How to read a match score

The number beside each result is a compatibility score, not a quality rating. It says how well an instrument's characteristics line up with the profile you built. It says nothing about whether the instrument is good, and it could not — CentDNA has no view on that and no data with which to form one.

Six weighted components produce it. Currency preference is 30%, stability preference 20%, liquidity preference 15%, risk compatibility 15%, geographic exposure 10% and your stated objectives 10%. Anything you marked as something you do not want then subtracts, in proportion to how badly the instrument offends rather than as a flat penalty.

The part worth understanding is the normalisation. Scores are scaled against the best result achievable for *your* profile, not against a theoretical perfect instrument. This is why a narrow profile produces a dramatic spread — pick one currency and one objective and the top result may sit at 96 while the fifth sits in the twenties — while a profile naming all seven currencies produces a much flatter set, perhaps 90 down to 69.

That flattening is the model telling you something rather than failing. If you ask for exposure to everything, more things genuinely fit, and the engine declines to invent distinctions that are not there. The corollary is practical: if your results look undifferentiated, remove a currency or two. A tighter profile is not a smaller question, it is a sharper one.

The ceiling moves as well. A profile that nothing in the catalogue serves well will top out in the eighties rather than showing a 99 for merely leading a weak field. When the top score is low, that is the honest statement, and it is more useful than a flattering one.

CentDNA provides recommendations, not investment advice. The score ranks fit against a profile you described. What to do about it is not a question this engine answers.