For any others still interested and have also been trying to keep their data in Price/Time Tables. I have posted above in this thread(as forrestang) following MANY attempts to organize my price data in a way to reference it. A post by another member a while ago in another related thread: My previous attempts, I had done this via collecting data in 'swings' organized by my very own ZZ indicators created. Recently, I've taken a step back, and organized my data simply in price bars(currently 15 min, although I can do any price bar type/size I want). The quote above about creating a historical database. The steps listed are: 1) Obtain the data 2) Create a way of sorting and organizing it logically 3) Then you can compare like with like I think I have something closer than I have had to accomplish the first 2 on this list. The third is the issue. All data is relative to some other data point. That could be the current days RTH High/Low/Open, or the current day's ETH High,Low,Open. Or similarly the prior day's values(OHLC values). I would think the goal is to be able to ask questions of your data as you are watching the current day develop. Likely typical price behavior. The example given was: Unpacking this question... Your desired prediction windows is the last 30 minutes of the day. As all of your tabular data is time-stamped, you simply want to look at data 30 minutes prior to 16:15EST. It is easy enough to query your data for this time window. Location is "X from a high." As I understand this could be either the HOD of the current Day's ETH or RTH session. It could also be some other high(weekly high, monthly high, 52 week high)? But also, relatively easy to query your data for this. Also on Location, mentioned is "volatility dependent." This could mean that this measurement in absolute tics or points won't get you the comparisons you need? All of my data is in absolute references. Volatility dependent makes me thing this distance needs to be measured in some measurement of the current volatility. This could be measured MANY different ways. Has anyone given any thoughts to your measurement means? Maybe a 30 day Daily Range, and the distance relative to such a metric? "Down on the Day," also sort of location. But this seems to me like a binary answer. In that we are either UP or DOwn on the day, likely relative to either the ETH Open or the RTH Open. "Other supporting conditions are in force (to do with both the prior action say from lunchtime EST onwards and the location in the bigger cycle)," this is the other thing that could mean quite a few things. Just addressing the 'location in the bigger cycle,' I am unsure how to ask this question of my data. This could mean days, weeks, months, relative again to the Highest point in a 52 week cycle, just yesterday, or possibly many other things. Has anyone given any thought on how to query your data for a 'location in a larger cycle?' Also just to give a general idea of how my current database looks. All the values aren't in the pic, as it's quite a few columns. BUt I've added quite a few metrics on there just to be able to sort the data in a few different ways. Things like bar number as they were pulled from a chart, session number, dates, times, datesTimes, DayofWeek, and of course all the data relative to the current days DEVELOPING OHLC values, as well as the prior day's OHLC values. Also this time I've pretty much included BOTH the RTH and ETH metrics, as previously I was unsure on how to model this.
Curious if anyone did any more work on this type of trading model and arrived at any interesting conclusions/applications?
For what it's worth I did an incredible amount of work on this with the end result being a custom made statistical application which I use to attempt to predict the trading day ahead. I have no idea if this is a similar model to the one suggested by @Cheese. I've made some interesting discoveries which in turn have given me a good understanding of how the market generally moves and the basic probabilities for any given day. No doubt this is useful. Some of the predictions I've made have been remarkable and once in a while I'm able to make a forecast which according to my model have a very, very high probability. Here's one example posted in real time pre-market. ES Journal - 2021/2022 Picture attached. But, most of the time I'm not able to make predictions with such high probability. Or even if I'm able to make a prediction that turns out right, such as a positive/negative close by EOD - it's still a bit too rough to be used for accurate trading. Meaning I would have to use fairly large stops to monetize it if I were to trade it mechanically as the market could move greatly against me over ther course of a trading session. Of course, if the leverage utilized is low, that's perfectly possible to do. This have led me to believe that either I have unrealistic expectations about what's reasonable to predict with any kind of accuracy OR simply that I have not put together the correct parameters or asked the right questions in order to make such predictions. Put differently, maybe if I asked the right questions or did the right query with the data I have at hand, I could more consistently be able to make predictions as accurate as the one shown. I'm not sure if there's anyone present at the moment who's doing anything similar and if anyone made any worthwhile discoveries I'm sure they're not going to share it. But would be interesting to know what's possible and not...