Taking the back from turtle

Turtle is what somebody does when they have decided that being on all fours is better than being on their back. Knees under hips, elbows in, head up or tucked. It is a defensive shape rather than a system, and it has no legs in front of anybody, which is the property that separates it from every guard.

That matters for how it is treated. A guard is a position that can attack. Turtle mostly cannot, which is why the rule books handle it as a state to be resolved rather than a position to be scored. The IBJJF awards takedown points against somebody who ends up on all fours only when the athlete who put them there controls the back, without any requirement for hooks, while keeping at least one of the opponent's knees on the ground for three seconds. The IJF's rules describe a kneeling athlete who may not grab the legs to defend a throw, and specify that once that athlete has two elbows and two knees on the floor, the standing player may apply a technique only to transition into ground work.

Both documents are doing the same thing from different directions: they are treating turtle as a temporary condition and setting a clock on it. The person on top is expected to convert. The person underneath is expected to move.

What follows is about that conversion. It is a transition rather than a submission, and it is worth its own treatment because the sequence people use to take a back is usually taught backwards.

The most useful evidence about back control is not a coaching argument. It is that three organisations have written definitions of it, and the definitions share a structure.

United World Wrestling's grappling rules award four points for back mount when a grappler controls the opponent from the back with their chest to the opponent's back and their heels between both of the opponent's thighs, or in a body triangle, with the legs positioned to trap up to one of the opponent's arms, held for three seconds. The IBJJF awards four points when an athlete takes control of the opponent's back placing their heels between the thighs without crossing their legs, in a position to trap up to one of the opponent's arms, without trapping the arm above the shoulder line, and remains there for three seconds. The ADCC rules give three points for back mount with hooks, accept both hooks or a body triangle, and specify that the hooks must not be over either shoulder.

Three tables and three point values. All three describe the legs; the IBJJF and UWW cap how many arms may be trapped, and ADCC bars hooks over the shoulders. Only United World Wrestling states a chest requirement in the clause itself; the IBJJF and ADCC definitions describe the legs alone. Where they differ is instructive: UWW and ADCC accept a body triangle for full points; the IBJJF gives only an advantage for it, and the same reduced score applies when the legs are crossed, when only one heel is in, or when both of the opponent's arms are trapped.

The chest connection is a coaching commonplace rather than a rule most of these documents state, and it is worth being precise about that: only one of the three requires it. What all three do require is that the legs are in, and that is what the scoring turns on.

The usual teaching order is hooks, then harness, then attack. The order that survives contact is the reverse.

The reason is geometric. Somebody escapes the back by rotating their shoulders toward you and getting their back to the mat. That rotation is fought at the shoulder line. If your chest is fixed to their upper back with the over-shoulder arm deep and the under-arm elbow tight to your own ribs, they can move their hips a considerable distance without changing which way they are facing. Two hooks with a loose chest, by contrast, is a position somebody can turn inside all day, and it is a common way to lose a back while believing you still have one.

From turtle specifically, this changes what you reach for. The first thing to establish is not a leg. It is the connection: chest onto the upper back, one arm over the shoulder and one under the far armpit, hands joined. If you can get that, the legs are a matter of time. If you cannot, putting a foot inside a thigh gives you something to be rolled over with rather than something to hold on with.

There is a common intermediate that the rules acknowledge. Controlling a turtled opponent's back with no hooks at all is a real position, and both the IBJJF's takedown clause and United World Wrestling's phrase about partial back control treat it as one. It is not a failed back take. It is the state that precedes a back take, and it scores or restarts in your favour under both of those documents.

The harness is usually described as symmetric, and it is not. One arm goes over a shoulder, one under the far armpit, and which is which determines what happens when the position moves.

The rule to carry is that you follow them to the mat on the side of your under-arm shoulder. If they reach the mat on the side of your over-shoulder arm, your chest has to travel over them first, or you end up underneath with the position half gone. If they go to your under-arm side, you arrive on top. Choosing the harness is therefore choosing which direction of escape you are prepared to travel with, before they have started travelling.

From turtle this decision usually gets made for you by where the opening was. If their far arm came up, the arm you thread under it is your under-arm, and you follow them to the mat on that side. If the opening was their hip, take the seat belt so that your under-arm is on the side they are turning toward. The mistake is to take whichever arm is closest and then discover, mid-roll, that you committed to the side you cannot follow to.

