As for how much is believable is debatable, but I am inclined to take his word since he states it was under blind, not just sighted conditions, and they have been testing their designs far longer than audiophile would probably demo the devices.
A test being blind is insufficient to draw any conclusions. The test needs to be repeated a number of times to reduce the chances that the person just got lucky. If I flipped a coin and predicted tails and that is what became, would you believe that I can always predict a coin toss?
I have run blind ABX tests where I got a number right only to get lost from there on. Here is an example:
foo_abx 1.3.4 report
foobar2000 v1.3.2
2014/07/19 07:09:23
File A: C:\Users\Amir\Music\Arny's Generational Loss\sb20x_original.wav
File B: C:\Users\Amir\Music\Arny's Generational Loss\sb20x_pass1f.wav
07:09:23 : Test started.
07:09:40 : 01/01 50.0%
07:09:54 : 01/02 75.0%
07:10:07 : 01/03 87.5%
07:10:26 : 02/04 68.8%
07:10:36 : 03/05 50.0%
07:10:51 : 04/06 34.4%
07:11:03 : 05/07 22.7%
07:11:14 : 05/08 36.3%
07:11:29 : 05/09 50.0%
07:11:44 : 06/10 37.7%
07:12:01 : 07/11 27.4%
07:12:16 : 08/12 19.4%
07:12:37 : 09/13 13.3%
07:13:23 : 09/14 21.2%
07:13:34 : 10/15 15.1%
07:13:47 : 11/16 10.5%
07:14:00 : 12/17 7.2%
07:14:13 : 13/18 4.8%
07:14:23 : 13/19 8.4%
07:14:37 : 14/20 5.8%
07:15:09 : 14/21 9.5%
07:15:22 : 15/22 6.7%
07:15:39 : 16/23 4.7%
07:15:59 : 16/24 7.6%
07:16:12 : 16/25 11.5%
07:16:34 : 17/26 8.4%
07:16:48 : 18/27 6.1%
07:17:09 : 18/28 9.2%
07:17:53 : 18/29 13.2%
07:18:11 : 19/30 10.0%
07:18:34 : 19/31 14.1%
07:19:13 : 19/32 18.9%
07:19:18 : Test finished.
----------
Total: 19/32 (18.9%)
As you see, I got 19 out of 32 right at the end which is not a valid result. But the bolded section appears to make a case that I was guessing correctly four times in a row.
You have to get results like this to be definitive:
foo_abx 1.3.4 report
foobar2000 v1.3.2
2017/11/26 10:50:22
File A: C:\Users\Amir\Documents\Test Music\Headfi RR Samples\07094.wav
File B: C:\Users\Amir\Documents\Test Music\Headfi RR Samples\27776.wav
10:50:22 : Test started.
10:51:03 : 01/01 50.0%
10:51:17 : 02/02 25.0%
10:51:25 : 03/03 12.5%
10:51:36 : 04/04 6.3%
10:51:42 : 05/05 3.1%
10:51:52 : 06/06 1.6%
10:52:00 : 07/07 0.8%
10:52:10 : 08/08 0.4%
10:52:17 : 09/09 0.2%
10:52:25 : 10/10 0.1%
10:52:36 : 11/11 0.0%
10:52:41 : Test finished.
----------
Total: 11/11 (0.0%)
Now probability of guessing is pretty much the same as zero.