I like the video! Also I wanted to say sometimes when I need a bit of motivation I watch your videos for that too, not just for specific ideas. It's an energy boost without all the sugar and caffeine. Thanks for doing what you do.
Cheers Kevin, wishing you well with your current situation. Appreciate these kind of videos, it was what made me buy your books. The mass testing to see what actually works. It makes sense to start with ideas what work.
So you can use the > or < for entry instead of crosses over/crosses under to get only one trade for each long and short signal. If totaltrades=0 or (marketposition(1)=-1 and marketposition(0)=0) or (marketposition(0)=-1) If totaltrades=0 or (marketposition(1)=1 and marketposition(0)=0) or (marketposition(0)=1)
@@AlgoTradingWithKevinDavey Yes, I know. One signal = one trade in signals direction. Needed to do this when I wanted to add multiple indicators to scalping system. After the scalp profit I don't want it entering again. It is also useful for systems that trigger too much. And if you want limit the number of signals per trend trade you can and another or marketposition(1)=1 for longs.
I'm starting to get the sense that no matter what simple strategy you throw at the entire set of markets, there always be a subset of 10 or so ticker symbols and timeframes for which that particular strategy might have worked. 8000 symbols x 15 bar sizes there are 120k combinations to probe with the given strategy....and it's no surprise that out of 120k there are 10 symbols&bars that worked. It seems that it will always be that way. "Even a broken clock is right twice a day"
Thanks for the very insightful comment. I am only testing about 300 combinations, not 120K. Even so, your point is very valid. As I allude to in the video, there are other steps to take to prevent the "clock right twice a day" issue.
Hi Kevin! Here's your answer: Yes, I very much appreciate you testing different algos and giving us your thoughts on them. Your work inspires me in my own search for something worth adding to my portfolio. Keep 'em coming. 🙂
I was literally optimizing your Keltner Channel strategy when I got your email. So, yes, you should keep providing awesome content! I appreciate how you have always made yourself available to, whether it's answering emails or TH-cam comments.
Good Day, yes please. It is helpful for a beginner algo trader like me. Knowing what you know and if you were like me. What would you do with only $10000 to invest. What kind of algo would you use? I like Futures like MCL, MYM, MES, MNQ. Thanks,
Using micros is good, although 3 of your 4 are stock indices, and will be highly correlated. There are also some low margin full size, like corn, wheat, lean hogs, etc. Some interest rates could work too. Margin frequently change though, so be sure to keep aware of current requirements.
Sir, when we practically follow your strategy factory process we can do walkforward test, Monte Carlo test , find correction strategy to strategy, incubation, lots of trading ideas are there , The BIG problem is coding the strategy idea ? if we learn coding for that idea and test the idea then the result is worst.again we start from scratch.again we have to code another idea until we find good strategy. We don't know how to code ? Please give any solution.
You can learn the fundamentals for free and online and then move onto Kevin's courses which are industry best. Take a look at tradestation "introduction to easylanguage".
You already know the answer, I can tell. You have to learn to code, or team up with someone who can code. I provide coding help as part of my course, but it really is an essential skill to algo trade effectively. You end up coding a lot of things that never work, so paying for a coder to do it would be expensive.
Glad you like them! If I didn't know you so well, Jason, I'd be a little creeped out by your last sentence!!! LOL! You are a great example of who a top notch Strategy Factory student is!!!
@@AlgoTradingWithKevinDavey Thank you Kevin, I was only half joking. You gave me the tools and education I needed, the other half:) I would love to think I could have succeeded on my own, I really don't think I would have. Not without your course & ongoing, unwavering support
Hi Kevin, thanks for another excellent video. I am a bit confused by the last row of the table that is shown at 2:00. For the Bollinger w/Momentum strategy, it looks like there are 18 combinations passing preliminary testing and 36 that did well in walkforward testing. Was walkforward testing carried out for Mkt + Bar Combos that didn't pass preliminary testing?
@@AlgoTradingWithKevinDavey Thanks for the reply! Is this something you typically do during strategy development, or is it just for illustrative purposes? Also, it would be fascinating to see what percentage of the successful walkforward results correspond to high potential Mkt + Bar combos in preliminary testing. Perhaps in some situations preliminary testing is discarding a significant number of promising strategies?
I liked the look of the equity curve on the JY 1440min Break Out with High Volatility so I tweaked it with separate lookbacks for long and short along with separate profit targets and came up with a "too good to be true" result. It uses stop orders marginally above and below the highs and lows and I've set the backtest to fill once order price is exceeded and I've included in slippage and commission but I'm still not convinced the result can be as good as the back test shows. Below is the code, the inputs aren't hugely different from your original code. input: lookback_L(50), lookback_S(35), profit_L(2250), profit_S(2750); if AvgTrueRange(10) > AvgTrueRange(50) then Buy next bar at highest(high, lookback_L) +.001 stop; if AvgTrueRange(10) > AvgTrueRange(50) then Sellshort next bar at lowest(low, lookback_S) -.001 stop; SetStopposition; if Marketposition < 0 then Setprofittarget(profit_L); if marketposition > 0 then Setprofittarget(profit_S);
@@AlgoTradingWithKevinDavey I tried both Look Inside Bar set to 5min and the full 1440min, it only impacted the performance by a small amount. It doesn't trade often, I guess I'll sit on it for a year in incubation to see what happens with new data.
@@garybooth3130 - I would also test with 1 minute LIBB. Are your results from in-sample optimization, or are they from walkforward testing? If the former, I would not trust the results at all.
@@AlgoTradingWithKevinDavey Originally I used 8yrs of in sample data and then tried it on 2yrs (most recent) of unseen data. Last night I added an extra 2 yrs out of sample on the front end of the timeline as well so now there is 2yrs in the beginning and 2yrs at the end out of sample and 8yrs in the middle in sample. The performance is still consistent.
