Having done some further measurements, I'm pretty sure zoom_sensitivity_ratio 0.793471 is correct.

According to the link trippa posted, in CS:GO at zoom_sensitivity_ratio_mouse 1 (equivalent to zoom_sensitivity_ratio 1 in TF2) the distance the mouse needs to be moved to move the crosshair from the middle to the left or right edge of the screen at 4:3 is the same when scoped and unscoped.

Having just checked this, the same is true in TF2 at zoom_sensitivity_ratio 1.

This is only true at 4:3 - the game does not compensate for different aspect ratios.

With this information, the following formula can be derived:

zoom_sensitivity_ratio = (unscoped FOV / scoped FOV) * tan (scoped FOV / 2) * cot (unscoped FOV / 2)

At fov_desired 90 and with a scoped FOV of 20 (the value used in TF2), this gives zoom_sensitivity_ratio = 0.793471.

This formula is the same as the one used in the CS:GO thread, the final answer is different because the zoomed FOV for the CS:GO AWP (40 and 10 degrees depending on zoom level, according to the internet) is not the same as for the TF2 sniper rifle (20 degrees).

Additionally, with the scope overlay removed, scoped FOV scales with aspect ratio in the same way as unscoped FOV (vertical FOV is constant, horizontal FOV increases as aspect ratio increases). A combination of this and TF2 ignoring aspect ratio when calculating scoped sensitivity means that the correct value of zoom_sensitivity_ratio is independent of aspect ratio.

I imagine you are correct that 0.81 is close enough to the correct value that it makes no difference. There is no gameplay benefit to finding a more accurate value, I'm just curious as to what the exact value is, and the maths behind finding it.

However, I consistently measured that, for example, 0.815 was too large, so I think that the differences between the value I measured as correct (0.81) and the values other people have linked me to are significant enough to suggest that the calculated values are incorrect.

Given that 0.833333 is even further from 0.81 than the other two values suggested, I don't think that that is right either.

The justification given is also pretty weak. As I said, my attempt at doing the maths properly failed, so I can't prove that what genemilder is saying is wrong, but I don't think that trig works like that.

The problem with this is that the zoomed horizontal FOV of the TF2 sniper rifle at 4:3 is 20 degrees, whereas the value for the CS:GO AWP is 40 degrees.

This post in the thread I linked initially provided the following equation:

(tan(zoom fov/2)*hip fov)/(tan(hip fov/2)*zoom fov)

Plugging in 90 degrees for hip FOV and 20 degrees for zoom FOV gives a value of 0.793471.

Again, the value I measured was 0.81, not 0.818933 or 0.793471. Admittedly my method was crude, but the difference is large enough that I don't think either 0.818933 or 0.793471 is the correct value.

EDIT: I now believe the exact value to be 0.793471.

Having found a discussion in this thread on what value of zoom_sensitivity_ratio makes your sensitivity feel as close to the same as possible both in and out of scope at fov_desired 90, I tried to find the correct value by using a ruler to measure the distance my mouse needed to be moved to move my crosshair a set number of pixels when in and out of scope.

The correct value is about 0.81.

This doesn't match any of the values suggested in the aforementioned thread, and I can't figure out why it would be the correct value - my attempt at calculating the exact value gave an answer closer to 0.75, which is obviously wrong. If someone who knows how TF2 calculates the scoped sensitivity could explain how to calculate the exact value, that would be very much appreciated.

As a sidenote, I used a similarly crude method to measure the horizontal FOV when scoped in. With the scope overlay removed, this is about 26 degrees at 16:9. Assuming that scoped FOV is affected by aspect ratio in the same way as unscoped FOV, the value at 4:3 is therefore 20 degrees with the scope overlay removed.

After some further playing around, I found that neither 1.0 nor 1.2 hide the spies PDA (disguise kit), even when set to do so. Both versions will hide the engineers build and destroy PDAs.

I also feel I should clarify that I never tried the 1.1 installer, so it could have the same issues as 1.2 for all I know. I never had any issues with the 1.0 installer.

When set to hide the spies sapper, both 1.0 and 1.2 hide the stock sapper but not the red tape recorder for me.
I also got the same error as Carn using the 1.2 installer on the first computer I tried, however the installer worked on a second computer (both Windows 10, I don't know of any difference between them that would cause the error).