MyHeritage Releases cM Explainer! The new standard of genetic genealogy prediction! 

I can’t wait to have videos on usage! Once people fully understand this paradigm changing prediction it’s going to flip things in its head! Well done myheritage!

thats awesome! I will have another video soon that will help with this because its so different that all the tools we have used in the past!

Thank you Delila! There will be another video VERY soon and i think when people see it, they are going to be amazed at this thing! and best of all, its FREE! no login even required!

@@DNAFamilyTrees More to come? …even better! I Can’t wait!

With pedigree collapse it will rarely be the top % Mine with PC tend to be third in probability

awesome!

Thank you so much! their team was AMAZING!

That is awesome! Remember these percentages are LEGIT! meaning if there are strange generation patters, endogomay, pedigree collapse those are less likely and it will be lower in list. its the first tool to have a DNA database (of millions) to review to create percentages that are REAL population oriented, age oriented, what they have done is really amazing. I just completed an analysis of all the tools and this one DOES in fact take the top spot from the cmsolver

thank you!

Very! Their team of engineers, scientists, web developers, project leaders and team were among the brightest I have ever met. it was incredible collaborating and then seeing how they improved things over 5 months to make something like this that is so INCREDIBLY accurate! I am humbled to even have been a part!

I appreciate you saying so :) !

Thank you!

In case they dont read this ill pass that along

@@DNAFamilyTrees That would be appreciated. Thank you.

AWESOME question! Ok, this is paradigm changing. 95% vs. 5% does not mean its #1 at 95%, this is a TRUE probability! Things to consider, if there is pedigree collapse or AGE extremes it unlikely to be most likely answer! however, even in infidelity situations if the age range is normal and no pedigree collaps suspsected (and in this case sounds unlikely since you picked up new ethnicities) then it will more likely be the top answer. I will be publishing a detailed comparison of this very very soon! Think of the answers from cm explainer as a single tier of 100% (however many it gives in that tier) and its broken out that 100 into specific percentages for each kinship. Statistically the higher the more likely, however, being low percentage in the 100% teir does not exclude like being low in a tier based system. it simply a universal comparison if 100 people ahd the same exact scenario, the number that would be in each case.

The Explainer is only showing the possible places on the tree where the relationship might be, and the probabilites for those, but there isn't any guarantee that the highest probability is the correct answer, except where you get a single answer with a 100% probability. In this instance it gives you two possible answers, with a roughly 2/3 to 1/3 split, depending on the exact ages, and one of those two answers actually is the correct one, it just happens to be the one with the 1/3 probability, rather than the one with 2/3 probability, so I would say that it is working as well as it's possible for the mathematical model to work. If you found a 100 different people who had a match of 1853cM and an age gap of 15 years then something like 67 would get their correct answer at the top of the list of 2 options, and 33 would find their correct answer as the second of the two options, but all 100 would find that the correct answer was one of the two that the explainer offered up, although those numbers do assume that the correct answer is a single relationship only, rather than multiple different relationships via further back ancestors, which is an assumption the Explainer is built around.

The predictive answer that is top answer may not always be THE answer The other answers have numbers for a reason and that is that under the given criteria they are also the correct answer that % of the time Things that will skew answer to secondary options are… pedigree collapse ( for me those tend to be third in list) and unusual age variants ( in your case 15 years. If you see the blog for the release there is a sibling age range slope 15 years apart is on the outside with a huge probability that it’s less year’s difference That statistically in your case made the other “more likely” I have branches this is spot on I have other branches this my ranges are so unusual it’s always 2nd option

