The most powerful predictor of MVP among all existing stats is Win Shares :http://popgates.com/projecting-the-nba-regular-season-awards/
In studying the Most Valuable Players of the last 30 years, I focused primarily on the advanced statistics Player Efficiency Rating (PER), Win Shares (WS), Box Plus-Minus (BPM), and Value Over Replacement Player (VORP). One aspect I also tracked was whether or not a given MVP played for the best team in the regular season. In the last 30 years, the representative frequency of the MVP leading the league in each statistic was 43% (PER), 56% (WS), 40% (BPM), and 47% (VORP). The MVP was also on the best regular season team 56% of the time. These percentages tell me that there is a good likelihood that this year’s MVP will lead the league in one or more of these categories.
Leading the league in WS is an equally strong a marker of MVP as finishing with the best team record (56%)
Harden leads the league in WS, 2nd in BPM and VORP, and 3rd in WS/48 and PER.
Westbrook leads the league in PER, BPM, and VORP, but he's only 5th and 9th in WS and WS/48.
Kawhi is 2nd in WS,WS/48 and PER, but only 5th and 8th in VORP and BPM.
LeBron is 7th in WS, 8th in WS/48, 6th in PER, 4th in BPM, and 3rd in VORP.Harden is the only player who is top 3 in all major advanced statistical categories
(except RPM.) He also has the best TS% among MVP candidates, and 2nd best team record.
Due to this consistency factor, basketball-reference has Harden way ahead in this MVP race http://static.bkref.com/friv/mvp.html