LG,
OK....I'm totally guilty of a completely botched analysis of the probability of the three possibilities....ggg
I had a good instinct that the "flat" possibility was less probable, but unfortunately did not have my thinking cap on when I counted up those Japanese Candlesticks. My data was erroneous in that a black or white or doji stick does not necessarily imply up, down or flat. The candlestick is based on the open, range and close which does not correlate with up or down days. So, apologies and I gave myself a mulligan since no one caught me...GGG My do-over is from the compx and considers the whole number close from one day to the next over the last 100 days. +/- 20 points I considered "flat", the rest up or down. I thought this was reasonable as this encompassed a 40 point range surrounding the previous close.
(FWIW, changing to +/- 25-30 points did not affect the results...)
The corrected data gives:
Up: 38
Down: 41
Flat: 21
So...a similar result but definitely showing a higher probability for flat than I originally suggested. Interesting was that the flat days tended to cluster together and were often near trend turning points....(ST trend)
Statistics have always fascinated me and when I saw your response to the individual predicting the flat/down day, I was inspired to check the guesses against the data of recent Nasdaq history.
In considering this reasoning, I was led to a more interesting idea which has to do with my prediction of 70% probability of a down day on Monday. Namely, my work with and the following of the Japanese Candlesticks. Since Friday's candle was a Black Marubozu, and being conscious of this fact, I went back and checked the same last 100 days of compx data to see how well this stick predicted a down day (more than 20 points) for the following session. If the shadows of the black sticks were in excess of roughly 10 points, it was not considered by me as a Black Marubozu. Very interesting was that during the time frame considered, there were 17 Black Marubozus. The following session yielded: (using the same +/-20 point criteria)
11 down : 65%
4 up : 23%
2 flat : 12%
Amazingly, my 70% expectation of a down day for Monday is quite close to the actual statistical probability for the day after the candlestick printed Friday and that the flat possibility after a Black Marubozu is only half the probability otherwise over that same span of time.
As the art and science of Japanese Candlesticks is quite ancient and prescient (IMO), I consider these observations valid and look forward to Monday's actual result now to see how well Friday's action predicts Monday's result.
Thanks for the opportunity to correct my flawed analysis and your generosity not blowing my error out of the water. Some interesting and original thought I think quantifying Japanese Candlestick analysis statistically, even though I think a few hundred more days of data would have been more convincing!
Best regards,
Eichler
I'll stick with 70% probability of down on Monday even though statistically, I can only justify 65%! |
