Which direction will the gymnastics tilt twist turn me? Figuring out rotation direction in torque-free precession

I'm trying to understand the physics of tilt twisting in gymnastics / trampolining, and I've run into a conundrum where theory doesn't seem to agree with practice, so I'm looking for where my mistake could be.

In tilt twisting, the gymnast performs a flip, and, during the flip, also starts turning. The motion is described by an Olympic champion in this video, and at 2:38, it says that, for a back flip, LEFT twist = LEFT arm drops first.

Here is how I formalized the problem: Let the gymnast be standing, with the positive X-axis pointing towards his right side, the positive Y-axis pointing towards his head, and the positive Z-axis pointing towards his front. Assume the gymnast is in the air doing a backflip; this would mean that the angular momentum only has a positive X component, and zero Y Z components (assuming I've correctly used the right hand rule; a backflip is the same rotation motion as when falling backwards from standing).

I also assume that the gymnast, when standing aligned with the axes, has a moment of inertia that is a diagonal matrix, and that a gymnast is much taller than either wide or long (fat?), meaning that the Y component of the moment of inertia matrix is lower than the other two non-zero entries.

Now, assume the gymnast has jumped into the air with his arms up, with a moment of inertia as described above, and, once in the air, the gymnast drops his left arm completely straight tracing a path in the X-Y plane, and so rotating around an axis parallel to the Z-axis. This would cause his entire body to tilt to the right, let's make a simple assumption and say less than $$45^{\circ}$$. A slight tilting to the right can be represented by a rotation vector with a negative Z component and zero X and Y components, or a rotation matrix of the form $$\mathbf{R}=\left[\begin{array}{l}a&b&0\\-b&a&0\\0&0&1\end{array}\right],$$ with $$a, b > 0$$, and we can use this to compute the updated moment of inertia using $$\mathbf{I}_\text{new} = \mathbf{RIR^T}$$, which, given the gymnast's height is much greater than his other dimensions, will yield a matrix of all non-negative numbers.

Computing the angular velocity based on the formula $$\mathbf{h} = \mathbf{I}_\text{new}\mathbf{\omega}$$ will yield a vector with negative values and zeros for the entries of $$\omega$$, as $$\mathbf{h}$$ is composed of all negative values and zeros, and $$\mathbf{I}_\text{new}$$ of all positive values and zeros.

So the angular velocity in the Y-axis is negative, which according to the right-hand rule implies spinning to the right, yet contradicting the video that dropping the left arm during a backflip results in a left twist.

What is my error here?

• The most highly voted comment for "Check-my-work" link provided says: "I think teaching is important, and doing worked examples is an important part of teaching. I'm not saying I want the site to fill up with posts asking how to do a worked example, but I think answering such questions can illustrate important concepts. [...] I would urge site members to consider whether the mistake is conceptual rather than arithmetical. If it's a conceptual mistake I think there is some justification for answering the question as long as your answer is mainly concerned with the concepts involved."
– Gabi
Commented Jan 7 at 9:12