One thing not to do, in any of these positions: do not cross your ankles in front of the opponent's body. Under the IBJJF's table it drops a four-point position to an advantage, and it is a weaker structure that offers an attack to a defender who is not even escaping.

Hooks are directional rather than symmetric: each resists the hips travelling toward its own side, so the hook on the side they are escaping toward is the one under attack. Knowing which of yours is under attack tells you which escape is arriving, which is a more useful piece of information than the fact that you have two of them.

The choice between two hooks and a body triangle is genuinely unsettled, and the rule books make the trade-off legible. The triangle is harder to strip and restricts the opponent more, but it commits both of your legs to one side and can be turned into. Two hooks are more mobile and give up less if the opponent stands. And under the IBJJF, the triangle is worth an advantage where two hooks are worth four points, which is a reason to have both available rather than a reason to pick one.

The general principle for anybody losing backs is easier to state than to do. Recover the harness before you recover the legs. Losing both hooks with the chest connection intact is a recoverable position. Losing the chest connection with both hooks in is often the start of an escape that has already worked.

This transition has real risk in it, and unlike a submission the risk is not concentrated in one joint. It is spread across a moment when two bodies are rotating and neither person has full control of the other's head.

The rule books have written down the specific dangers, which is the most direct evidence available for what they are. The ADCC rules prohibit spiking an opponent's head when the opponent is on your back, and prohibit the person on the bottom from rolling forward with somebody attached. The IJF makes it a direct disqualification to intentionally fall backwards when the other athlete is clinging to your back and either athlete has control of the other's movement, and separately makes it a disqualification to take any action that may endanger or injure an opponent, singling out the neck and the spinal vertebrae. Three separate clauses, in two documents, all about the same short stretch of a fight.

In a training room that becomes a short list. Do not roll forward or backwards to escape when somebody is attached to your back, because their head and neck are not where you can see them. Do not drive a defender's head into the mat to stop a turn. If you are riding a back and the position starts to invert or roll, the correct response is to let it happen slowly rather than to hold harder through the rotation.

The other item belongs to the person on top, and it is the body triangle. Locked around the midsection and squeezed hard, it stops being a control and becomes something people tap to. That tap deserves the same immediate, complete release as any other, and a body triangle should not be squeezed to make a stalled position feel productive. All of this belongs in a supervised room, with a partner who agreed to the terms of the round, and with both people tapping early rather than proving a point.

Rule books are good evidence for what a position is and poor evidence for how often anything works. Little has been published on this. Lamas and colleagues, modelling the 93 matches of a single 2019 no-gi event as transitions between actions, report that the back-take had the highest probability of any action of leading directly to a submission, at 0.45. That is one event and one model, and the abstract we read does not say exactly how a back-take was coded. We found no measurement of how often a turtle becomes a back take, or of whether the chest-first ordering argued for here outperforms the alternative.

The closest relevant work is small. Spanias, Kirk and Ovretveit, coding 26 no-gi submission-only matches, found that time spent in dominant positions, a category that included back control, correlated with winning by an upper-body submission and did not correlate with winning by a lower-body one. In their sample the most frequent finish overall was the heel hook, and the athletes who won that way had spent no time in dominant positions at all. Under a ruleset with no points, position and finish came apart.

That is worth holding onto as a caution against the version of this article that could have been written instead: the confident one, in which taking the back is the point of grappling and everything else is a route to it. Under the scoring tables quoted here it is the highest-value position under ADCC, and it ties with mount at four points under the IBJJF and UWW. Under a submission-only format with no points, the sample we have suggests it is one route among several.

What survives both readings is narrower and better supported. Back control is defined by three organisations through the legs, and by one of them through the chest as well. The transition from turtle is the business of getting the chest connection before the legs. Everything the position later does, including the strangle it is famous for, depends on that ordering rather than on the legs everybody reaches for first.

Correction, 29 September 2026: an earlier version of this article said three rule books define back control as a chest connection, and called it the highest-value position under all of them. Only United World Wrestling's definition includes the chest, and the back ties with mount at four points under the IBJJF and UWW. It also said nobody had measured how often a back take produces a finish; a 2024 model of a 2019 no-gi event by Lamas and colleagues reports one such figure, and is now cited. The IBJJF all-fours rule is about the opponent's knee, and the seat-belt and hook directions are now named by arm.