How wider do you think is SPY's open range vs the mid-day range?

[MarkBrown]

For example, if the first hour has a High/Low range of 10 points, by what percent was that range actually expanded in the other two time frames?

percentages are one of the hardest things to program on the fly because you have to keep track of all the highs and lows. then when a higher high or lower low occurs it takes the place of the previous high or low.

many people short cut and use the function highest(h,lookback) and lowest (l, lookback) but that will either restrict or over shoot your goal and or induce delay into what your trying to do. where lookback is the number of periods to search for a highest high or lowest low.

however what you are asking is the most valid way to build a significant model that will hold up with time because it would use true percentages and not a restricted static lookback period.

 

[tommcginnis]

My initial result was 17%, hence I put the lowest answer in the poll as 20%. But then I found an issue with my calculation and found out its actually only 6%! I wrote an article about it, it contains the data used and the spreadsheet with the stats
http://www.beathft.com/?p=146

If anyone finds a mistake or has any comments, I'm all ears

Nicely done study, and a very fun read.
There were lots of points I had 'responses' to your statements, but then you resolved them all by the end. 'Engaging!' Nice ol'

 

[tommcginnis]

i sort of dont see Midday as 4.5 hours and as such given its 4.5hr vs 1hr i think way too much skew in stats . i think the pertinent thing would be to break it down to closer time periods or just compare lets say 15m bar ranges over entire day , now this data i found on net here is from early 2018 and i appears to be a clearer view of range distribution over day . I cant verify the accuracy of these stats but from my observations over 20 years trading it looks good to me . I could write my own code to verify it but it is just not a priority for me .

View attachment 192354

Along that line, I'd love to see a study on 'continued opening trend' days -- where, after a first-hour climb[fall] of x-points, how does the market behave[close] after that? Usually, the bigger the pop[drop], the more likely a continuation. But I'd love to see empirical evidence. (Especially since Nov. 2016, eh. )

 

[PistolPete]

Along that line, I'd love to see a study on 'continued opening trend' days -- where, after a first-hour climb[fall] of x-points, how does the market behave[close] after that? Usually, the bigger the pop[drop], the more likely a continuation. But I'd love to see empirical evidence. (Especially since Nov. 2016, eh. )

OK wrote a box around AM , midday and PM boxes . Got a daily cash range measure and box on total cash range , Will do the range measures for the 3 time periods as well when i have time or can even break it down into smaller parcels . Then we can do some exploration

 

[Laissez Faire]

Not exactly what you're asking, but for the ES, the 1st 90 minutes capture the entire daily range only about 7% of the time in the last 5 years or so.

 

[Laissez Faire]

And if my call for today is correct - this is one of those 7% days...

EDIT 28 minutes later: 2nd scenario being a move towards 2930.

 

[PistolPete]

And if my call for today is correct - this is one of those 7% days...

EDIT 28 minutes later: 2nd scenario being a move towards 2930.

Not today

 

[They]

Not exactly what you're asking, but for the ES, the 1st 90 minutes capture the entire daily range only about 7% of the time in the last 5 years or so.

The question then becomes, when it is not one of the days that falls into the 7% by how much does price exceed that 90min range?

 

[Laissez Faire]

Not today.

Most certainly not.

A good trader needs to plan for and anticipate multiple scenarios on any given day. Although we trade probabilities this means that the unlikely can and will happen.

I still need to close the gap between my statistical predictions and my intuition and experience as in how I observe the chart. Today, I felt certain we'd roll over, but my data (or at least how I interpreted it), said we'd most likely range sideways today.

 

[Laissez Faire]

The question then becomes, when it is not one of the days that falls into the 7% by how much does price exceed that 90min range?

The latest record in my data set (currently not updated) has a daily range of 31,25 points and the 1st 90 minutes being 7,75 points.

Thus, to answer your question for that single day, the daily range is 31,25/7,75 = 4 times bigger than the 1st 90 minutes.

If I repeat the same arithmetic for my entire data set (1130 days), I get an average of 1,8.

For this kind of data to be useful, I suggest more detailed analysis (which I have of course done).

Bottom line - It's uncommon to find both the day high and low inside the 1st 90 minutes, but it's not uncommon that a large percentage of the day range is contained inside those 1st 90.