How would one go about assigning a value to how "trendy" a day was? Take today for instance, I'd rate it fairly high as a trend day - we basically had one trend up for most of the day, then a trend down. Some days we trend all day, those would get the higher score. Most days (lately) the market sets it's highs and lows pretty early then respects them for the rest of the day. I'm curious because it might be interesting to plot "trendiness" and see if it is increasing (as I suspect it is). Perhaps summer spells lower trendiness and winter higher.
I can tell by my p/l If it is up it is trendy, if it is down, it is not. My basic trade is breakouts. Whatever the direction. Breaks of pivot points. When they keep falling back into the range and chopping me to pieces, it is a non trendy day. Actually, what I would say is non followthrough day or non volatile. I need follow through to make money. Every thing else junks me up.
For day trading, Toby Crabel once mentioned: "... Look at trendiness based on an expanding day's range with a close in the bottom 10 percent of the day's range and an open on the top 10 percent of the day's range, and vice versa."
That would only work for the days where there is one prevailing trend the whole day (a long red or white candle), not like today
I agree. Of much more use has been a record I've been keeping of the 10d ADR since SnoSur4 first posted his OBR strategy last year. For one thing, it puts the lie to this NR4/NR7/NR9 business. Assuming that one is looking for trending because he likes breakout or retracement strategies, the range itself is likely to be of more interest than whether or not the day began at one point and ended at the opposite end of the spectrum.
As if Crabel's trendiness (it might be applicable to intraday bars, I guess) is measured on daily bar basis, it's quite alright, I would think.
You could take the data in whatever bar length you use for a day and compute a linear least squares regression line. Then use the absolute slope of the line as the value for the day. The more it trended, the higher the value. This would give you a rough approximation. To get a more accurate number you could use the non-linear least squares regression line with it's slope in the same way.