How it reads your numbers
No secret sauce. Four ratios, four bands, and a rule for which one wins when more than one is out. Here is the whole thing so you can disagree with it.
The ratios
| Ratio | Healthy | Below the band | Above it |
|---|---|---|---|
| Overhead squat ÷ snatch | 1.05–1.35 | You can snatch nearly everything you can hold overhead, which means the bar is landing on a position with no margin in it. Overhead strength and stability is the ceiling here. | Plenty of room overhead — holding the bar is not what is stopping you. |
| Power snatch ÷ snatch | 0.78–0.88 | Your power snatch is a long way behind your squat snatch, which points at how fast you get from the finish of the pull to a locked-out bar. | Your power snatch is nearly your squat snatch, so the squat is buying you almost nothing. You are catching high and standing up rather than receiving in the bottom. |
| Snatch ÷ back squat | 0.58–0.68 | You squat far more than you snatch. The legs are not the problem — the lift is leaking somewhere between the floor and the lockout. | You snatch a lot relative to what you squat. The technique is carrying you; more leg strength is the cheapest kilos available. |
| Snatch ÷ clean & jerk | 0.76–0.84 | Your clean & jerk is well ahead of your snatch, which makes this snatch-specific rather than a general strength gap. | Your snatch is unusually close to your clean & jerk — if anything the jerk is the lift with room in it. |
Coaching a squad? Read the whole team sheet at once — every lifter through these four ratios, grouped by the session each needs, with a link for each athlete to their own read.
Which one wins
Each ratio gets a severity — how far outside its band it sits, measured in band-widths, so a narrow band and a wide one are compared fairly. The overhead ratio then carries double weight, because it is causal as well as limiting: a lifter who cannot hold the bar overhead also catches high, and fixing the catch without fixing the overhead fixes nothing.
It is a multiplier and not a veto, on purpose. An overhead ratio that is barely outside its band does not get to hijack a dramatic finding somewhere else.
Where this stops being reliable
These bands describe a technically sound lifter. They are conventional weightlifting benchmarks, not laws, and plenty of people sit outside one and are completely fine — so every reading here is phrased as the thing worth looking at next, never as a diagnosis.
They are also only as good as the numbers you feed them. A snatch max from two years ago and a fresh back squat will produce a confident, wrong answer. Ratios computed from a max you have never actually tested are guesses wearing a decimal point.
Bodyweight is in none of them, on purpose. How a snatch compares with lifters of other sizes is a fair question — the Sinclair calculator answers it — but it says nothing about what is capping the lift, which is what the ratios between your own lifts are for.
The same ratios run the other way are the lift ratio calculator: one lift you know, and where each of the others would sit if every ratio were in the middle of its band. That is a place to load a drill from, never a max — the planner only ever reads lifts you have tested.
Meet day is a different question again: not what is capping the snatch, but what to open with and when to warm up. The meet plannerworks that out from your best, with the same warm-up rungs as the rest of the site.
And they cannot see you lift. Where your feet land and whether the bar leaves you at the hip are not in any ratio — that is what the video read is for, and it is described below. Neither is the same as a coach watching you.
What a lift you have not tested does
Nothing. A ratio with a missing lift is reported as unchecked rather than filled in with an average, and if the only number you have is your snatch, you get a session and no diagnosis. That is the honest answer, and it is a better one than a limiter invented out of one data point.
How it reads a video
A pose model runs in your browser and turns each frame into twelve body joints — no face among them, and the video itself is never uploaded by anyone. Signed out those joints stay in the browser too; with anaccount they are kept with the read, so a lift filmed on a phone can be watched back on a laptop. The wrists are taken as the bar, because in a snatch the bar is in the hands the whole way; the detector has never seen a barbell and the numbers below are honest about being the path of your hands.
Full extension is found first — the hips are higher there than at any other moment of a snatch, since you are stood up on your toes — and the catch, lift-off, knee, pocket and lockout are located from it. If the hips never drop into a catch, or the bar never ends up above your shoulders with the arms locked, there is no rep and nothing is measured. It does not guess at where the phases might have been.
Then nine measurements, all in torso lengths (shoulder to hip, roughly 45–50cm on an adult) and degrees rather than pixels, so a clip shot from three metres reads the same as one shot from one:
| What is measured | Allowed | Called |
|---|---|---|
| Bar away from the mid-foot | up to 0.2 torso lengths in front of the mid-foot | The bar swings out |
| Elbow angle before the finish | 165° or straighter until the hips finish | Pulling with the arms |
| Hip opening at the finish | the hips open to 168° or more | Cutting the finish |
| Back angle change to the knee | up to 10° of back-angle change to the knee | Ripping it off the floor |
| Foot travel | up to 0.2 torso lengths of foot travel | Jumping forward |
| Hips against the knees in the catch | hips level with the knees or below them | Catching high |
| Time from the finish to lockout | up to 0.35s from the finish of the pull to lockout | Slow under the bar |
| Elbow angle in the bottom of the catch | 165° or straighter in the bottom of the catch | Pressing it out |
| Bar wander in the catch | up to 0.12 torso lengths of drift overhead | Wobbly overhead |
These are conventional coaching margins — the amount a coach would let go without saying anything — reasoned from practice and not fitted to anybody's data, exactly like the bands above. Past the allowance a reading gets worse smoothly rather than flipping from fine to broken, and the score is the sum of what that costs, clamped at zero.
