Shared Matches (aka In-Common With Matches or ICW) is a term applied in autosomal match analysis when doing next tier analysis after basic match lists. Shared Matches (or ICW) between two testers are simply matches that appear on both testers autosomal match lists. That is, simply the intersection set of two testers match lists. The result is a Shared Match List which is a subset of each testers match list.
Appearing on a shared match list does not imply that all the shared matches are related to each other. That is what the process of segment triangulation does more accurately. A shared match list is simply a hint as to a group of testers that may ALL share a common ancestor that the two "base" testers also share. Verified segment triangulation does guarantee there is a common ancestor. (More confusing is the fact that just because a shared match does not segment triangulate does not mean they do not have a common ancestor; they still could. Because for more distant matches, the chance of segment triangulation is small even if their is a common ancestor.) Confused? Let's give some examples.
If you have three full-siblings, then when taking any two of them, the third sibling will always appear on the shared match list made from the other two full-siblings match lists. If you have the parents tested as well, both parents will always appear on the shared match list created from any two full-siblings. When creating a shared match list of the match lists of the parents of full-siblings, the children will always appear (and any of their children's descendants that tested). BUT, a parent will generally NEVER appear on the other parents Matches as they are generally not related. Pretty much always, for second cousins and closer, if you are truly related, then you will always appear on the shared match list created from any two other relatives who are second cousin or closer and share the same ancestor.
Just because person C appears on a shared match list made from person A and person B's match lists, does not mean that person C is related to both A and B through the same ancestor. And that is the crux or limitation here. Only segment triangulation can verify that there is a common ancestor — that all three are related through the same ancestor. So as another example, C could be related to A through C's father and C related to B through C's mother. Assuming C's parents are not related, C will appear on a shared match list created from A and B's match lists. But A and B are not related. Specifically, think of a 1st cousins' cousin who you are not related to because they are a cousin due to the parent they have that is not your ancestor. Lets say you are person A and related to the cousin C's mother (your aunt). The cousin C's cousin is person B and related to C through C's father (your Aunt's husband who you are not related too). The common cousin C that both A and B have would be on each other's Matches and thus appear on the shared match list created from A and B's match lists. But A and B are not related. Hence why shared match lists are hints at their being a common ancestor but not a confirmation; especially if A and B (used to create the shared match list) are not related in the first place.
It is very important to understand, when doing genetic genealogy matching, specifically autosomal match analysis, the difference of simply matching testers and matching segments. Often, tools and companies try to simplify the analysis and only present the matching of testers via match lists and maybe shared match lists. Without delving into the underlying matching segments actually used to create these upper-level summaries though, they may over state a possible relationship. Tools may present summary information of the total matching size (summation of all matching segments), the largest matching segment, and the number of matching segments to help understand the match further. A Shared match list is one of these higher level summary concepts based simply on the match lists each tester has. To really do detailed analysis and develop proof-level work, one often needs to delve into segment matching which underlies all autosomal match analysis.
GEDMatch terms shared matches as "People who match both, or 1 of 2 kits". Some others use the term In-Common With (or ICW) instead of shared matches. They are synonymous.
Not In Common With (or NICW) is often as important to check for as ICW. See the "or" above in the GEDMatch description. That is an exclusive-or. Meaning, a kit will either appear in the (a)ICW (shared match list; that is, be on "both" match lists), or (b) the match kit will be in the NICW list as it appears in only one of the two kits match lists. NICW lists can help greatly when only one parent is tested. The NICW list of a child when the child is compared to one parent will be the people related to the child because of the other parent. A NICW list between two full-siblings will likely be more distant cousins (3rd or further) where only one of the two siblings got the DNA shared with the more distant cousin. Hence, simply "merging" full-siblings match lists is often useful before doing deep autosomal match analysis. Support for providing an NICW Matches is spotty among the analysis sites and tools but very useful.
Much can be gleaned by manipulating match lists and shared match lists; doing intersections, negations and such. But realize, these are often hints only. A search must still be done for a likely and true common ancestor.
Some mention the shared matches term in regards to yDNA analysis but it is not really applicable.
Clustering is often presented in a graphical form and thus letting you "visualize" all the Shared Match groups that may exist with all your Matches matches. Clustering is a gaining-popularity tool that is similar in concept to segment triangulation but applied to Shared Matches. The caveat being that segment triangulation assures all members of a triangulation group are matches on that segment with every other member and thus must share a common ancestor. Clustering simply looks for groupings where many of the matches are on each others shared match list; but not necessarily all. And not necessarily the match is due to the same segment. If you develop a segment triangulation of matches, then they truly are all related to each other through a common ancestor. Clusters indicate a strong likely shared common ancestor "path" (some may have closer shared ancestors than others) but does not guarantee it. The term triangulation is never applied to the shared match list: the process or result. In fact no term has been used to define a fully matched cluster. A fully matched cluster exists only when you take all members, two-at-a-time, and confirm all other members of the group are on the resultant shared match list. In cluster charts, this would represent a solid, fully-filled cluster square with all the smaller squares within it solidly filled. Such strict clusters are rare when including matches at approximately 150 cM and below in total match strength.
A higher-level clustering process has been recently refined and regularized, now called the Leeds Method by most, to map more distant matches to likely (great-)grandparents (not segments, simply matches). it is simply a quick and simple form of clustering that tends to work and generate only 4 strong clusters. The clusters then, likely when looking at how cousins match into them, are likely representing each of the original testers grandparents as the path to the ancestor in common with all the other matches in the cluster. Key to using this method is to remove all 1st cousins, 1st cousins once removed, and similar who share more than one of the original testers grandparents. Starting in 2019, many tools exist and are being added to genetic genealogy test company websites to do cluster analysis of your 30 to 300 cM total match strength matches on your shared match list.
