|
TOP 6 DATES TO PICK  1. May 92. April 283. April 234. April 275. May 106. May 8 |
TOP 6 DATES TO AVOID 1. May 12. April 303. April 204. April 185. May 66. May 11 | ||||||||||||||||||||||||||||||||||
|
The easiest technique to choose an ice-out date is to open a calendar to April/May and throw a dart. However, if you want to improve your chances of winning you might review the ice-out facts and throw a dart at a smaller number of dates. Or you can review the same ice-out data and look at the dates to pick/avoid (shown above) and choose a date based on the theory that the actual ice-out percentages will, over time, match the normal (Bell) curve. This is happening with each passing year but it's certainly not foolproof for any given year, although the above dates improve your chances over the dart technique. Are you wondering why May 8, 9 and 10th are in the top picks? Answer: Over 10% of the ice-outs have occurred after May 10th. There is only 1 ice-out date for May 8,9,10th combined, around 1%. This creates a hole in the distribution. There should be 3-4 "ice-outs" on these dates to fill in the hole and match the Bell curve shape. | |||||||||||||||||||||||||||||||||||
|
Last year we looked at the impact of the cold weather months on the ice-out. We did this by calculating the multiple correlation coefficient (R). A correlation coefficient (R) of 1.0 is when the regression perfectly fits the data. All the data deviations are accounted for by the regression. This year we've added the warmer months of March and April and adjusted the R-values based on 2007 data. The high temps in December and the snowfall in February are insignificant factors in determining the ice out day. We could drop these data elements without sacrificing accuracy. Using regression, the cold weather months contribute to predicting ice out, but it's the weather in March and April that correlates best with the ice out day. When the weather data for the months of March and April are combined, the R-value is .755 compared to a combined R-value of .467 for January/February. A 62% difference. The .755 value means that 75.5% of the deviations in the ice out are accounted for by applying the weather in March/April and only 46.7% are accounted for by applying January/February weather. Even the temperature lows in March/April have a higher correlation to the ice out day than the temperatures and snow fall in January/February. It's a myth to look out the window after several days of -20 degree weather and say, "the ice out day will be late this year". The weather in March/April will have more to say about the ice out day than the cold you are experiencing. The combined R-values shown above provide the odds of predicting the ice out day on a month-by-month basis. By combining the weather for December, January, February, etc. your odds improve. When the above weather conditions are all applied the R-value equals .827, meaning that 82.7% of the deviations in the ice out day are accounted for by the above weather conditions. By using the equation generated by the regression, you should be able to pick the ice out day 8 out of 10 times. The 'Iceman' has always been a little leery of some of the data that is available prior to 1960. So before I provide you with the equations for predicting the ice out day, I'd like to run the same analyses on the data from 1960 and on. We'll do this next year and compare conclusions. The deviations in the data may be less because of better methodologies in capturing the weather conditions. Thus the R-values and equations may be more accurate, giving you better odds to predict the ice out day. Until next year... © Iceman - January, 2008 | |||||||||||||||||||||||||||||||||||