That's why according to profitable traders an edge exists, and that's why according to new traders an edge does not exist. And they are both right (for their personal situation).
Edge edge edge , what is it ?
- Thread starter TurtleLearner
- Start date
That's why according to profitable traders an edge exists, and that's why according to new traders an edge does not exist. And they are both right (for their personal situation).
Here is how I determine if an entry signal has an edge. This process is mostly automated so it is super easy to do.
1. Run Machine Learning routine on in sample data where strategies generated have a limit order and a stop order that are 2 x ATR(14) away from the point of entry.
2. Filter strategy results from step 1 that have a 75% win rate, thus beating random entry.
3. Run the filtered strategies on out of sample data with the same exit criteria.
4. Filter strategy results from step 3 with the same filter criteria outlined in step 2.
5. The strategies remaining are worthy of further investigation.
The whole process can test about 500,000 strategies in 20 minutes running on my 5 year old laptop. I have a similar process for determining exit criteria.
1. Run Machine Learning routine on in sample data where strategies generated have a limit order and a stop order that are 2 x ATR(14) away from the point of entry.
2. Filter strategy results from step 1 that have a 75% win rate, thus beating random entry.
3. Run the filtered strategies on out of sample data with the same exit criteria.
4. Filter strategy results from step 3 with the same filter criteria outlined in step 2.
5. The strategies remaining are worthy of further investigation.
The whole process can test about 500,000 strategies in 20 minutes running on my 5 year old laptop. I have a similar process for determining exit criteria.
Just curious. Have you tested the Turtles trading system of buying new highs or is that 20 day highs? Or Trend following for that matter? I would think both systems work but, no way to test it myself.
Impressive, but you must only be testing this on either one instrument (like SPY or ES) AND OR your data set must be very small??? How many instruments and how many years are you testing across?
How did you come up with "2 x ATR(14)" as your stop? Let me save you hours of research, anything that has a Long Only Bias and a really really wide stop that trades a stock market index like SPY will show a profit. Simply doing buy and hold would show a profit as well.
The real test (assuming you are trying this out on SPY or ES) would be to see if your system can produce profits by going Short.
Sorry, how is it super easy when you still have to use human judgement to investigate what is leftover from your filtering criteria. How do you know that you have not over fitted your strategies
Meaningful edge for a trader is just how to make good profit within some reasonable set time and within some reasonable set risk parameters. For a trader, it doesn't really matter if you beat your "competition" or not, because if you are happy with what you get, why should you care, are you a retard??
If you trade on insider info, are you a retard criminal??!
So to get any respect, you seek some technical and/or fundamental legal way to accomodate the above, and you're all set. That is enough "edge". If it's not, you need to examine yourself in some way. Because if the latter, something is seriously BROKEN.
If you trade on insider info, are you a retard criminal??!
So to get any respect, you seek some technical and/or fundamental legal way to accomodate the above, and you're all set. That is enough "edge". If it's not, you need to examine yourself in some way. Because if the latter, something is seriously BROKEN.
The Turtles were mainly trading 55 day breakouts.
I think the initial stop they used was 2 ATR.
My guess is both those numbers have to be larger these days as the markets no longer trend as cleanly as they did in the 1980s and prior decades.
There was also a 20 day breakout system, but that wasn't traded by all the turtles and it had a filter that skipped some signals. The 55 day breakout was the more simpler (i.e unfiltered) setup traded by all the turtles.
Last edited:
I said the process is easy to do. Writing the platform to do this was certainly not easy to do. In and out of sample data is used for testing to minimize curve fitting.
What are the complete rules of the system?
Currently I use all data from SPY for signal/strategy formulation and test each strategy against SPY, QQQ and EEM where all the data is loaded in memory. Out of the hundreds of thousands that might be tested, there might only be a couple of hundred that go on to the next phase.
2 x ATR(14) plus or minus from entry for stop and limit seem to be a good indication of whether the strategy has any predictive value. But yes, you are correct, stock related instruments have an upward bias so it is much easier to find Long strategies with stocks. The question is can you outperform the market with the less volatility. That is my goal.