I played an altered dominant - the tense, bluesy chord that ends a hundred jazz tunes - and tapped it into a chord finder.
"No chord found."
There's a chord. It has a name. I was holding it. The tool just couldn't see it, because the voicing does the most ordinary thing an altered dominant does: it bends a couple of notes and drops the fifth.
Every chord finder looks good on an open C. The real question is what happens the moment you leave a note out or bend one - the way actual players actually voice chords. That's where the field comes apart, and that's what this post is about.
TL;DR A chord finder that aces beginner shapes can still go silent - or quietly wrong - on the chords you reach the moment you play past open-position triads: altered dominants, slash chords with real extensions, Lydian colours, omitted-tone voicings. This post is the deliberate zoom-in on that hard tail. It's a companion to the fair, scored test - the chords that surfaced while I stress-tested my own lookup, with theory as the referee and every shape published so you can re-run it. Not "every other tool is wrong." Three honest degrees of breaking, the gentlest of which is just less explicit than Fretscape, not wrong.
First, the honest framing
These are hard cases, and the honest thing is to tell you where they came from. I didn't assemble a reel of competitor failures to win an argument. They came out of building the thing.
Before Fretscape's chord identifier shipped, I stress-tested it against something like seven hundred voicings - deliberately awkward ones, the kind chosen to break a lookup - checking every result, changing the engine, checking again, until the hard ones came out right. Cross-referencing other tools was just part of that: when my answer and theirs disagreed, I wanted to know who was right, and why. The examples below surfaced in that process. I noticed the field stumbling on them while I was busy trying to stop my own tool from stumbling too.
That's also why this is the supplement, not the proof. In the main post I ran the balanced version: a set of 47 voicings I didn't choose, scored by two independent blind judges who didn't know which tool was mine. This isn't that. This is the zoom-in on the chords that separate the field - hard on purpose, because hard is where the separating happens.
But hard doesn't mean obscure. Not one of these is a freak. They're altered dominants, slash chords, Lydian major-sevenths - bread-and-butter the moment you play anything past a campfire chord. That's the actual point: the tool you use is probably fine for beginner shapes and silently wrong exactly where playing gets interesting.
So I'll do three things on every example. State the notes. State what theory says it is. Then let you check. Don't trust me - trust theory, which has been refereeing this since long before any of us shipped a chord finder. And I won't name a competitor anywhere. "One tool," "another," "a couple." The behaviour is the point, not the brand.
Seven hundred chords is a lot of staring at note names. These are the handful where the gap was clearest - I stopped noting them down somewhere around fifty. Not because I ran out. Because I had enough.
The bridge: the fair test already cracked here
I don't have to assert that the field breaks on the hard tail. It already did, in public, under blind conditions.
On the main test's hardest cases, the two judges - scoring anonymised tools they couldn't identify - documented exactly this fracture, in their own words:
- one tool refused around eleven textbook-nameable voicings, returning "no chord found" on ordinary omitted-tone dominants;
- one buried answers in a 26-to-62-name unranked wall, leading with a rootless reading over the obvious chord;
- one inferred notes that weren't played to complete a fragment;
- even the strong runners-up reached for contorted enharmonic spellings on the hard ones;
- and one tool - revealed only afterwards as Fretscape - scored a perfect 94 out of 94, from both judges independently.
That's two referees who'd never heard of Fretscape, finding the gap blind. The full transcripts are linked at the bottom. And that was the tame set.
Here's the hard tail.
Degree one: "no chord found" - when a tool just gives up
The most damning failure. A legitimate guitar voicing goes in, and nothing comes back. Each of these three drew an outright "no chord found" from at least one tool - and they're standard altered and extended shapes, not puzzles.
