I know, how to evaluate the success of the prediction against the random probability. I use the Roger Hunter´s probability estimator and some of my predictions had random probability less than 5%. For example, the newest prediction of Kurily - Kamchatka EQ of M7+ had a random probability only 13% . I used rectangle area 140 - 210, 30 - 90, M7 - 9 and 48 days time window. Try to predict such event. I predicted only strongest EQs, because I can not observe the small stress waves, which precede the small earthquakes. I am able to recognise the stress waves longer than 3 days, which implies the possible magnitude M6 from the area 3000 km far from my stations od M7 from Eurasian lithosphere plate. Sometimes I can see the stress waves from other lithosphere plates, for example I saw the Baja California EQ in 2010.