cnb: well done. I happen to have created something similar, and what you created is very useful. Equally important, your chart will spark other good ideas. Here's an additional idea for you. Create a supplemental version by replacing your left numbered column (1-31) with a day-of-week column (M, Tu, W, Th, F). You'll see the day-of-week effect, sliced by month. Here are some small suggestions that may or may not appeal to you: Carrying the significant figures out to two decimals is too much (not useful) precision and interferes a little with the mind's pattern recognition. Consider dropping the decimals, and just round to the whole number percentage. Consider categorizing as 'neutral' anything between, say, 49-51% or anything within, say, 1 standard deviation. Maybe add true heat map color to it. Let, say, orange and red down days and green or blue be up days. Gray can be neutral days within a standard deviation or whatever. For the month slice, you can also aggregate the months into seasons to see a valid and tradeable seasonal message. Very nice.
Not sure if you understand the purpose of my inquiry. This example posses an opportunity to challenge our(my) general understanding or lack there of, for statistics by using the appropriate tests for significance. It seems a useful exercise, even for just to be able to say we can through the numbers away, but only based on reason, and not conjecture. If perhaps I am thinking through this problem in the wrong way, I challenge a response that may explain how and then to pose an alternative measure. Significant= yes or no? Explain process by which to formulate.
Perhaps I misunderstood your intentions. If so, I apologize. I just think that the premise of this "grail" is absurd at face.
I have been fooled too many times ( both type 1 and type 2 errors) during my time educating myself with rearguards to probability that I can assume any belief I hold has potential for fallacy. My education involves not only learning, but also unlearning fallacies of reasoning that I hope will help me make profitable decisions. This example posses a good exercise and hopefully something of value may come of it, perhaps other than the potential of a trading system from it. I can understand how one can become cynical with the tremendous amount of bunk that is designed to take advantage of people that don't know any better, perhaps of no fault of their own. The defense against that is a grounding in the concepts.
so anyone has tested the accuracy of this during the last month or you guys are too busy trying to figure out which moving average will cross next??
Just came across your chart. Are you using it in your trading - succesfully - ? It looks seductive and somehow not convincing at the same time imho.
I just saw this thread. I think it's a pretty cool chart. I don't know how useful, but it can't hurt to add one more item to check. I will note that late March seems fairly poor in this chart, and then early Apr seems much better. Now, with the HC bill coming, it looks like it might follow the chart. Of course, I realize the chart isn't perfect or anything - if a certain day had say 38%, maybe many of those days are still huge up days, and then downs are real minimal, and vice-versa. So, you have to take it for what it's worth. JJacksET4
Be careful... the more times you slice a dataset the more chances you give it to lie to you. These kinds of effects are very unstable. You can actually make a better case for using shorter data histories than longer. For instance, repeat your work with 1990 - 1995, then 1995 - 2000, then 2000- 2005, then 2005-2010. You have an idea that is logical on the surface, but you need to do a lot more work to see the flaws. FWIW, I think it is unlikely that you will find this approach profitable. Don't mean to rain on your parade, but far better to find flaws in research than with real money on the line.