Question on the model results (% winners and average P/L on each trade) that you report - are those based on backtest results or live operation after the models built? If the former, what kind of holdout period was used, and was the model generation iterated and the ones that performed the best over the holdout period selected? Cheers
1) we have been running the models live YTD and they have been performing very well and within the backtest expectations.
2) I have intentionally kept the signals as raw values as opposed to implementing any type of neutral position or drawdown control so that everyone can see exactly the statistical significance of macro factors.
I have been trying to walk the line of showing the statistics of the actual macro factors and this is why I have the tactical trade portfolio in the chat with a real time P&L of the highest conviction trades as well as the trade write up section.
So everything is a work in progress in terms of our own internal development as well as how we are sharing these actively.
If you have any feedback or questions please always feel free to let me know
Yeah just having a fair bit of hands on experience here I know the pitfalls of overfitting that can even apply when a proper holdout period is used, if the process is iterated too many times and the model that performs the best over the holdout period is selected - it effectively turns a holdout period into another test set.
Reason I asked is because the backtest results look suspiciously good, from memory 60%+ winners with average win being more than 2x average loss - obviously this is a license to print money if replicable live.
Having said all that, the small sample of live results you have so far matches up results wise, so there’s definitely hope!
Not saying any of this to knock the product, I’m using it and like it and about to take advantage of your yearly sub discount. But would like to know how much confidence to assign it.
If you have any additional colour about the ML process used that would be awesome
So I dont really have a way to do this without having you unsubscribe and then resubscibscribe at the newer price. Substack makes this difficult which is why I originally was hesitant because I didn't want to make loyal subscribers like yourself to go through any logistical difficulty.
So I dont really have a way to do this without having you unsubscribe and then resubscibscribe at the newer price. Substack makes this difficult which is why I originally was hesitant because I didn't want to make loyal subscribers like yourself to go through any logistical difficulty.
Shat.....I clicked on monthly instead of annual. Do I also have to unsubscribe and resub?
Yes. Dm me your email and I’ll reset it so we get you locked into the right one
Hey I am getting 1750 for the monthly fee. Please help to reset?
is it possible to switch from monthly to annual, locking-in the $100/mth annual price, if I am 1 week into my current month period? thx
Yeah if you just unsubscribe and resubscribe 👌🏻
Question on the model results (% winners and average P/L on each trade) that you report - are those based on backtest results or live operation after the models built? If the former, what kind of holdout period was used, and was the model generation iterated and the ones that performed the best over the holdout period selected? Cheers
Great questions
So several points:
1) we have been running the models live YTD and they have been performing very well and within the backtest expectations.
2) I have intentionally kept the signals as raw values as opposed to implementing any type of neutral position or drawdown control so that everyone can see exactly the statistical significance of macro factors.
I have been trying to walk the line of showing the statistics of the actual macro factors and this is why I have the tactical trade portfolio in the chat with a real time P&L of the highest conviction trades as well as the trade write up section.
So everything is a work in progress in terms of our own internal development as well as how we are sharing these actively.
If you have any feedback or questions please always feel free to let me know
Yeah just having a fair bit of hands on experience here I know the pitfalls of overfitting that can even apply when a proper holdout period is used, if the process is iterated too many times and the model that performs the best over the holdout period is selected - it effectively turns a holdout period into another test set.
Reason I asked is because the backtest results look suspiciously good, from memory 60%+ winners with average win being more than 2x average loss - obviously this is a license to print money if replicable live.
Having said all that, the small sample of live results you have so far matches up results wise, so there’s definitely hope!
Not saying any of this to knock the product, I’m using it and like it and about to take advantage of your yearly sub discount. But would like to know how much confidence to assign it.
If you have any additional colour about the ML process used that would be awesome
Is there a way for current subs to lock in the price?
If you are currently a subscriber, you already have the price locked in so you should be good to go
I am on the 1750 annual rate though. Subbed 3 months back
So I dont really have a way to do this without having you unsubscribe and then resubscibscribe at the newer price. Substack makes this difficult which is why I originally was hesitant because I didn't want to make loyal subscribers like yourself to go through any logistical difficulty.
Do you know if I unsubscribe and resubscribe, will the new charge be effective immediately or
If you do it before whenever you’re monthly charge goes through at whatever rate you’re at.
I am on the annual one though, is it etiol the same?
Curious about this as well
So I dont really have a way to do this without having you unsubscribe and then resubscibscribe at the newer price. Substack makes this difficult which is why I originally was hesitant because I didn't want to make loyal subscribers like yourself to go through any logistical difficulty.
Does early subscription mean that the current price will be locked in for life?
yes