0 7 6 7 8 8 is E7#5#9 - E, G#, D, G, C. A classic altered dominant: the dominant shell, plus both altered colours, #5 and #9. The honest name is short and useful. One tool returned no chord at all. Another called it an augmented chord and quietly understated the dominant seventh. A third produced a contorted minor-chord spelling that treats the raised 9th as if the whole chord were minor - reading the tension as a mistake rather than the point.
x 3 2 3 2 4 is C7#5b9 - C, E, B♭, D♭ and an altered fifth. Root, third, flat-seventh, flat-ninth: every ingredient of an altered dominant, nothing ambiguous about the family. One tool again returned no chord. Another gave an augmented spelling that dropped the dominant seventh - the one note that makes it a dominant in the first place.
3 x 4 4 2 x is Gmaj7#11 - G, F#, B, C#, with no natural fifth played (strictly, Gmaj7#11(omit5)). A bright Lydian major-seventh sound. One tool returned no chord. Another spelled the C# as a flat-fifth - Gmaj7b5 - which is pitch-equivalent, but reads a deliberate Lydian colour as a damaged major chord. Same notes, completely different story about what you're playing.
A real chord went in, and the tool returned nothing. That's not restraint. That's a blind spot where the music gets interesting.
Degree two: a related name, but not the useful one
Subtler, and more common. These tools don't fail - they answer a slightly different question. They re-root to the bass, or promote an upper-structure reading over the chord you're actually holding. Not wrong. Just the wrong headline.
2 x 0 2 3 0 is Dadd9/F# - a D add-nine chord with F# in the bass, the open high E ringing as a genuine added ninth. Several tools re-rooted it to an F# minor chord - things like F#m7♭13 or F#m7#5 - purely because F# is the lowest note. They built a complicated F# chord rather than hearing a simple D chord over an F# bass. This is the clearest slash-with-real-extension case in the set: the bass note is not always the root, and a good tool knows the difference.
x 3 2 2 1 2 is Am6/C - a complete A minor-six chord, inverted so the third sits in the bass. The same notes can be respelled in stranger ways - C6#11(no5), F#m7♭5/C, a rootless D9 - and a couple of tools led with exactly those. None of them is an impossible match. But none is the recognisable object a guitarist is actually holding. In several of these the clean Am6/C was sitting right there in the list - just not at the top, which is its own kind of failure.
x 3 3 3 3 5 is C9sus4(13) - a functional C dominant-sus voicing. A couple of tools promoted an upper-structure slash, B♭maj9/C, to the headline. It's pitch-related and it's not nonsense - but as the primary name it's close to useless, because it hides the dominant-sus function that's the whole reason you'd play the chord.
The hardest part of naming a chord was never finding a label that fits. Dozens fit. It's choosing the one you'd actually write on a chart.
Degree three: acceptable shorthand - we're just more explicit
This is the bucket that earns the other two. Here the other tools are not wrong. They use valid, common shorthand. Fretscape's only edge is precision - and I'm going to say that plainly, because pretending these are wins would poison the cases where the gap is real.
x 5 3 5 6 3 is Dm7add11 (strictly Dm7add11(no5)). Most tools call this Dm11, and that's perfectly reasonable shorthand. The only quibble: a full Dm11 implies the ninth, E, which isn't in the voicing. Dm7add11 says exactly what's there and nothing that isn't. And to be honest about it - G7sus4/D is also a completely valid reading of the same notes; Fretscape returns that one too, just second. This isn't an "everyone else is wrong" case. It's a "we're a notch more literal" case.
The altered dominants from Degree one belong here too, in part. The same E7#5#9 and C7#5b9 also drew valid jazz alternates from some tools - E7(#9,♭13), C7(♭9,♭13). Those aren't wrong; the C in that voicing can honestly be heard as a ♭13 rather than a #5. The point of Degree one was never that the alternate spelling is wrong. It's that on the very same chord, some tools found a clean name and others found nothing at all. That spread, on one voicing, is the whole story.
And the sus voicing: C13sus4 or Csus13 instead of the more explicit C9sus4(13). Valid shorthand against a fuller content label. A taste difference, not a correctness one.
