<html> <head> <style></style></head> <body class='hmmessage'><div dir='ltr'>THIS MESSAGE IS FOR THE PERSONAL USE OF THE READERS OF THIS LISTSERV AND IS NOT TO BE REPRODUCED FOR ANY OTHER PURPOSE, INCLUDING PUBLICATION IN THE SPUT.   I recently heard speculation that Old Faithful is not as "regular" as it used to be.   Using data from geyser times for the last 99 intervals (one of the intervals in the most recent 100 intervals was a double interval so I had to exclude it), I calculated minimum, mean, maximum, and coefficient of variation for "short" intervals and "long" intervals.  The following table shows a comparison with the last calculation that Ralph Taylor posted on geyserstudy.org for the week ended August 17, 2011:                              Geysertimes                Taylor Long:     count                      95                             104     minimum              1:19                           1:15     mean                     1:36                           1:34     maximum             1:57                           1:50      coefficient of       variation            7.81%                         10.46%   Shorts:     count                       4                                   5     minimum               0:52                           0:54     mean                      1:01                           1:00     maximum              1:06                           1:09          coefficient of        variation             8.86%                       11.78%   A comparison of the results shows the percentage of "short" and "long" eruptions is about the same for both the 2011 and 2016 data sets.  The sample sizes for the "shorts" is small but there are no striking differences in minimum, mean, and maximum.  The minimum for longs is 4 minutes higher for the geysertimes ("current") data, mean is 2 minutes longer, and maximum is 7 minutes longer.  The coefficient of variation is smaller, indicating that the "longs" have less variability in the current data set than they did in the 2011 data set.  But mean for longs for the "month" for Taylor's data set in 2011 was 95 minutes, which is only one minute different from the "current" mean of 96 minutes.   Assumptions: The OFVEC staff correctly identified the short duration eruptions such that predictions for the short intervals were within the window.  The OFVEC staff is using 94 minutes for the prediction following a long duration eruption.  I don't know if the staff is using a third prediction for "medium" duration eruptions.   For the 95 "long" intervals, 3 were less than 84 minutes so would occur before the prediction window of 94 minutes +/- 10 minutes.  There were 16 intervals greater than 104 minutes, which meant the eruption occurred outside the far end of the window.  Thus, 19 of the 95 eruptions, or 20% of the prediction windows did not contain the actual time of the eruption.  Moving the prediction to 96 minutes would have captured 4 additional eruptions, resulting in 16% of the eruptions occurring ouside the prediction window.  There was no 20 minute window that would capture 90% of the eruptions, which was the goal when I was a geothermal volunteer doing Old Faithful predictions.   Lynn Stephens </div></body> </html>