People keep on criticism the methodology of the study. It's fair game.
Here is a more detailed report about the study:
What I like in the methodology:
-Real significient price for winners of the games.
-Random pairing of player.
-the grouping of similar behavior was statistic (5 groups was the optimal solution according to "Davies-Bouldin index". This index might be bullshit, but ....hey... sometime you have to be trusting )
-The grouping seems statistically sound and simple (minimize dispersion in cluster)
What I dont like:
-sample size coming from a fair (might skew the population toward certain type)
-They could have report other statistical ways of grouping people to see it the result were still coherent.
And btw it's not a personality test per se. It's a solution to the old problem of rational agent in economy.
It might be a more precise separation than risk-adverse and risk-takers. And it only apply in collaborative interaction with perfect information (everyone know the chance, and payoff of their decision before hand.. Which never happen in real life)
Still a nice tool to have.