I’ve started to look through the data a bit; Here’s one of the basic calculations I ran.
I’m working with ~3500 days of data x 24 hours, giving me pretty good sample size.
As it turns out, over 50% of bars are inside move bars. However remember that these bars are only as such in respect to how I categorize them. An inside bar does not make a new high or new low, taking into account the current days range. What’s also interesting is how cell shaped this data is. There is very obvious clustering in the mid levels, and spread out the ends.
Originally, I thought I would have to flip the percentages in order to obtain a “true retracement” percentage. See, if price moved from 0 to 10 and then to 8, I would calculate the 8 as 80% [ 8/(10-0), or more specifically, (8-0)/(10-0) ]. Now, the move is at the 80% level, but price retraced 20%, or 100-80%. But with the data so uniform and obvious, I don’t need to perform such a statistic; both versions will show a bell curve.
What do these results really mean? One of the conclusions I can draw is that price isn’t floating. Price doesn’t make new highs or lows, and then sit in range forever. If I want good odds, I can’t favor, say the 40-60% range over the 30-40% + 60-70% range. (40-60% covers 28%, the latter is still covering 26%, only a 2% difference) I can only really favor the entire 20-80% over the areas outside of that (0-20 and 80-100). While it may not give great probability areas to trade, I would like to think it gives areas not to trade. If price is bullish and makes a 10 or 20% retracement, odds are favoring further retracement before a new move to make a high.
What’s next is to think about what could possibly be happening at these levels. For example, how often at retracements back to the 40-60% range leading to breakouts, and how often are 80-99% retracements leading to a breakout on the other side?