Glad you liked it

Thank you

Lance, Think of the set of final answers as being 100% one of these. (which is better than any other program) THEN from that list, instead of saying good on ya, good luck, it statistically breaks out the likelihood of each of the kinships (something on the the cmsolver did) when there are odd generation ages or pedigree collapse it will NEVER be the number 1 "most probable" answer as its not the most common scenario. (full kinships, half kinships then pedigree collapse and gen differences) example from a FB post. a person said their grandchild entered information, it got it WRONG! it said 98.8% likely to be AUNT and 1.2% to be grandmother! They railed that this tool was wrong! Here is the data 1880 cm, ages 20 and 55. Now.... two generations means to births in 35 year (55-20) which means average age 17.5 when she gave birth and 17.5 when her child did as well! That IS IN THE ONE PERCENT! for the first time ever we get a singular 100% one of these with an average of 3.4 kinships if both ages given and instead of tell ing you good luck its one of those 3.4 (which is better than even tier 1 in any other tool) its THEN breaks that 100% tier answer to give you the break out of that 100% per kinship! yeah its 1.2% for her, but thats not wrong! 1880 on shared cm, has 100% grandparent, Aunt/Uncle, Half Sibling, Niece/Nephew, Grandchild This woman was praising the accuracy of this LOL 100% one of these 5! Yet the cm Explainer said its 100% one of these TWO! and the likelihood of each is 98.8 / 1.2 aunt/grandmother. she is grandmother because giving birth at 17 and your child also at 17 is NOT the normal expectation! if 1000 people entered the same information 988 woudl be aunt, 11 others besides her would be grandparent lol by knowing person A was 20 and B 2was 55 it eliminated GRandchild (cant havea grandchild 35 yrs older than you! i cant give away all the internal logic involved but 3 of the 5 possible kinships for 1880 cm were eliminated. Others may see this as "wrong answer" for kinship or MRCA but they are not looking at the list as the 100% tier! and not realizing its a REAL quantifiable percentage of probability that it is a given kinship! if yours is "wrong" then i would 100% bet you have a generation anomaly or pedigree collapse (or the top two answers were a single digit separation) So which is in play? pedigree collapse or Age Variance? I know its one of those :) Knowing which it is, and expecting this can help pick as well! if 3.4 results per final answer its typically a single path, dual path and multi path answer! if i know its pedigree collapse involved im know to look at #3 FIRST! if i know there is an age variance in that line that is unusually high or low, #3 is first as neither will be THE MOST COMMON outcome. This is paradigm changing! It will take a while but once fully understood (those with cm solver background can get even more out of it!) this tool is going to skyrocket in usage! So im curious lance, which caused yours to vary, age of pedigree collapse?

I love this post! Ty! It is NOT always the #1 answer that’s a fact! In fact we’re pedigree collapse is involved it’s most often #3 answer for me! BUT that’s the beauty it nails it in list! Sure it was 7% vs 64% but that’s the correct ratio of pedigree collapse vs normal inheritance! And the ones about changing the answer from one to another I would guess are sue to unusual ages making the gaps unusually low or high comparatively This is a blind prediction taking all factors into consideration! Answers that are #2 or #3 mean unusual age or pedigree collapse involved! Those are huge clues for future! Can’t change it… well.. yep my pedigree collapse kin show #3 in list but that doesn’t mean wrong means it’s not the most likely aka unusual but I can’t change CNN news either on there website for anything I see is not entirely correct Expecting to change this would be like expecting to change wato or sharedcm or auto cluster or … other tools answers 😇

For the 99% aunt vs grandmother that has to be young ages for births (teens?) that had two gens pass instead of 1 I’d love to see the numbers on this one It took a while but found 1880 with a 35 yr range gave the percentage for aunt over grandparent 99/1 That’s 18 and 17 year for generation which is where the chart shown in blog begins to uptick from 0 to probable (1% !). That’s why it’s 1% prediction because with that cM its 99% of the time going to be the other answer! Other tools give it 100% as grandparent or aunt or half sibling with no prediction which is most likely #7 Half aunt as cousin has half aunt in list I’m sure if that just not too answer That’s from pedigree collapse ( unusual that close but possible ) or age abnormalities ( most likely ) This person is older of older while the half aunt is most likely young of the siblings of her parent thus the closeness skewing it towards same generation over a split generation Blind probability it would most likely be cousin more often than aunt

Thank you for this As for the FB admin and group I can probably guess which and who In all of these cases mentioned no other tool is closer to correct prediction as top answer! As no other tool has the guts to say this is THE most likely based on the simplicity of cM and ages! This is light years ahead of any other predictor! And I have yet to find one that the “correct answer wasn’t listed with percentages as possibles Once people gather understanding of pedigree collapse and of ages influence they will be able to spot the correct from list of not top answer ! If this was cM solver the ones listed are in RED multipath or I’m the yellow percentile area as less than 2% likely! These “wrong answers” are actually incredibly right once that is understood! Again GREAT feedback ty!

Interesting. 🤔 I was wondering about that and why that is. It’s as if your name pops up and the response is “nope. Not legit”

I concur with this .. One of my father's pedigree collapse double connection cousins has the following actual connections followed by the percentages the tool shows via mother's father - 2nd cousin's son - 22.9% via mother's mother - 2nd cousin's grandson - 1.4% It suggested 3rd cousin - 38.0% MRCA wise. It suggested great great grandparents being 54.9% likely a connection and great grandparents being 33.7% likely. The latter is correct via his mother's father. But via his mother's mother the connection is via his great great great grandparents which had a 4.6% probability.

