Scoring Method

We cannot appraise extra long-rang daily forecast by standard deviation (S.D.). For example, predicts heavy rain today, but no rain now and there are storm the second day; it predicts descent of day’s temperature, but real ascent, the cool air should come the second day. Thus deferent between predict event and real weather in fact for a day, but weather process have been predict, that it should is very good. If as per S.D. it is very bad. So I have to choose a scoring method of appraise of predict. My method to predict is testing, many parameters is indeterminate. In order that predict is reliable and stable, I must do a great deal of tests, in other word, I must do predict in much times. Of course, I do all those trial with computer, and appraise of predict not by person one after one. In order to compare of different predict, it require a standard of scoring deal by computer too. Thus making of scoring method is difficult. In case of predict arise or decline of temperature and amplitude variation correctly, but the difference between predict and real temperature is 3----5℃, good or bad? If predict value always equal to average in all years, the S.D. may less, but not any fluctuate, obviously insufficient. The best scoring method is my heavy demand. The scoring standard is better, the forecast accuracy. The standard in using is rough-wrought, really bad. Now I announce what I used in order to get some good advice, especially from tactful forecaster’s experience and advice.

Scoring of Temperature predict

①. Real ascent in two days, each day’s ascent is over 1℃:

add 100 score if it predicts first day ascent over 1℃,

subtract 100 score when descent over 1℃,

add 100 score again if it predicts first day ascent over 3℃,

subtract 100 score again when descent over 3℃,

add 100 score again if it predicts first day ascent over 8℃,

subtract 100 score again when descent over 8℃,

①. Real descent in two days, descent in each day is over 1℃:

add 100 score if it predicts first day descent over 1℃,

subtract 100 score when ascent over 1℃,

add 100 score again if it predicts first day descent over 3℃,

subtract 100 score again when ascent over 3℃,

add 100 score again if it predicts first day descent over 8℃,

subtract 100 score again when ascent over 8℃,

③. The subtraction between sum of predicted value in 3 days and sum of real temperature in 3 days:

add 20 score if the subtraction is less 1℃,

it means the average temperature is less 0.3℃

add 20 score if the subtraction is less 3℃,

it means the average temperature is less 1℃

add 20 score if the subtraction is less 9℃,

it means the average temperature is less 3℃

add 20 score if the subtraction is less 15℃,

it means the average temperature is less 5℃

add 20 score if the subtraction is less 24℃,

it means the average temperature is less 8℃

Scoring of precipitation predict

①．It predicts rain in the day, cumulative sum of precipitation in 3 days before and after:

add 10 score if the sum of precipitation over 0 and

add 10 score again if the predicted value and sum precipitation increase 1mm

subtract 10 score if the sum precipitation equal 0 and

subtract 10 score again if the predicted value increase 1mm

②．There is real rain in the day, cumulative sum of predict value in 3 days before and after:

add 10 score if the sum of predicted value is over 0 and

add 10 score again if the predicted value and sum precipitation per increase 1mm

subtract 10 score if the sum of predicted value equal 0 and

subtract 10 score again if the real precipitation per increase 1mm