Quite often I get Smart Matches that I have to reject, not because they are not family, but because close relatives seem to be swapped, e.g. two sisters are reported as smart matches in my and somebody else's tree, but "Mary" in my tree is reported to be "Elisabeth" in the other tree and vice versa - and therefore I also get two sets of these matches! In both trees names are spelled correctly and birth/death dates are correct - so it seems the smart matches aren't that smart? How does this happen? I hope you can sort it out and correct it!
The second problem seem to be that the confirmed smart matches seem to be unconfirmed again when one of the confirming parties publish a new version of his/her tree - this produces a lot of extra work, and I sincerely hope this will be changed soon so that confirmed matches will remain confirmed?
I am very glad for the possibility to fetch information from matching trees into my own, but this is a little difficult to find - I saw it once, but it disappeared again before I had the possibility to use it.
After a Smart Match was found for an individual in your tree and confirmed by you, the conditions for matching this individual's first degree relatives with the first degree relatives of the match are less "strict";
This is why sometimes these "false" matches are found, but once you reject them the program knows not to show these results again.
What brings me to the second issue, receiving approved / rejected matches again; Please note that Smart Matches are calculated per family tree, so if you delete your tree from your site and upload a different version of your tree, it will be considered a new tree, and will receive the confirmed / rejected matches as new matches.
Therefore, it's better to publish your tree with the same tree name, so that the existing tree in your site will be updated, and won't be considered a new tree.
Also, I think you'll find the following FAQ helpful:
It is fine that these false matches that we reject will not be shown again - but as long as they indeed should have been there - just sorted the correct way - means that the correct matches also will remain rejected? As I tried to say: the false matches are indeed family, but they are mixed and wrongly sorted by your program. So how can we correct this and accept the real relationship? And why are the less precise "matches" chosen and not the far more correct ones which even has more or less identical names? May be the matching algorithm has a small bug and chooses a wrong match which is found before the real ones are read? These also seem to come more than once in the same collection, which makes me even more suspicious. Real mismatches are OK and quite easy to pick, and they usually occur only once.
Somehow I have got unwanted copies of my tree, then I have managed to delete the unwanted versions, but when I started to work with the correct version again on my PC, both the unwanted copies suddenly were published automatically together with the correct version - so it was a long and tedious work to sort out, detect and delete the unwanted versions. Fortunately I am a quite experienced programmer, but I am sure this would be a hopeless task for normal users.
When time allows I will certainly look at the merging FAQ! Thank you!
The wrong matches are shown in addition to the correct matches, not instead of them.
Also, you can disable publishing for a family project in the Family Tree Builder by opening this project and going to Publishing > Publishing Progress > disable the Automatic publishing (online backup) option.
That is indeed strange. I've found that whether or not that is an issue in any particular case, if one person keeps a bunch of trees which just repeat smart matches I confirm one and just 'ignore' all of their other trees so that they don't keep coming up. One person had about 15 trees, all of which I had to respond to until I ignored 14 of them.