It SORT OF works in stable and predictable situations.
What seems to be happening is that your brain, in all of its neuroplastic glory, is learning to make better inferences from the broken information it receives.
The issue is that this works as long as the inferences are correct. That's fine when you are reading a newspaper, using the computer, etc. as the situation is stable and predictable.
The problem is that when you are suddenly confronted by a situation where the inferences are NOT correct--such as a nighttime emergency situation while driving. Now you are relying on the "uninferenced" data coming in from your eyes and that data is subpar with all the resultant problems.
The best solution is both: fix the data coming in with corrective lenses for unpredictable situations, and train your brain to make better inferences so you can deal with predictable situations better.
But for example, right now I'm in Thailand, traveling here for the first time. Everything is new, unfamiliar and unpredictable, daytime, nightime. I have no problems seeing things, everything is almost perfectly clear and sharp. I needed some time to adapt to a smaller screen of my laptop (I was using 24 inch before), but now I'm doing fine.
Actually, the more you look, the better you see. Like in Aaron Swartz blog post, if you want to retrain your weak legs, you need to walk more. Same with eyes.
I have already written extensively about natural eyesight improvement in my native language, and also I'm a web app designer and developer. I can combine these things into seeing.fun
One thing I know, is that this practice should be relaxed and even fun, definitely not serious. If you become serious, there will be no results.
I also have trouble with low constrast settigs and nighttime, vs daytime. I think it may have something to do with the brain recognizing small things even if the image coming in is blurry, and it uses high contrast to do it. That's my guess.
But does Bates method really work, or is it all just placebo effect?
I have even more dioptres than you, around 5 for each eye and a nasty cylinder (astigmatism).
I don't use glasses or any kind of lenses (don't even have them) for the last ~10 years.
Maybe this sounds to good to be true, but the main principle is that this natural eyesight improvement is not about exercises, but about re-learning the natural seeing habits.
When you re-learn to see things in a natural way (without strain, without fixation), you then see things clearly 24 h/day without any exercises.
It's all about habits, not exercises.
I practice this for more than 10 years. Each day I work with computer for ~10 h., drive a car and do other things, and never wear glasses, even though the traditional ophtalmologic measures clearly indicate that I need strong glasses and I shouldn't see even the biggest letter on the Snellen chart, but I see not only the biggest, but sometimes even the 20/20.
Doctors can't explain this, and only congrats me on my achievement. Of course, the eyesight is not perfect. I see clearly in the daytime, but in the nighttime or low light conditions it becomes much harder to distinguish faces.
The best book I found so far is "Relearning to See" by Thomas R. Quackenbush. The originator of this theory was William Bates.
Actually, there is no clear unified theory on how to achieve this. Everyone interprets it differently and the results are inconsistent. There is also a lot of criticism from the medical establishment.
Natural eyesight improvement really works. And the unified theory, in a form of an app, or a good book, maybe including findings from neuroplasticity, would be a great gift for humanity.