If you read the actual research, they clearly state that they aren’t asserting that their model currently holds predictive accuracy, while also noting that it did have successes.