I think the amazing part of this is lost in how the result is shown a bit The results are a set of 99.99% to be one of these few kinships Something no other prediction tool does stating it’s @one if these@ But rather than leaving you hanging it then gives you scientifically and mathematically predictions of the probability of each and states THE most probable (which for a majority will be correct) However the ones with lower percentages are also correct. A 7% third in list may well be your answer. This means you had a non normal age or pedigree collapse in that line and as a result is less likely than those without collapse or extreme ages ( which is the most common aka most probable) Once people realize that for the first times the percentages are in fact accurate rather than the SWAGs that every other prediction tool has this is going to skyrocket in usage! It’s a paradigm shift because we have become accustomed to 7 kinships per tier with three to four tiers And allowing those numbers to be completely meaningless for predicting! There will be resistance at first but the math holds up and eventually people will be on board despite people naysaying because they want answers My grandfather was born 1896 and his father 1860 That’s an obvious high age three in a row The most likely would never be that scenario. I don’t expect that But from 20 possibilities it turns to four and prioritizes them by probability. If I know I have extreme high (or low or pedigree collapse or both) I know yo skip #1 and look down This now I have even closer predictions and fewer noise to work through This is amazing!

@@DNAFamilyTrees Thanks for a detailed answer. I know my direct Darroch lineage back to before 1575 with certainty. Of course I am endeavouring to connect my Darroch DNA with the Darroch family in Stirling in the 1400s, which I believe I have done. Off to Scotland again on April 28th. 🙂

If I read this correct you have connected this person via your 13th great grandparent?

I have a 103 match (3c) and our mrca is my 4xGGP, another link being my 5xGGP. Our ages put us in the same generation, but trees show she's one generation below me. 🤔

@@PaulJHawkinsJr I dont think i follow the example, can you provide ages entered? I read a 103 cm match, check, 3c (actually that or prediction?) mrca is 4xggp (not 3c actually then?) another link at 5ggp (ok pedigree collaps) your ages put you in same generation, hmm.. maybe it does or doesnt because this tool has slope that the farther out you go are not what we are used to thinking of as generation, so it may be same or might be 1 or 2 different, so how did you determine same relative generation?) tree shows her one below you, i cant replicate that if i entered the 103 their are 5 kinships with a prediction a generation lower in tree, 3rd cousins child and the rest under great grandparents or closer which you didnt allude to, so im not sure i follow. If you could provide both ages, i think i can follow better and perhaps underestand the question to help answer it.

There can be exceptions to the rule. Age range in a generation could be as large as 20 years but the tool may not take that into account. Also, their blog explains that pedigree collapse and multiple connections can mess things up. In my case I have a match who shares 31cM match with someone and via our shared matches either I have several ancestral lines that are interconnected or he does (pedigree collapse-multiple connections). We have a possible connection via a possible secret father (half blood 8th cousin) and we have connections via the mothers ancestors currently unidentified. They also don’t take into account the individual segment sizes. The tool shows the connection as 3rd cousin. I’ve looked at most of my non complex matches and the suggested connection is correct or one generation out due to generational gap in the family. But when I look at most of my closer matches yet to be identified, the 3rd cousin suggestion is nowhere near found out to 5th cousin level yet. So, similar to ethnicities, it can give you an idea of how far back you need to go to find a clue, you just might have to go back an additional step or two in case of these potential multiple links. But, outside of my own inability to use the tool it has promise to be useful under normal familial links.

