Battle of the high-res mirrorless IS systems

Both the Nikon Z7 and Sony a7R III feature 5-axis in-body image stabilization. When tested against one another at 24mm, 55mm and 200mm both cameras' IS systems proved similarly useful. Users can expect around 2 to 3-stops IS advantage when shooting with wide angle lenses, 3 to 4-stops when using normal lenses and 4 to 5-stops when shooting with telephotos (helped by both cameras making use of in-lens IS at 200mm).
Nikon Z7 vs a7R III IS test results:

2.3-stops 4-stops 4-stops
2.6-stops 3.3-stops 4.6-stops

These results correlate pretty similarly to our Sony a7 III IS test. That camera is rated as being 0.5EV less well corrected than the a7R III, so it's not surprising to see that it performed about 1 stop behind its high-res counterpart at 200mm.
How we test and why

Our standard image stabilization test is performed by shooting 10 frames of our test target, starting with a shutter speed of 1/focal length and working our way downward in intervals of 2/3-stops using a neutral density filter (to maintain the same aperture). The test is done at 24mm, 55mm and 200mm, since the challenges of stabilization change with focal length and exposure duration. Note: We use electronic first curtain shutter for this test, when available, to avoid shutter shock skewing our results.
We first do this with IS on, and then with IS off. The images for both are then assessed to the following standard:
Excellent Some softness
Noticeable blur Unusable
1) Excellent - the best the camera is capable of
2) Some softness - only distinguishable from Excellent by comparison
3) Noticeable blur - blur can be perceived without reference to another image
4) Unusable - significant motion perceptible
After each frame is assessed, we plot the results in the graphs you see below. We then compare the lowest shutter speed at which the camera was able to maintain a 50% or greater hit rate with IS on and off at a given focal length (images that are acceptably sharp are those rated 'Excellent' or 'Some softness'). This allows us to determine how effective the IS system is in real world shooting.
Our results compared to CIPA

CIPA offers a good baseline to compare cameras, but CIPA tests tend to use focal lengths around 50mm (which are relatively easy to stabilize) and do not include any rolling motion, which can give very high numbers. For instance, the Sony a7R III's in-body stabilization is quoted as offering a '5.5-stop shutter speed advantage,' according to CIPA testing (the Nikon was rated at 5-stops). That suggests you can get usable shots, consistently, at 0.8 secs (5.3-stops below one over focal length), which we could not reproduce: at 1/2 sec (4.6-stops below) we were only about to get one usable shot in ten with IS on.
Nikon results

For 24mm and 55mm we tested the Z7 using the Z 24-70mm F4. At 200mm we used used an adapted Nikon 70-200m F2.8E FL ED VR.
At 24mm with IS on, the Z7 maintains a 60% acceptable hit rate at 1/5 sec and a 40% at 1/3 sec, which indicates 1/4 sec would likely be where it'd hit the 50% acceptable threshold. With IS off, the acceptable rate is 60% at 1/20 sec and plummets to 10% thereafter. The results indicate about a 2.3-stop IS advantage at 24mm.
With IS on, the 50% acceptable hit rate at 55mm should be 1/4 sec - the Z7 maintains an 80% acceptable rate at 1/5 sec and a 40% at 1/3 sec. With IS off, we weren't able to hit the 50% acceptable threshold at 1 over the focal length (40%), but it's safe to assume at a third a stop faster (1/60 sec), 50% should be manageable. 1/60 sec off vs 1/4 sec results in a 4-stop IS advantage at 55mm.
At 200mm, the Z7 maintained a 50% or greater hit rate down to 1/8 sec, our lowest speed on the graph. We pushed it an additional 2/3rds stop slower in testing (down to 1/5 sec) and found it dropped to a 30% hit rate. With IS off we were able to get 50% or greater usable down to 1/125 sec. The result is a 4-stop IS advantage at 200mm. This is helped by the use of a VR lens: Nikon says the pitch and yaw corrections are handled by the lens, where possible.
Sony results

At 24mm we tested using the Sony 24-70 F4 OSS, at 55mm we used the Sony 55mm F1.8 and at 200mm we used the Sony 70-200mm F2.8 GM OSS.
At 24mm with IS on, the a7R III maintains a 60% hit rate down to 1/5 sec and 40% hit rate at 1/3 sec, meaning 1/4 sec is where it's likely to hit the 50% threshold. With IS off this threshold is reached around 1/25 sec. This gives the a7R III about a 2.6-stop advantage at 24mm. Note that this result includes the use of a lens with its own stabilization (though sensor shift is usually better-suited to the corrections needed for wide-angle lenses).
At 55mm we opted to de-couple the Sony's sensor IS from lens IS by using a non-stabilized lens (the same we tested on the a7 III). With stabilization on, the hit rate was 60% at 1/8 sec and 40% at 1/4 sec meaning 1/6 sec is the likely 50% point. We weren't able to get a 50% hit rate with IS off at one over the focal length, but it's safe to assume we would have by 1/60 sec (and surely by 1/80 sec). This gives us at least a 3.3-stop advantage at 55mm using sensor IS alone.
Like the Z7, we tested the a7R III at 200mm with IS on down to 1/5 sec. Good thing we did: it's not until this speed that a7R III's IS system also drops below the 50% acceptable threshold with 2 images rated 'some softness.' This means you can expect a 50% or greater hit rate down to 1/8 sec with IS on vs 1/200 sec with IS off, providing a 4.6-stops advantage at 200mm. Like the Nikon, this is helped by the use of a lens with IS, which is able to provide the large movements needed to correct long focal lengths.


More...