Questionable result on Search for Matches

Question

The Search Request I used (note that I actually unchecked the Century box):

Limit where we look for matches?	No, search for matches across WikiTree
Limit to one surname in your Watchlist:	(e.g. Jones)
Limit to one person in your Watchlist:	(enter a WikiTree ID, e.g. Zenger-3.)
Strictness for dates:	Match complete dates exactly
Skip people without dates?	Yes, only include people with birth date or death date
Allow for century typos in dates:	(e.g., if checked, 1661 will be treated the same as 1861.)
Skip Pending Merges, Unmerged Matches, and Rejected Matches:

 

The Results I got were:

Possible matches for Mary Margaret Cowan April 17, 1936 Rodney, Aldbourough, Elgin, Ontario, Canada -

• Mary Cowan Abbeville, South Carolina, USA - October 1839 ~ Managed by Joe Sneed. [compare] [merge] [match] [reject]
• Mary Cowan Abbeville, South Carolina, USA - 1874 ~ Managed by Joe Sneed. [compare] [merge] [match] [reject]

It seems to me that, given what I entered for the Match search, neither of these two results should have been displayed. Did I do something wrong?

Answer 1

No, I don't think you did anything wrong, I think the search engine found the closest matches it could.

 

In this case you are blessed, no matches, no confusion.

 

Wail till you get 14 matches, all the same person. Or, 14 matches, son, father and grandfather. Which one do you merge with?  (The correct answer should be 'the earliest one.') but it isn't always.

Answer 2

Dave -

Based on the search you described, at first glance it would appear that you shouldn't have received those results, but if you'll humor me for a moment, I'll try to explain why you did.

The hyphens make all the difference.  (huh?)

Based on what's listed for Mary Margaret Cowan (Cowan-660), there is a birth year of 1936 and no death date (still living I presume).  The results, however, show Mary Cowan (Cowan-65) with no birth date and a death year of 1839 and Mary Cowan (Cowan-76) with no birth date and a death year of 1874.  Since neither of the profiles that were returned had a birth year to compare to Mary Margaret Cowan (Cowan-660), technically the results are correct.

Sometimes when you get what would appear to be strange results, you have to look closely to see where the dates fall, before or after the hyphen (birth date - death date) for it to make sense.

 

Hope this helps!

-Allen

Comments

Dave, great explanation of your question. Showing the example makes all the difference!

Tom and Allen are exactly right. FindMatches is doing the best it can, though admittedly it could do better.

Like Allen says, your Mary Cowan has a birth date but no death date. The Mary Cowans she was matched with have no birth dates but have death dates. Therefore, the birth dates and death dates didn't conflict and the results weren't tossed out.

Of course, your Mary Cowan was born 50-100 years after the other Mary Cowans died! It's obvious that they're not matches, so why doesn't FindMatches know that? Coincidentally, I just wrote the explanation here a few minutes ago: http://www.wikitree.com/g2g/4087/should-wikitree-require-a-dob-range
(See my answer, the question is less directly relevant.)

---

I understand about computer researces and that you have to manage them. But it seems to me that showing apotential match between two people where one died 80 years before the other one was born is decidedly wrong. Is there a middle ground where you could at least check, at the end of the process right before the display, that the potential matches are at least a possibility. The search takes a while now. If it costs us time to reject the impossible matches, that seems the right investment.

It is not productive for me to go through 30 matches for one person who are mostly not born yet, not alive still or not ever lived in the same country; the matching process is nearly a waste of time to do anything with. I know you have a lot of worries, but I am looking at adding another 18,000 people to my tree on Wiki and I don't want to have to go through 200,000 matches where 95% of them are impossible.