FTDNA kits match on Gedmatch but not here?

[sfrank]

I put my kit on GEDMatch and found a lot of FTDNA kits that were relatively strong matches (3rd and 4th cousins) but did not show up on my matches on FTDNA. Has anyone else experienced this? Its frustrating to know there are relatives out there on FTDNA, who have matching people in their Gedcoms, but I can't see them because for whatever reason the algorithm didn't put them in my match list.

[MitchellSince1893]

It could be that your total shared cM is less than FTDNA's threshold of 20cM.

For example if you shared a single 19cM strand with another individual, it would show up on gedmatch as possibly a 4th cousin, but it wouldn't show up on FTDNA.

[Earl Davis]

Does anyone understand Family Tree DNA’s rationale for this? It’s something I have ever really understood.

I can understand to some degree them not showing say an 8cm largest with not other segments as probably being an IDS match but I don’t understand why they don’t (if this is true) for example show a 14cm largest with additional 2cm and 3cm segments and 19cm total.

Earl.

[Rebekah Canada]

The Matching Rules

Hi,

I think in most cases it is the rule for the second largest block that causes matches to be dropped. We really need phasing to sort out many of the true IBS cases so that we can re-evaluate matching. It will happen this year.

Does anyone understand Family Tree DNA’s rationale for this? It’s something I have ever really understood.

I can understand to some degree them not showing say an 8cm largest with not other segments as probably being an IDS match but I don’t understand why they don’t (if this is true) for example show a 14cm largest with additional 2cm and 3cm segments and 19cm total.

Earl.

[sfrank]

It could be that your total shared cM is less than FTDNA's threshold of 20cM.

For example if you shared a single 19cM strand with another individual, it would show up on gedmatch as possibly a 4th cousin, but it wouldn't show up on FTDNA.

As an example, total shared for one was 48.8, longest segment 11.1. It seem like it would meet the threshold. And that's at default settings. If I lower gedmatch to 1cM on one to one comparison, it comes up as 98.3cM. Not sure where FTDNA puts its cutoff, but either way, should be a match.

[madman]

gedmatch cm is way different then here.


ftdna match 20cm

gedmatch same match 60cm


another one

ftdna match 20cm

gedmatch same match 45cm

As an example, total shared for one was 48.8, longest segment 11.1. It seem like it would meet the threshold. And that's at default settings. If I lower gedmatch to 1cM on one to one comparison, it comes up as 98.3cM. Not sure where FTDNA puts its cutoff, but either way, should be a match.

[madman]

Those gedmatch totals were at a Total of segments of 5 cM or more so if it was set at 1 cm the total would of been way higher.

gedmatch cm is way different then here.


ftdna match 20cm

gedmatch same match 60cm


another one

ftdna match 20cm

gedmatch same match 45cm

[Rebekah Canada]

Hi,

How have you set the SNPs per block level?

Those gedmatch totals were at a Total of segments of 5 cM or more so if it was set at 1 cm the total would of been way higher.

[madman]

I didn't use the SNP analysis utilities it was One-to-many matches

Hi,

How have you set the SNPs per block level?

[S9 H9]

Hi,

I think in most cases it is the rule for the second largest block that causes matches to be dropped. We really need phasing to sort out many of the true IBS cases so that we can re-evaluate matching. It will happen this year.

Am I understanding you correctly, that FTDNA will phase autosomal SNP files?

If so, are you going to implement a BEAGLE-like approach? Will you require a parent to be tested also?

[hansonrf]

GedMatch, as they compare between different testers and datasets, is more forgiving of no-calls within the data and places where the different datasets may look at different SNPs.

To truly answer your question would require both data files and knowledge of both GedMatch and FTDNA algorithms; Not likely to happen but I'll check back regularly...

Bob H.

Those gedmatch totals were at a Total of segments of 5 cM or more so if it was set at 1 cm the total would of been way higher.

[sfrank]

I actually see the opposite of what madman described. For the people I know on both FamilySearch and Gedmatch, I see lower cM totals on Gedmatch. For example, using the default for one-to-one match on gedmatch (500 SNPs, 3 cM) I have three people that come in slightly under the FTDNA totals when I look them up on Gedmatch. So I don't think the analysis methods should be so dissimilar as to cause a total shared 40-50cM match on gedmatch to not register on FTDNA.

What exactly are the minimum thresholds for cM lenth, SNPs, etc for a match on FTDNA? Is there somewhere in the faq I can read some basics about the matching alogrithm?