There are a couple of flaws in that write up. The write up mentions that the Cohen SAI Index measures Oct snow advance 'north' of 60N in Eurasia. As stated in the Cohen paper, it should be 'south' or 'equatorward' of 60N. In spite of that, the model he shows does indicate above normal precip south of 60N. Secondly, he states that the SAI did not perform well last winter, which is not true. The SAI correlation is used to predict the AO, not how cold it will be in a given location. Last October, the SAI indicated that the wintertime AO would be positive, which it was...particularly if you include the month of March, which Cohen has been including in the correlation in more recent years (his initial paper linked above for the SAI used the Dec-Feb period for the correlation).