Matches in Common: No correct unweighted shared DNA and no longest segment

[Riverman]

Today I encountered a false unweighted shared DNA being shown if I look at matches in common. Like if I go to
Unweighted shared DNA
directly from one kit, its 10 cM more than when I look at the same value for this match from another kit and just doing it from the matches in common. The matches in common click on the shared DNA is therefore wrong, it displays the same value, in this case, for weighted and unweighted, which is incorrect. Also, there is no longest segment shown.

Is this on purpose and was it reported? Why showing unweighted DNA and longest segment if its either wrong or there is no value at all?

Was there a debate on the issue anywhere?

[geebee-1015]

I'm not entirely sure what you mean.  When I look at Matches in Common, they don't even show either unweighted sharing or longest shared segment segment -- just Timber's adjusted amount.  That will be the same as the unweighted sharing only in cases in which no adjustment was made, which in my experience is not the majority of matches.

For example, if I click on Shared Matches for the first of my below 90.0 cM matches -- "DB" -- I first see how much he shares with my closest relatives, who include my "alternate self", my daughter, and three of my nieces.  

As it happens, DB is not a below 90.0 cM match for my "alternate self", but I still only see one number, 101 cM.  It's just that I know that there would be no Timber adjustment with this much sharing.  But, for the account assigned to me directly (not as "manager") I reportedly share 85 cM.  This represents a reduction of 3 cM from the unweighted sharing of 88 cM.  My daughter's sharing with DB is shown as 24 cM, with no number of segments shared or unweighted sharing or longest segment shared.  However, if I switch a direct comparison between my daughter and DB, I can see that this represents a reduction of 13 cM from the unweighted sharing of 37 cM.  It's in two segments, and the longest shared segment is 25 cM.

My niece GL reportedly shares 90 cM, which is likely to be at least 90.0 cM since otherwise it would probably have been adjusted downward -- simply because Timber wouldn't be constrained against doing so.  My niece SS shares 96 cM, so there definitely should be no adjustment there.  And my niece SJ shares only 25 cM -- which is very likely the result of a downward adjustment by Timber.  But of course I can't tell, since it's less than 90.0 cM.

Now, if what you're saying is that like me you have been tested more than once and show different amounts of sharing -- that's true for me, too, at least much of the time.  This is likely in part a function of having somewhat different phasing for the multiple kits.  For example, as myself I share 3,448 cM with my daughter; in my managed account also based on my DNA from a previous test, I share 3,468 cM.  That's a 20 cM difference.

That difference is even greater with another relative -- RB, who is a 2nd cousin once removed -- I share 100 cM in 10 segments, with a longest shared segment of 37 cM.  As "ID" -- who as I've said, is also me -- I share 145 cM in 9 segments, with a longest shared segment of 41 cM.  That's absurd, and clearly "SideView" has some incorrect phasing going on here for at least one of these "selves".  Likely the one that shows less matching, since incorrect phasing isn't likely to increase the amount of matching by such a huge amount -- but might reduce it by that much.

(This is where a chromosome browser would really be useful.)

[Riverman]

Well, if you have the shared matches list and click on the blue link for the shared DNA. You get a value line for "Unweighted shared", but this value is just wrong, its still just the weighted shared value. And it also misses the longest segment, even though this line exists.

Like:
Shared DNA: 20 cM across 2 segments 
Unweighted shared DNA: 20 cM (wrong!)
Longest segment cM (missing!)

[geebee-1015]

Well, if you have the shared matches list and click on the blue link for the shared DNA. You get a value line for "Unweighted shared", but this value is just wrong, its still just the weighted shared value. And it also misses the longest segment, even though this line exists.

Like:
Shared DNA: 20 cM across 2 segments 
Unweighted shared DNA: 20 cM (wrong!)
Longest segment cM (missing!)

Yeah, I can see that.  I think this is a change, at least for me.  What I'm remembering -- though I could be wrong -- is that previously when I clicked on the link I didn't see anything but the final reported amount, plus additional information about possible relationships (with probabilities).  