I'm not going to tell you the other tools are wrong. On the easy chords they're fine, and even on these, some land a perfectly valid name. The difference is what they reach for first - and whether they reach for anything at all.
The pattern underneath
Strip the three degrees back and they're degrees of the same single thing: a good identifier's one real job is choosing the most useful name first.
Degree three does it, just less precisely. Degree two leads with a less useful relative. Degree one refuses to do it at all. And under the worst of them sits the exact fault from the main post - a tool that either fakes knowledge (invents a note, walls you with maybes) or refuses knowledge it should have ("no chord found"). A good identifier lives in the honest middle: it names what's defensible, says what's missing, and puts the useful answer on top.
It's the same call Fretscape keeps making everywhere else - the honest answer over the impressive-looking one.
The hard part was never finding a name that fits. It's choosing the one you'd actually write on a chart - and being willing to say "this, with the fifth left out" instead of either a wall or a shrug.
So go break your own tool
The takeaway here isn't "Fretscape wins." It's that you now know what failure actually looks like - not on an open C, where everything works, but on the chords you reach the moment you play something interesting.
So go test your own. Take the tensest, sparsest, most awkward voicing you know - the altered dominant, the slash chord with a real extension, the shape with the fifth left out - and tap it into whatever you use. Does it name it? Does it lead with something you'd recognise? Or does it give up? Then try the same shape in Fretscape's chord identifier and compare.
One honest caveat, the same as the main post: because I've published the shapes, a determined reader could re-run them and work out which anonymous tool was which. Keeping them unnamed is my editorial choice, not a guarantee - so I'd rather say so than pretend otherwise.
Technically correct was never the hard part. Useful is.
A few questions this raises
Aren't you just cherry-picking chords your tool happens to win?
Not to win - to build. These chords came out of stress-testing my own lookup against around seven hundred voicings before it shipped, hardening it until the hard ones came out right; I cross-referenced other tools as I went. So yes, they're selected for being hard - but for hardening my engine, not for flattering it. The representative proof is still the main post: a set I didn't pick, graded by judges who didn't know which tool was mine. The shapes here are all published, the theory is checkable on a fretboard, and the honest Degree-three cases are in precisely so the harder ones are believable.
Is "no chord found" ever the right answer?
Sometimes - when a voicing genuinely has no defensible name, an honest blank beats a confabulated label. But none of the Degree-one chords are that. They're standard altered and extended voicings with clear, compact names. Returning nothing there isn't restraint; it's a gap in what the tool can model, usually around an omitted fifth or an enharmonic spelling it couldn't reconcile.
Why does the bass note matter so much?
Because on guitar the lowest note does most of the work of telling your ear what it's hearing. Dadd9/F# and "some F# minor chord" contain the same pitches - but one hears a D chord over an F# bass, the way you're actually playing it, and the other invents a complicated F# chord just because F# sits lowest. A good tool defaults to the bass for rooting and lets you override it.
Where's the scored proof these tools really differ?
In the companion post. Seven tools, 47 cases, two independent blind judges scoring against theory, full transcripts published. Read that one first if you want the evidence; this one is the teaching tour of the edges.
Appendix: the shapes, so you can re-run
Every voicing above, low string to high - paste them into any tool and check against theory yourself.
| Shape | What it is |
|---|---|
0 7 6 7 8 8 | E7#5#9 (altered dominant) |
x 3 2 3 2 4 | C7#5b9 (altered dominant) |
3 x 4 4 2 x | Gmaj7#11 — Gmaj7#11(omit5) (Lydian colour, no 5th) |
2 x 0 2 3 0 | Dadd9/F# (slash with a real added 9th) |
x 3 2 2 1 2 | Am6/C (minor-6 inversion) |
x 3 3 3 3 5 | C9sus4(13) (functional dominant-sus) |
x 5 3 5 6 3 | Dm7add11 — Dm7add11(no5) (precise vs the broad Dm11) |
And the two blind-judge transcripts from the fair test referenced above, shared exactly as they happened:
