following:

• A specific weakness named — not a generic "we improved it"
• A change that's small and targeted, not a mechanical redesign
• Updated calibration curve with new R² and usable range, even if the result didn't improve

Reflection

Answer the following questions before submitting your work.

1. Did your targeted change actually improve your model, make it worse, or leave it about the same? What does that outcome tell you?
2. Why do you think this project's rule is "small changes only" instead of allowing a full redesign at this stage?
3. Your usable range from today is what you'll rely on for real on Field Test Day. Are you confident in it, or is there a distance range you'd still avoid being tested on if you could choose?

Submit

Submit your updated notebook entry, updated calibration curve, and your reflection.

Milestone Complete