The numbers worth tracking
Every read keeps the number behind each measurement, andProgress shows each one as a trend of its own: the mean of your last few lifts against the few before them. The score cannot do that job — it averages everything together, so the one thing you have been drilling for a month can improve without the score saying so.
Two of them are speeds and are never judged: how fast you drop under the bar after the finish, and how fast your hands rise through the second pull. There is no target for either — how fast is fast enough depends on the weight, your build and how deep you catch, and nobody here can defend a number for it. They cost nothing in the score and they are only compared with your own earlier lifts, so compare lifts at similar weights. Hand speed is exactly that: the detector sees hands, not a barbell.
The ones to watch first are the five the score weighs heaviest, and the speed under the bar beside the turnover time — the time says how long it took to get under it, the speed says how hard you went.
| Number | Why it matters | What is measured |
|---|---|---|
| Bar away from the mid-foot watch first lower is better | A bar that swings away from you has to be chased forward to be caught, and a bar you chase is a bar you miss in front. Keeping it close is what lets the pull finish straight up and the catch land over your feet. It is the heaviest single item in the score. | How far in front of your mid-foot your wrists got between lift-off and the finish of the pull, in torso lengths. Side-on only — filmed head-on there is no forward to measure. |
| Elbow angle before the finish watch first higher is better | Arms that bend before the hips finish hand the lift from your legs and hips to muscles that are far weaker, and they usually cut the extension short as well. Long arms through the pull are what let the legs do the lifting. | The tightest your elbows got between lift-off and the finish of the pull, in degrees. Straighter is better. |
| Hip opening at the finish watch first higher is better | The finish is where the bar gets most of its speed. Leaving for the bottom before the hips open throws away the part of the pull that does the most work, and it is usually the first thing to go as the weight climbs. | How far your hips opened at the moment they were highest, in degrees. More open is better. |
| Time from the finish to lockout watch first lower is better | At the finish the bar is only rising for a moment. Every fraction of a second between the finish and a locked arm is time the bar spends slowing with nobody under it, and it is usually the first thing to slow down as the load goes up. | The time from the finish of the pull to the first frame with your arms locked overhead, in seconds. Shorter is better. Withheld on a recording too slow to time it. |
| Speed under the bar watch first higher is better · not scored | How fast you drop under the bar after the finish decides how high the bar has to be pulled: the faster you go, the lower you can meet it. It is the number to watch while drilling the turnover — the turnover time says how long it took, this says how hard you went. | The average speed of your hips from their highest point at the finish to their lowest in the catch, in torso lengths a second. No target is given, only your own earlier lifts to compare against — and the load matters, so compare reads at similar weights. |
| Hips against the knees in the catch watch first lower is better | Receiving high means the bar had to be pulled higher than it needed to be. The heaviest weights cannot be pulled high, so being fast and comfortable meeting the bar low is what lets them be caught at all. | Hip height minus knee height in the bottom of the catch, in torso lengths. Below zero is below parallel; lower is better. |
| Back angle change to the knee lower is better | If your chest comes up before the bar reaches your knee, the hips have shot up first and the back is doing the legs’ job. The bar then arrives at the hip in the wrong place to finish. A patient first pull sets up the second. | How much your back angle changed from lift-off to the bar passing the knee, in degrees. Less is better. |
| Foot travel lower is better | Feet landing in front of where they started usually mean the bar went forward and you followed it. It is often the visible symptom of a bar-path problem rather than a problem of its own, so watch it alongside the bar. | How far your ankles travelled forward from lift-off to the catch, in torso lengths. Side-on only. |
| Elbow angle in the bottom of the catch higher is better | Arms still bent in the bottom mean the lockout was finished by pressing — a no-lift in competition and an unstable catch in training. The bar should arrive already locked. | Your elbow angle in the bottom of the catch, in degrees. Straighter is better. |
| Bar wander in the catch lower is better | A bar that wanders overhead is being saved rather than held. Less drift means it landed over your base, and that is what makes standing up a heavy catch possible. | How far your wrists moved front to back between lockout and standing, in torso lengths. Less is better. |
| Peak hand speed in the second pull higher is better · not scored | The speed the second pull produces is what the finish is for. If it falls away as the weights rise, the extension is where the work is. | The fastest your wrists rose between the hip crease and the finish, in torso lengths a second. The detector sees hands, not a barbell, so this is hand speed and not a bar speed off an encoder. No target is given; compare reads at similar weights. |
Speed in a drill session
A filmed drill session is not read for faults — the video read judges a full squat snatch, and its tolerances say nothing true about a pull or a balance. What a drill session does get, when speed is what the drill is for, is how fast your hands rose and how fast your hips dropped anywhere in the set. Each is the fastest stretch of three frames, so one frame where the detector lost a wrist does not count. Like the speeds above they are never judged, andProgress shows each week's beside the heaviest load you logged that week, because a speed means little without the weight it was moved at.
