I agree that the variation I recommended above is considered less than satisfactory these days but that is besides the point.
First of all after 17.Qa3 the move 17... a5 is considered the weakest out of the four alternatives (many dragonplayers I know
). Either rook to d8 should in my opinion be played and there are a lot of intriguing variations. Some variations that are considered as better for white in several dragonbooks, Fritz assesses as about 0.00 even though white is a pawn up or two, which shows blacks potential (knowing that Fritz is very materialistic...)
Secondly, one shouldn't emphasize so much on the statistics. A little homestudy should be enough to get a position which is more than playable. As I said I've seen players with 2300-2400 Elo sometimes chicken out with 16.Qc3 and sometimes been slaughtered after 16.Qxe7 and 17.Qa3
There are a lot of opening variations that are considered dubious or even refuted, but stil give rise to exciting play with winning chances such as the Schliemann variation of the Ruy Lopez or the Latvian gambit to name a few.
Before one (hopefully
) reaches International Master status, such variations are in my opinion ok once you know a bit of theory and have looked at the position at home.
Even Grandmasters have troubles against a few openings or opening variations that are considered less satisfactory than the rest.
Our own Chris Ward in his book "Unusual queen's gambit decline" says on page 110: "I once went through a two-year period losing only three games with White; amazingly, all of them were to the Chigorin Defence!"
Now I know most 1.d4 players don't fear the Chigorin more than, for instance, the Grunfeld or something...
Just one more comment on this topic: In the Najdorf there is a variation against the English Attack that is considered totally refuted but it didn't stop Vallejo from using it and achieve an advantage as black against Leko himself just recently in Linares.
Bottom line: look at the variations yourself and if you like them, try them and trust yourself, not statistics!