Using shared matches to break brick walls

Why shared matches do the heavy lifting

Most of us have a match list full of names we cannot place. There is the person who shares 47 cM and has no tree, the one with a tree made entirely of private profiles, and the dozens of faint matches hovering under 20 cM. Staring at them one at a time gets you nowhere. Shared matches change the shape of the problem.

A shared match is simply someone who matches both you and the person you are looking at. That single fact is quietly powerful: it usually means all three of you descend from the same ancestor, and it tells you roughly where in your family that ancestor sits. If you are stuck on a great-great-grandmother whose maiden name you cannot confirm, shared matches let you work sideways instead of backwards, sorting distant cousins into lines until a pattern appears.

Start with your closest match, not the most interesting one

Work down from the top of your list. Take the closest match you can actually identify, even if the connection is dull, and open their shared match list. Everyone there sits on the same side of your family as that person, with a few important exceptions we will come to shortly.

• Begin with a known second or third cousin, ideally someone sharing 20 to 100 cM.
• Note the names that appear in their shared match list, along with how much DNA you share with each.
• Repeat for the next closest identifiable match and compare the two lists.

Two matches who keep turning up in each other's lists are almost certainly from the same branch. That overlap is your first real clue, and it costs nothing but an afternoon.

Give every branch a colour

Colour coding is the most useful habit in genetic genealogy. Assign one colour to each of your four grandparent lines, then split into shades as you narrow things down. If your testing site offers automatic clustering, let it do the first pass and check its work afterwards. If it does not, a spreadsheet does the job perfectly well.

Whatever you use, keep the same columns: match name, shared cM, predicted relationship, shared matches noted, whether a tree is attached, colour group, and a one-line note on what you think the link is. Ten minutes of tidying each week saves hours of confusion later.

When a colour group grows to six or eight matches who all match one another, you have a genuine branch rather than a coincidence. Now look for the paper trail. Read the surnames that recur in the trees of that group, and note the places and dates attached to them. A cluster is only useful once it points at a parish, a townland, or a street.

Use the missing match as evidence

Negative evidence is badly underrated. If you have built a confident cluster on your grandmother's line and a new match sits entirely outside it, that tells you something valuable, even if their surname matches one you have been chasing for years. Similar names prove nothing on their own.

Try this test. Pick a match and list their shared matches. Then list the shared matches of the person you believe is their cousin. Any name that appears in one list but not the other deserves a second look. So does any name that ought to be there and is not, because the gap usually means the connection is further back than you assumed, or that a tree in the cluster is wrong.

Watch for the traps

• Endogamy and pedigree collapse. In communities where cousins married cousins across generations, shared cM figures are inflated and clusters behave unpredictably. Treat the numbers as a guide, never a verdict.
• Matches from both sides. If your parents came from the same town, a match may be related to you twice over. Their shared match list will look chaotic, and that is expected.
• Very distant matches. Ignore anything under roughly 8 cM unless the cluster around them is unusually tight. Small segments are often noise.
• Silent cousins. No reply to a message is not proof of no relationship. Plenty of people test once and never log in again.

Turn a cluster into a name

Once you have a branch, the goal is a specific couple. Build a floating tree for the cluster, assembling the matches side by side until two of those trees overlap in the same place and the same decade. That overlap is your target. Then go to the records for that place: parish registers, census returns, wills, and nonconformist registers, which are often overlooked.

Message the strongest matches in the cluster with something short and specific. Say which line you think you share, offer one detail, and ask one question. Vague enquiries rarely get answers.

Above all, write down your reasoning as you go, even in a notebook. Brick walls tend to break months after you last thought about them, and you will not remember why you dismissed a branch back in March. Shared matches will not hand you an answer, but they will tell you where to dig, and that is usually the hardest part.