@@xrystal First thank you for the detail, i love the comment. === First ill speak to part you said "the age range in a generation could be as large as 20 and it doesn't take that into account". In fact it does and in fact it handles MUCH MORE than that! if you look at their release paper, look at parent graph. this is the age at which people are identified as becoming parents. you see the uptick begin around the 15/16ish age, but you see it slope up to around the 27/28 age then down to the 45ish and trail off about the 50ish age (hard to tell exactly but that is the slope appearance. This means that from 15 to 50 is the range of the include slope for ages. this also means the range of possibility is 35yrs. Now the beauty of this, that for the parent. if you look at the next graph below it, you see aunt/uncle and the 1c1r slopes, you see from 0-70 inclusive due to the compounding nature of the 35yr range at the parent level. This is the best generational predictor based on ages probably ever produced quantitatively! i find its application in the numbers to be beyond amazing, im awestruck at the overlapping and exponential included! === lots of juicy tidbits in that next part. first ill speak to the half 8th cousin. thats 20 generational distances. No company reports cm to fractional percentages less then a single 1cm (*and i wouldnt trust any that did as the test just doesnt acoomodate that level of observation). === Segment size do help but we have 3 entry fields, cm, my age and their age. BUT there is more behind scenes going on, i have to leave that right there. === ah the suggested connection, aka most probable. this is 100% the most misunderstood feature! it will be right MOST of the time (statistically lol) BUT we need to change paradigm to thinking that the resulting list after ages entered are a set of 100% answers, within that set, they are individually likely at the percentages listed. having it be the 1% rather than the 99% isnt an indicator of error, its stating your conditions are in the 1 percentile of likelihood. this is a result of extreeme age ranges (as you mentioned) or pedigree collapse or both. === i love the fact you stated that it has promise, because it means your open to learning how it is working and gaining the insights it can give. Ill give you an example you can enter. i have a 488 match, im 60 they are 75, this will refine a list down to just 4. Now if you have used the cmsolver or learned my pathing techniques, you will observe the listed answers in order are A) dual path aka full kinshipi, B) single path aka half kinshp, C) MULTI PATH aka pedigree collapes and D) half kinship aka a single path I knew at the start this was a pedigree collaps scenario, but i had no idea the effect on the cm and kinship when i was researching to predict a kinship. know i understand and the fact i can spot the multi path kinship from the list despite it being 7.1% likely overall comparatively i KNOW 100% its the ONLY multi path option and thus the only possible right answer, 2C! this tool is so complex but also so simple! it is the most innovative predictor ever created and when people fully grasp it, it is going to help so many people! thanks again for the observations! Good luck in your researching!

I’ll assess this in a video soon!

On simple level do not assume this isn’t occurring behind scenes for this with dna in myheritage! There are MANY factors in play! Users only need enter cM and ages and if myheritage dna member not even that Much of what you mention is included in the intellectual properties and can’t be discussed but well said

@@DNAFamilyTrees Thanks I hope you are right and that more is being done as I shared with them their recent hiatus from what I consider "helpful" work aside I think they need to leave the graphics alone and get back at some serious work. But to give them credit at least MyH pays some heed to their customer base and as a member of a few of their beta teams know that it's far more than just I they need to please eh. Anyway thanks for the reply.

I cant discuss future changes at all, if any (grin) BUT there will be another video soon explaining the cM Explainer! we have been so used to accepting SWAG that when we get precision of this caliber its actually a bit of a challenge to fully understand what we are seeing and why, but I have made preparation videos for the final draft that will explain this and once you see this i think you'll be able to get what you need every time inside the predictions given!

that video will be SOON to release, *withing the week)

I guess it's different horses for different courses, but personally I'd say that 'parent's 2nd cousin' is more quickly understandable than '2nd cousin once removed', and also more specific, as it narrows it down to one type of 2nd cousin once removed, instead of lumping the two types together.

@@MrDannyDetail I suppose it depends on what one is used to. For me, it changes the focus from the primary person to the primary person's parents--a little confusing. My oddly wired brain, perhaps.

@@debbeb4499 To keep the focus on the primary person I guess they could instead say '3rd cousin's parent', so essentially a 3rd cousin with a remove of negative one, but that would go against the usual convention of using the smaller cousinship which is what makes it a 2c1r.

This was a good thread. I am still getting used to the change to second cousins child from 2C1R as I’ve used that so long To a new person I think this is easier For vets it’s hard to change what we have used forever lol

@@DNAFamilyTrees The expression is a little bulky, don’t you think? I suspect people won’t be inclined to write it or type it that way. Abbreviations stand out-more absorbable without thought, so to speak. For example, it is editorially correct to type ages one through nine as words, and higher ages as numbers, but in the medical reports I type, it is correct to type all ages as numbers because it is an important fact in the document and is more noticeable that way. P2C for parent’s second cousin, 2CC 2nd cousins’ child, perhaps? No one would be likely to think it meant 2nd child’s cousin. 😊 I still like 2C1R, though (sigh).

227 as 3c would indicate extreme pedigree collapse I find this high of cM unlikely for that kinship ( nearly impossible) If I saw this in person I would suspect the person in question is a 3c as you said BUT there are SEVERAL paths to them with several shared ancestors

@@DNAFamilyTrees Thanks Larry, this DNA stuff is fun!

@@DNAFamilyTrees I (aged 62) do find this confusing (and fun!) as my 3rd cousin (Female aged 78, 227.9 cM) has an older brother (aged 82, 133cM). Can "extreme" pedigree collapse really be the explanation as neither they nor their parents came from or to the UK where our common ancestors had lived?