So from my standpoint, Ancestry has just made things even worse.  Now they're actively misstating the truth in many cases.

For example, "EC" is my 1st cousin twice removed -- the granddaughter of one of my 1st cousins.  I share 150 cM in 7 segments with EC, with a longest shared segment of 56 cM.  And in my shared match list with EC, she and my daughter are reported to share 34 cM.  But now if I click on the link, Ancestry is claiming that this is also the unweighted sharing amount -- which is bovine excrement.

The only reason I know this, however, is because my daughter's data is in my account and I can do a direct comparison between the two of them. When I do that, I find their unweighted sharing is actually 40 cM -- so there's a 6 cM adjustment by Timber that Ancestry is no longer simply not mentioning in my shared match, but pretending does not exist.

It's even worse in the case of "BH", another 1st cousin twice removed.  Like EC, BH is the granddaughter of one of my 1st cousins.  In fact, the granddaughter of the same 1st cousin as EC is, and the two are 1st cousins to each other.  Here, too, my daughter is in the shared match.  The reported sharing is 46 cM, which after clicking the link is reported as "Shared DNA: 46 cM across 4 segments" and "Unweighted sharing: 46 cM."  In this case, the direct comparison doesn't just show a 6 cM adjustment by Timber, but a 35 cM adjustment!  It also shows that that the longest shared segment alone is 50 cM.

I hope this is unintentional.  Best case scenario is that they're getting ready to actually report unweighted sharing and longest shared segment length but have just jumped the gun a little.  This is definitely something Ancestry needs to fix, because as it stands now they're actively lying -- whether they mean to or not.  Unfortunately, I've never had much success in dealing with Ancestry's "Customer Service".  

(Anybody else want to give it a go?)

[Riverman]

This is definitely an error from their side. Because either they don't want people to see that, in which case they should simply delete the two bottom lines, or they wanted it to work, but it just didn't.
In any case it is a drastically misleading information in some instances, since you know very well that the difference between weighted vs. unweighted can be huge. And I compared the results for endogamous vs. non-endogmous people and results. The final verdict is that the whole Timber algorithm almost never works out fine, up to relevant samples being deleted.

[geebee-1015]

This is definitely an error from their side. Because either they don't want people to see that, in which case they should simply delete the two bottom lines, or they wanted it to work, but it just didn't.
In any case it is a drastically misleading information in some instances, since you know very well that the difference between weighted vs. unweighted can be huge. And I compared the results for endogamous vs. non-endogmous people and results. The final verdict is that the whole Timber algorithm almost never works out fine, up to relevant samples being deleted.

Well, as I said I'm hopeful that they are planning to add the info for the bottom two lines and just got ahead of themselves.  We'll see.

At least they did add the ability to sort "Shared Matches" by their relationship to your match, not just their relationship to you.  This is really useful, because all that sorting by relationship to me does is list my alter self 1st and my daughter 2nd -- that is, when she also shares the match.  What's funny is when I'm looking at my daughter's shared matches and don't find either her mother or me in the list.  That shouldn't happen, but it does.  (Usually because Timber has reduced sharing for the connecting parent to less than 8.0 cM and then simply dropped it.)

What I've found seemed to undermine Ancestry's use of Timber to combat "excess matching".  Typically, the way it turns out is that there simply are a lot of very close matches on one side or the other, which affect the odds of sharing the same segment or segments with more distant matches.  Ancestry claims that you're unlikely to share the same segment with multiple 4th cousins, which may seem reasonable; but often, it turns out that the multiple 4th cousins aren't 4th cousins to each other, but sometimes siblings.  Two full siblings may have a greater that 50% chance of sharing the same segment or segments with you, even if they are 4th cousins to you.

And, of course, if someone happens to share more than 90.0 cM with me -- and is therefore Timber is restricted from "adjusting" the sharing -- it really shouldn't make an adjustment for my daughter, either, if all of the shared DNA came from me.  It wouldn't suddenly become "excess" between the two of us!  The same thing is true on the other side.  If none of my sharing with a match should be considered excess, why should any of my sharing with the match's child be excess if it all came from the connecting parent?