That accuracy rate over 8 years means it deviates by that much over that span. Which is good correlation and means that within a year one would not expect that model to continue to deviate by much from the actual measured. Which was why in December after it had departed for nearly 3 months the call for a serious return to form in this invited blog: https://cointelegraph.com/news/2018-blockchain-and-cryptocurrency-outlook-expert-blog.

No we should not be worried but we should be reasonable about our expectations. The model showed bitcoin market value doubling every 288 days. Which is impressive! But when it nearky doubles within a month from Nov 2017 to Dec 2017 (10K to 19K) we check our expectations, especially when user adoption or any other metric does not nearly match that rise. And maybe not be surprised when it corrects significantly after. And we should also be prepared to slowly let go of bitcoins as the flag bearer for the blockchain space if its fundamental issues are not resolved (and they are still working hard on this right now with solutions like Lightning and we are still watching if Segwit will have any real impact.) In the end, technology and utility always wins out and that would happen as well in this space.

The DUA published on the link above is one place to go to for a sort of measure of DAR. Other measures such as transaction volume includes a lot of nonsense such as dice games and other automated transactions that does not solely capture daily adoption. I wrote a rough program to distill the DUA in real time from the blockchain some years ago and I plan to maybe publish its daily value, maybe here at some point. But even the 2 day lagged DUA from bitcoin.info should be good for a value investor just maybe not for day trading.