Project: HAL-63

Last night some thunderstorms rolled in, my trading partner Lou-da-dog got scared, so my wife woke me up.
Apparently I mumbled something about "Don't worry I got it sorted and tagged" I was dreaming about spreadsheets.

Anyhow when working at this scale, I constantly have to validate what I'm doing. The chart below is the next week's performance after the Action Word and Conviction Status. Action word (as it implies tells you what to do, but conviction is equally as important.) It is a mathematically weighted occurrence, how often it happens is very important.

Each of the 13 Pattern groups and their signals are segregated, no bleed-over, so we get 13 signals and one total-weighted signal for the entire database. The returns below are for the the week after the signals.

Screenshot_2025-06-05_18-52-01.png
 
Last night some thunderstorms rolled in, my trading partner Lou-da-dog got scared, so my wife woke me up.
Apparently I mumbled something about "Don't worry I got it sorted and tagged" I was dreaming about spreadsheets.

Anyhow when working at this scale, I constantly have to validate what I'm doing. The chart below is the next week's performance after the Action Word and Conviction Status. Action word (as it implies tells you what to do, but conviction is equally as important.) It is a mathematically weighted occurrence, how often it happens is very important.

Each of the 13 Pattern groups and their signals are segregated, no bleed-over, so we get 13 signals and one total-weighted signal for the entire database. The returns below are for the the week after the signals.

View attachment 69368
This looks awesome JTH, but I need some clarification of what is triggering the Action. Is it the weekly return? What is each Pattern group responding to?
 
This looks awesome JTH, but I need some clarification of what is triggering the Action. Is it the weekly return? What is each Pattern group responding to?

The weekly return is what transpired the week after the action, really just a back-test to confirm the trigger words and conviction levels are working.

So if you see the word Sell, the stats are what happened the following week after (or with that chart it was the average of 3272 weeks). Right now it's a read-only engine, so the action is information for the user. The action is a 5-scaled word derived from the trade score. Since each week gives 13-trade scores, I created the action words for a quick read.

The trade score is measured against the delta, each pattern compared within its group, which was ran though an AI optimizer that analyzes each pattern group against the historical delta ΔAvg+, ΔAvg, ΔAvg–, and ΔWin Ratios. Those 4 delta per-group are assigned weights & optimized to get the trade score as close to a correlation of 1. AI claims it's called "Correlation-Optimized Weighted Scoring Model"

Of course we know this is "fitting the system" but with a rolling 63-years, when a new week is added, the oldest is removed. For now that's the weekly setup, here's what it generated over the most recent 26 weeks.

As an example, row 10 was a High Conviction Strong Buy, the following week the index earned 5.7%. So that basically tells the user the 13 patterns performed often (with high conviction) and well (with good Delta). This is not perfect, I still need to do a roll-forward blind back-test, but for now I'm just trying to get a working copy before moving over to the daily.

Screenshot_2025-06-06_01-48-15.png
 
Today's battle appears successfully won.

A Tradingview pine script alert sends the closing bar's data to a webhook URL, which then relays this to google sheets app script, which then routes it to its assigned column and row. For example, when a 30-minute bar alert is received, the script routes that data to start at row 21 in column A. Alerts for other timeframes like daily, weekly, or monthly bars are directed to their own designated areas in the sheet.

The rows above at 3-19 collect the most recent bar's data, which I can then route as needed. This will be a much needed addition, my normal routine is to download the .csv for every ticker and timeframe, then paste it into the sheet.

Screenshot_2025-06-06_18-11-39.png
 
Back
Top