Limiting the test to two variants removes an incredible amount of variables from the test and severely limits the ability of Phrasee's deep learning engine to determine what works and doesn't work for your audience(s).