Successful Results for the Experimental Earthquake Forecast of November 2018
Astrological Model: 8 EQ/12Frcst = 67% correct, (p) .03 Control Group Model: 5 EQ/12Frcst = 42% correct, (p) .22
November 30, 2018
Even though the M7.0 earthquake that occurred in Alaska today was not in my forecast, it certainly serves the excellent purpose of calling attention to the need to understand when these things are coming. The experts on the TV today are saying that they can forecast where the geologic stresses are building up and a range of years & probabilities when they might happen, but they just don’t know what day earthquake activity will occur. This, I believe, is an area where astrology can be of enormous benefit.
First, a word of disclosure and apology. About mid-way through the month, I had a severe case of ‘thinking too much’ and did a minor revision of the forecast days of the 24th & 27th. This was a mistake that I regret and rescind and pledge not to do that again. As is usually the case, one’s first instinct is usually the right one. The following analysis is of the original forecast posted on this website and NCGR’s Facebook page on or about 10/30.
This somewhat mysterious chart below is a proprietary tool that I call Zodoc and I use it and a number of other metrics to forecast extreme mundane events such severe storms, tropical cyclones and, in this case, earthquakes.
This is a list of the earthquakes from USGS converting the Z time of record to Eastern Standard Time. Each forecast window is 36 hours, centered at noon of the forecast date.
This graphic summarizes the actual EQ’s (red) compared to the expected EQ activity days (black). Of the 12 EQ’s forecast, 8 occurred in the forecast window, for an overall score of 67% and a binomial distribution probability of (p) .03, or about 3 chances in 100 tries to get a score of 67%.
Of the 20 days that no EQ was expected 15 were correct, for a score of 75%, and a binomial probability of (p) .09, or about 9 chances out of 100 of getting 75% correct.
There are 32 days (trials) in this test and 23 correct EQ or no-EQ days, for an overall score of a 72% correct forecast. Adding the (p) values for the EQ/no-EQ results (.03 +.09) gives a value of (p) .12, or about 12 chances out of 100 of getting 72% correct for the month.
We get a much different result for the Random Dates Control Group. Of the 12 EQ’s forecast, 5 occurred in the forecast window, for an overall score of 42% and a binomial distribution probability of (p) .22, or about 22 chances in 100 tries to get a score of 42%.
Of the 20 days that no EQ was expected 12 were correct, for a score of 60%, and a binomial probability of (p) .18, or about 18 chances out of 100 of getting 60% correct.
There are 32 days (trials) in this test and 17 correct EQ or no-EQ days, for an overall score of a 53% correct forecast.
Adding the (p) values for the EQ/no-EQ results (.22 +.18) gives a value of (p) .40, or about 40 chances out of 100 of getting a low overall score of 53% correct for the month.
The final result is 72% success for the astrological model, with a (p) .12
The control group got overall low score of 53% and an insignificant probability of (p) .40.
More analysis to follow, but given the general belief that earthquakes are totally unpredictable, there is some evidence here to challenge that null hypothesis.