​@@dancub1 ID if extreme but id suspect two paths additional somewhere between yall. thats a very high cM (228) for that distance. id give it a 99% chance there are more paths between yall. depending on location, endogamy could be in play causing the pedigree collapse as many regions in europe and in usa so lines would be unusually strong with similar DNA to the point of common ancestor. 53cm would be a "normal" 3C cm value. so the 133 is double that and the 228 nearly double even that. so im pretty confident your gonna find more common ancestors and its quite possible those lines intermix (endogamy).

ok i think i understand, ok i entered that and yep i get predictive 2nd cousin and PAIR of great grandparents. So this means a married couple in common is most probable. Which pair is the question. wel you have 4 pairs. 2 on mother side, 2 on father side. So I would group/cluster the matches and see if this was maternal or paternal. You already know its maternal. SO this means you have 2 places to look, pair A (mothers mothers parents) B (mothers fathers parents). you should see shared DNA match with similarities to one or the other group. if so thats the group. Now upon review 3 of the kinships are via great grandparents (of the 7 so you examine 3 in the same spot!) 3 are via great grandparents (2nd most likely) so next try would be there. and one at grandparent (which if on maternal side should be easy to either spot or eliminate) If you suspect endogomay/pedigree collapse then its most likely 3C if not its most likely 2nd cousin (77.3% to be able to make that determination) hope this helps.

@@DNAFamilyTrees Cheers Larry, you’re advice is appreciated and I have identified my DNA match.

@@tonyweston5049 THATS AWESOME! congrats!

is it way off? OR is this actually predicting to a level that is unexpected? what i mean is, think of the result list as its 100% one of these. From that list, instead of sending you on your way the way other predictors do, this small result list is refined to give relative percentages. What this means then is what is the most probably might NOT be the answer for you, because for the first time EVER we are getting accurate precise numbers at the individual kinship level! other tools list 4,5,.. 7, 8,9 kinships in a grouped tier of say 45% likely with a total of 3 or 4 similar tiers. totalling 15-20 kinships for all tiers. Now yoru getting a list after all entered of only 3-5 kinships of which the answer is 100% one of them. rather than say its 100% of these 4 (or figuring out which of 20), it "sub groups" it by relative probability. If in that line the age of birth is left or right of normality or if there is pedigree collapse (or both) then the most probable is not the most likely for you. I have some that are 64% probably parents 1st cousin. it is actually 2nd cousin *which it shows as 7.1%* is it wrong? No, its 100% right! The scenario of ages and cm i gave it are only 7% likely to have the outcome i got! this one had SIX distinct common ancestors with FOUR most recent common ancestors AND the great grandparent in this line was born in 1860! (aka far right of the age slope!) The likelihood overall for pedigree collapse containing SIX paths compared to normal full or half kinships combined with the FAR extreme age gap AND the fact the person in the same generation is with all this 15yr different, that combination is only 7% likely OUT OF THE FOUR POSSIBLE kinships! Its not the most probable but thats not WRONG, its actually incredibly accurate to the real probability (rather than groups of kinships in a shared TIER of percentage of "probability". One person told me it was wrong their grandchild predicted most probable as AUNT 98.8% to 1.2% however, upon examination, she gave birth at 17 and her child gave birth at 17! if you look at the slope of ages at which mother begin having children (in the release paper) you see each of those is .006 and .006 probable so for that in two successive generation is .012 (or 1.2%) so predicting 98.8% that it would be an aunt is correct (even for her) despite the fact it was not her singular answer. That scenario is for 1000 people, 988 will be Aunt, 12 people of 1000 will be Grandparent. So stating its 1.2% isnt wrong, its incredibly precise! the 1880cm on EVERY OTHER TOOL shows FIVE kinships at 100% tier, this reduced it to only TWO! (think of it as 100% one of those two as the total percent adds to 100%!) the fact it s99% aunt and 1% grandparent is in fact, exactly the probability, they just happen to be the 1% because of the 17 and 17 birth ages! if you share the ages and cm that are "wrong" ill gladly review them for you :) I think if you look at it vs any other tool, you will start to see that the results are ALL 100% likely and individually the percentages given. its paradigm changing as we have gotten used to the other predictive models. this is incredibly accurate in ever case ive reviewed. In every "WRONG " answer i find that there are either AGE anomalies or cm anomaly due to pedigree collapse that make it be the not "normal" or most probable answer.

@@DNAFamilyTrees i think its off on account of to many cousin marriages,,, its putting alot of my DNA matches closer kin then they should be ,,(TENN,KY thing)

What’s not accurate