| Speed | Why it matters | What is measured | Measured in |
|---|---|---|---|
| Hand speed torsos/s · not scored | The speed the pull gives the bar is what the finish is for. Watched against the load, it says whether the extension is keeping up as the weight goes on — and a pull that slows down sooner than it should is where the work is. | The fastest your wrists rose anywhere in the recording, in torso lengths a second. The detector sees hands, not a barbell, so this is hand speed and not a bar speed off an encoder. | Squat snatch, Power snatch, Tall snatch, Tall squat snatch, High-hang squat snatch, No-foot squat snatch, Muscle snatch, Dip snatch, Snatch from the power position, No-contact snatch, No-hook squat snatch, High-hang power snatch, Hang power snatch, above the knee, Hang power snatch, below the knee, Block power snatch (knee height), No-foot power snatch, Snatch pull, Snatch high pull, Floating snatch pull, Tempo snatch pull, Deficit snatch pull, Paused snatch pull, Hang snatch high pull, Hang snatch, above the knee, Hang snatch, below the knee, Snatch from blocks (knee height), Paused squat snatch at the knee, Slow-pull squat snatch, Segment squat snatch, Deficit squat snatch, Snatch from low blocks (below the knee), Snatch from high blocks (above the knee), Paused squat snatch below the knee, Floating squat snatch, Muscle snatch + overhead squat, Snatch high pull + tall snatch, Snatch pull + squat snatch, Snatch high pull + squat snatch, Hang snatch + squat snatch, Muscle snatch + overhead squat + squat snatch, Snatch balance + squat snatch, 3-position squat snatch (hip, knee, floor), Power snatch + overhead squat, Squat snatch + overhead squat, Power snatch + squat snatch, Power snatch + snatch balance, Drop snatch + squat snatch, Snatch deadlift + squat snatch |
| Drop speed torsos/s · not scored | How fast you go under the bar decides how high it has to be pulled: the faster you drop, the lower you can meet it. It is the number to watch while drilling the turnover and the receiving position. | The fastest your hips went down anywhere in the recording, in torso lengths a second — in these drills, the drop under the bar. Only measured for drills that get under the bar. | Squat snatch, Power snatch, Snatch balance, Heaving snatch balance, Drop snatch, Snatch balance with a receiving pause, Tall snatch, Tall squat snatch, High-hang squat snatch, No-foot squat snatch, Dip snatch, Snatch from the power position, No-contact snatch, No-hook squat snatch, High-hang power snatch, Hang power snatch, above the knee, Hang power snatch, below the knee, Block power snatch (knee height), No-foot power snatch, Hang snatch, above the knee, Hang snatch, below the knee, Snatch from blocks (knee height), Paused squat snatch at the knee, Slow-pull squat snatch, Segment squat snatch, Deficit squat snatch, Snatch from low blocks (below the knee), Snatch from high blocks (above the knee), Paused squat snatch below the knee, Floating squat snatch, Snatch high pull + tall snatch, Snatch pull + squat snatch, Snatch high pull + squat snatch, Hang snatch + squat snatch, Muscle snatch + overhead squat + squat snatch, Snatch balance + squat snatch, 3-position squat snatch (hip, knee, floor), Power snatch + overhead squat, Squat snatch + overhead squat, Power snatch + squat snatch, Power snatch + snatch balance, Drop snatch + squat snatch, Snatch deadlift + squat snatch, Snatch balance + overhead squat |
No speed is measured in Snatch-grip press, Behind-the-neck snatch press, Sots press, Overhead squat, Paused overhead squat, Snatch-grip push press, Pressing snatch balance, Tempo overhead squat, Overhead hold in the squat, Snatch deadlift, Paused snatch deadlift, Halting snatch deadlift, Snatch-grip Romanian deadlift. They are slow on purpose, or the bar is already overhead, and the fastest moment of one says nothing about the drill.
What the video read refuses to say
Filmed head-on there is no forward, so nothing at all is said about bar path or foot travel — those two are listed as unchecked and the clip is told to film side-on. A recording too slow to time a turnover does not get a turnover time. Fewer than four measurements and there is no score, only the readings themselves. Same rule as a lift you have not tested: no measurement, no finding.
Sending a read
After a read, /train offers a link to send it — to a coach, a training partner, the club chat. The link holds the skeleton of the trimmed lift and nothing else: the twelve joints of every frame and when each was taken, packed to 64 bytes a frame, with no video, no face and none of the read's words or numbers. It sits after the # in the address, which a browser keeps to itself rather than sending to any server, this one included, and the page it opens takes it out of the address before the visit counter loads, so the counter sees /read and nothing more. The page it opens runs the same engine on those joints in the recipient's tab, so they see the read the engine makes, not one the link claims — and what this clip could not check stays unchecked there too. A lift too long for a link is not thinned to fit, because fewer frames would quietly stop the turnover being timed; the page asks for a tighter trim instead.