This is an old revision of the document!
Validating Match Scores
If you are getting unexpected matches when fuzzy matching, it might be because of a balance issue with scoring.
It's very common to have to tune your scores over time as you encounter new situations and find the right balance between all of your different scoring criteria.
You can see the calculated score for a match by pressing the Explain button when entering match details.
Match Details
You'll see an explanation of your scoring on one side, and the actual numbers on the other in a table.
For example, if your captures included Temperature, Pressure, Vibration and Speed, you'll see…
| Attribute | Amount | Total |
|---|---|---|
| Temperature | 0 | 0 |
| Pressure | 0.2 | 0.2 |
| Vibration | 0.8 | 1 |
| Speed | 0 | 1 |
The Attribute column is the name of the capture attribute you're searching for.
The Amount column is the total amount of score penalty you're applying because of the difference in value between your search criteria and the actual captured value. This includes any multipliers you've assigned.
The Total column is a running total of the score being given for this match.
So in the example above, the largest contribution to the score is vibration. If you wanted to make the vibration matter less to the final result, you could reduce its multiplier. If you wanted it to matter even more, you could increase it.
You can also choose to make some penalties exponential.
Looking At Other Captures
Your URL in the address bar contains a parameter at the end called id. This is the unique identifier of the capture, which is found on the left-hand side of your capture list.
By replacing the id in your address bar, you can compare the scoring of the match you actually got to to the scoring of the match you wanted to get to see what caused the differences.
For example, if you are searching and getting batch 2345 instead of batch 1234 which you want to get, you can take the existing URL…
http://myserver/capture/matchexplain?speed=123&id=2345
…and replace the '2345' with '1234'.
You can then directly compare the two results to find out why one scores less than the other. Remember that the lowest score is considered the best match.
You can then adjust your fuzzy matching settings to help avoid this collision in the future.