At maybe a 1500 OTB player, I just like 1.b4 because
a) am tired of e4 and THEM controlling somewhat because they have canned responses to it
b) it surprises them and takes them out of their comfort zone
c) it can contain a lot of traps for black if they don't play sharply
d) I like to have the attack and make them guess at what they're doing
e) if it's good enough for Bobby Fischer and Tartakower...
Funny, I got the book yesterday and devoured the first 20 pages...he starts off with the f6 reply lines. Jumped on chess.com to try it out in LIVE games and it gave me Black three times in a row. 
Will try again today, hoping to get white.
Someone thought I wouldn't get much out of the book...well, I've found that it's the little sidelines where you may get the most out of it...that's where they make a not so great move and then you can hit them with good/great tactics. I learned what to do if they brought Nc6 out, push my b pawn and most likely they trot on down toward my men, of which I respond with c4; then their N is a goner. I hadn't thought of that response.
Chapter 2 is Black taking the b pawn on the 3rd move.
At my level, or sub 1800, I don't think the +- 0.3 moves matter so much either.
The last game, while playing Black, dreaming about the Orangutan, we were toward the end of middle game, and I had a family fork tactic, hitting his King and both rooks. I went temporarily blind and didn't see it. Two moves later, I still hit the fork and eventually got one rook and went on to win the game; but it so aggravated me. Does that ever happen to higher players? You seemingly go blind and don't see an easy tactic?
As one that has played the Sokolsky a good 100 to 200 times as White in Over the Board competition, here's how to look at the b-pawn:
1) Against 1...e5, don't protect it. Attack the e-pawn. After 1.b4 e5 2.Bb2 Bxb4 (Black's best try at equalizing) 3.Bxe5 Nf6 4.c4 (the other option is 4.e3/5.Nf3 going for fast development but completely relinquishing the center to Black), White gave up a wing pawn for a center pawn with an open diagonal for his Bishop.
According to all top engines after 3.Bxe5 Nf6, white is around -0.1 to -0.3 worse. Let's try to understand why:
1) Black has two pieces developed to white's one. In addition, white's bishop on e5 will have to lose another tempo.
2) In the upcoming middlegame, white lacks a plan. Black's development is comparably easier and he has strong central pieces and can occupy the center with pieces (Nf6, Re8) or pawns (c5,d5).
What does white have going for him? Not much, really. But he's more solid and has the open long diagonal which he can use to his benefit. There is no reason why white should be better.
Just didn't really understand the terminology of "black's best try at equalizing." If anyone's equalizing, it should be white.
1.b4 isn't the way to go for an advantage, anyway. 1.b3 is much better.
IM title or no IM title, I am going to counter what you just said 100%
You, an IM, of all people, should know that engines are clueless at evaluating any position after 3 moves! I don't care whether those moves are 1.d4 d5 2.c4 e6 3.Nc3 Nf6 or 1.a4 g5 2.Nc3 g4 3.Na2 Nh6, engines are clueless at evaluating these positions!
Secondly, even database statistics aren't always valid either. You've got average rating of players, and number of games. There's more validity statistically in a line with 45,000 games than in a line with 200 games. If you went strictly on Database percentages, White scores 90% after 1.b4 e5 2.Bb2 Bxb4 3.Bxe5 Nf6 4.h3!! on 365chess.com. I think we all know that you can't expect a 90% score from that position as White.
Thirdly, How can you possibly claim database validity when the percentages jump around just based on rating? And it's not like lower rated is always better! Take the database at ChessTempo. The position after 1.b4 e5 2.Bb2 Bxb4 3.Bxe5 Nf6. Now consider the following. Let's just look at White's 2 main options. 4.Nf3 and 4.c4
If you restrict the database to players over 2200 (both sides), you have 119 games combined, White's performance rating is lower than his average rating in both cases (4.Nf3 and 4.c4), and Black scores roughly 54 to 55 percent in both lines. (Black's win percentage plus half the draw percentage = Black's total percent score)
But now let's restrict it to even stronger players. 2400 or above for both players. We see White scoring in the mid-50s after 4.Nf3 and upper-50s after 4.c4. White's performance rating in both cases are well above their average rating, 50 points above after 4.Nf3 and about 80 points above after 4.c4.
Go to games where Both playes are over 2500 and suddenly Black is 3 and 0! 100% score for Black, but just like the 5 games of 4.h3 in 365chess.com scoring 90% for White, the validity is almost zero with only 3 games.
1.b4 is not "bad". Is it as strong as 1.d4 (best by "database percentage" if you really are gun-ho on that) or 1.e4 (A full 2 percent lower than 1.d4)? Of course not!
I forget where I read it, but a number of years back, I had read somewhere an article that had gathered almost every database that was out there at the time, like chessbase, newinchess, etc, and did a score comparison across all of them rather than 1, and 1.b4 as I recall had scored 50.2% (I was playing it as far back as then, hence why I remembered), compared to 1.d4's score of 56.1% and 1.e4's score of 54.1%. 1.c4 scored about 52%, 1.Nf3 was near 53%. Most other flank openings sat between 50 and 51%, with 1.f4 standing out as the worst with about 46 and a half percent.
If White is happy with a basically equal game that he understands and Black doesn't, there's nothing wrong with 1.b4. At the world championship level where you need to try to win with White and simply try to hold on in your Black games (worked for Kramnik in 2000 going with this approach), then yes, 1.b4 is probably not recommended.