Appearing on a shared match list does not imply that all the shared matches are related to each other. That is what the process of segment triangulation does more accurately. A shared match list is simply a hint as to a group of testers that may ALL share a common ancestor that the two "base" testers also share. Verified segment triangulation does guarantee there is a common ancestor. (More confusing is the fact that just because a shared match does not segment triangulate does not mean they do not have a common ancestor; they still could. Because for more distant matches, the chance of segment triangulation is small even if their is a common ancestor.) Confused? Let's give some examples.
If you have three full-siblings, then when taking any two of them, the third sibling will always appear on the shared match list made from the other two full-siblings match lists. If you have the parents tested as well, both parents will always appear on the shared match list created from any two full-siblings. When creating a shared match list of the match lists of the parents of full-siblings, the children will always appear (and any of their children's descendants that tested). BUT, a parent will generally NEVER appear on the other parents Matches as they are generally not related. Pretty much always, for second cousins and closer, if you are truly related, then you will always appear on the shared match list created from any two other relatives who are second cousin or closer and share the same ancestor.
Just because person C appears on a shared match list made from person A and person B's match lists, does not mean that person C is related to both A and B through the same ancestor. And that is the crux or limitation here. Only segment triangulation can verify that there is a common ancestor — that all three are related through the same ancestor. So as another example, C could be related to A through C's father and C related to B through C's mother. Assuming C's parents are not related, C will appear on a shared match list created from A and B's match lists. But A and B are not related. Specifically, think of a 1st cousins' cousin who you are not related to because they are a cousin due to the parent they have that is not your ancestor. Lets say you are person A and related to the cousin C's mother (your aunt). The cousin C's cousin is person B and related to C through C's father (your Aunt's husband who you are not related too). The common cousin C that both A and B have would be on each other's Matches and thus appear on the shared match list created from A and B's match lists. But A and B are not related. Hence why shared match lists are hints at their being a common ancestor but not a confirmation; especially if A and B (used to create the shared match list) are not related in the first place.
It is very important to understand, when doing genetic genealogy matching, specifically autosomal match analysis, the difference of simply matching testers and matching segments. Often, tools and companies try to simplify the analysis and only present the matching of testers via match lists and maybe shared match lists. Without delving into the underlying matching segments actually used to create these upper-level summaries though, they may over state a possible relationship. Tools may present summary information of the total matching size (summation of all matching segments), the largest matching segment, and the number of matching segments to help understand the match further. A Shared match list is one of these higher level summary concepts based simply on the match lists each tester has. To really do detailed analysis and develop proof-level work, one often needs to delve into segment matching which underlies all autosomal match analysis.
GEDMatch terms shared matches as "People who match both, or 1 of 2 kits". Some others use the term In-Common With (or ICW) instead of shared matches. They are synonymous.
Not In Common With (or NICW) is often as important to check for as ICW. See the "or" above in the GEDMatch description. That is an exclusive-or. Meaning, a kit will either appear in the (a)ICW (shared match list; that is, be on "both" match lists), or (b) the match kit will be in the NICW list as it appears in only one of the two kits match lists. NICW lists can help greatly when only one parent is tested. The NICW list of a child when the child is compared to one parent will be the people related to the child because of the other parent. A NICW list between two full-siblings will likely be more distant cousins (3rd or further) where only one of the two siblings got the DNA shared with the more distant cousin. Hence, simply "merging" full-siblings match lists is often useful before doing deep autosomal match analysis. Support for providing an NICW Matches is spotty among the analysis sites and tools but very useful.
Much can be gleaned by manipulating match lists and shared match lists; doing intersections, negations and such. But realize, these are often hints only. A search must still be done for a likely and true common ancestor.
Some mention the shared matches term in regards to yDNA analysis but it is not really applicable.
Clustering is often presented in a graphical form and thus letting you "visualize" all the Shared Match groups that may exist with all your Matches matches. Clustering is a gaining-popularity tool that is similar in concept to segment triangulation but applied to Shared Matches. The caveat being that segment triangulation assures all members of a triangulation group are matches on that segment with every other member and thus must share a common ancestor. Clustering simply looks for groupings where many of the matches are on each others shared match list; but not necessarily all. And not necessarily the match is due to the same segment. If you develop a segment triangulation of matches, then they truly are all related to each other through a common ancestor. Clusters indicate a strong likely shared common ancestor "path" (some may have closer shared ancestors than others) but does not guarantee it. The term triangulation is never applied to the shared match list: the process or result. In fact no term has been used to define a fully matched cluster. A fully matched cluster exists only when you take all members, two-at-a-time, and confirm all other members of the group are on the resultant shared match list. In cluster charts, this would represent a solid, fully-filled cluster square with all the smaller squares within it solidly filled. Such strict clusters are rare when including matches at approximately 150 cM and below in total match strength.
A higher-level clustering process has been recently refined and regularized, now called the Leeds Method by most, to map more distant matches to likely (great-)grandparents (not segments, simply matches). it is simply a quick and simple form of clustering that tends to work and generate only 4 strong clusters. The clusters then, likely when looking at how cousins match into them, are likely representing each of the original testers grandparents as the path to the ancestor in common with all the other matches in the cluster. Key to using this method is to remove all 1st cousins, 1st cousins once removed, and similar who share more than one of the original testers grandparents. Starting in 2019, many tools exist and are being added to genetic genealogy test company websites to do cluster analysis of your 30 to 300 cM total match strength matches on your shared match list.