Confidence Levels
Adjust the confidence parameter to control matching accuracy:
Strict (0.8 - 1.0)
python
find(driver, 'button.png', confidence=0.9)Use when: Element must match exactly, no false positives allowed
Balanced (0.7) - RECOMMENDED
python
find(driver, 'button.png', confidence=0.7)Use when: Good balance between accuracy and flexibility (default)
Relaxed (0.5 - 0.6)
python
find(driver, 'button.png', confidence=0.5)Use when: Element varies slightly, more flexible matching needed
Recommendation: Start with 0.7 and adjust based on results. If elements aren't found, lower it. If getting false positives, raise it.
Measure Instead of Guessing
match_score() reports how close a template actually came, so you can pick a number rather than working down from 0.7 by trial and error:
python
from pyxelator import match_score
print(match_score(driver.get_screenshot_as_png(), 'button.png'))| Score | What it means |
|---|---|
| 0.9 - 1.0 | Solid match. If find() still fails, your threshold is too high |
| 0.7 - 0.9 | Match. The normal range for a real element |
| 0.5 - 0.7 | Weak. Element is probably there but looks different, or the template caught surrounding layout |
| below 0.4 | Not on screen. Lowering confidence will not help |
| None | Unusable template: solid colour, bigger than the screen, or not a readable image |
Do not reach for 0.5 by reflex. A low threshold does not find a missing element, it invents one, and the click then lands wherever the false match happened to sit. If the score is below 0.4, the problem is the page or the timing, not the number.