I like the video! Also I wanted to say sometimes when I need a bit of motivation I watch your videos for that too, not just for specific ideas. It's an energy boost without all the sugar and caffeine. Thanks for doing what you do.
Love that!
Hey KD, keep making more of these trading videos! They are very helpful.
That's the plan!
6:30 "And I'm just myself" Champion Trader
Thanks for the comment!
Cheers Kevin, wishing you well with your current situation. Appreciate these kind of videos, it was what made me buy your books. The mass testing to see what actually works. It makes sense to start with ideas what work.
Awesome, thank you! It is a good starting point!
Thank you for these videos! Very helpful!
You are so welcome!
So you can use the > or < for entry instead of crosses over/crosses under to get only one trade for each long and short signal.
If totaltrades=0 or (marketposition(1)=-1 and marketposition(0)=0) or (marketposition(0)=-1)
If totaltrades=0 or (marketposition(1)=1 and marketposition(0)=0) or (marketposition(0)=1)
You can, but realize that marketposition(1) is not the marketposition 1 bar ago. It is the marketposition 1 trade ago.
@@AlgoTradingWithKevinDavey Yes, I know. One signal = one trade in signals direction. Needed to do this when I wanted to add multiple indicators to scalping system. After the scalp profit I don't want it entering again. It is also useful for systems that trigger too much. And if you want limit the number of signals per trend trade you can and another or marketposition(1)=1 for longs.
@@HateDietPepsi understood, hope it continues working for you!
I'm starting to get the sense that no matter what simple strategy you throw at the entire set of markets, there always be a subset of 10 or so ticker symbols and timeframes for which that particular strategy might have worked.
8000 symbols x 15 bar sizes there are 120k combinations to probe with the given strategy....and it's no surprise that out of 120k there are 10 symbols&bars that worked. It seems that it will always be that way. "Even a broken clock is right twice a day"
Thanks for the very insightful comment. I am only testing about 300 combinations, not 120K. Even so, your point is very valid. As I allude to in the video, there are other steps to take to prevent the "clock right twice a day" issue.
Hi Kevin! Here's your answer: Yes, I very much appreciate you testing different algos and giving us your thoughts on them. Your work inspires me in my own search for something worth adding to my portfolio. Keep 'em coming. 🙂
Thanks for sharing!
I was literally optimizing your Keltner Channel strategy when I got your email. So, yes, you should keep providing awesome content! I appreciate how you have always made yourself available to, whether it's answering emails or TH-cam comments.
Awesome, thank you!
Go blue!
Just don't be stealing signals!!!!
Thank you! I love these videos!
Glad you like them!
Good Day, yes please. It is helpful for a beginner algo trader like me. Knowing what you know and if you were like me. What would you do with only $10000 to invest. What kind of algo would you use? I like Futures like MCL, MYM, MES, MNQ. Thanks,
Using micros is good, although 3 of your 4 are stock indices, and will be highly correlated. There are also some low margin full size, like corn, wheat, lean hogs, etc. Some interest rates could work too. Margin frequently change though, so be sure to keep aware of current requirements.
Kevin, can you show some portfolio equity curves of a combination of your best tested, non-correlated strategies please?
I will put that on the list!
Hi Kevin, I appreciate your videos very much and they are very inspiring for me. Please keep providing that helpful content!
Glad you like them!
Sir, when we practically follow your strategy factory process we can do walkforward test, Monte Carlo test , find correction strategy to strategy, incubation, lots of trading ideas are there , The BIG problem is coding the strategy idea ? if we learn coding for that idea and test the idea then the result is worst.again we start from scratch.again we have to code another idea until we find good strategy. We don't know how to code ? Please give any solution.
You can learn the fundamentals for free and online and then move onto Kevin's courses which are industry best. Take a look at tradestation "introduction to easylanguage".
You already know the answer, I can tell. You have to learn to code, or team up with someone who can code. I provide coding help as part of my course, but it really is an essential skill to algo trade effectively. You end up coding a lot of things that never work, so paying for a coder to do it would be expensive.
Good observation Mal!
Thank you Kevin, I always learn something new from your videos and all your writings. You complete me:)
Glad you like them! If I didn't know you so well, Jason, I'd be a little creeped out by your last sentence!!! LOL! You are a great example of who a top notch Strategy Factory student is!!!
@@AlgoTradingWithKevinDavey Thank you Kevin, I was only half joking. You gave me the tools and education I needed, the other half:) I would love to think I could have succeeded on my own, I really don't think I would have. Not without your course & ongoing, unwavering support
@@jasondesmul - You have done a great job with the material!!! Keep up the great work!!
Hi Kevin, thanks for another excellent video. I am a bit confused by the last row of the table that is shown at 2:00. For the Bollinger w/Momentum strategy, it looks like there are 18 combinations passing preliminary testing and 36 that did well in walkforward testing. Was walkforward testing carried out for Mkt + Bar Combos that didn't pass preliminary testing?
Yes, I ran the walkforward on everything, not just Prelim tests.
@@AlgoTradingWithKevinDavey Thanks for the reply! Is this something you typically do during strategy development, or is it just for illustrative purposes? Also, it would be fascinating to see what percentage of the successful walkforward results correspond to high potential Mkt + Bar combos in preliminary testing. Perhaps in some situations preliminary testing is discarding a significant number of promising strategies?
@@brianshoemaker112 Thanks, I discuss these details and more in my paid workshop.
@@brianshoemaker112 - A lot of this is what I do, some of it is primarily for video. The workshop has many more things in it.
Clever video.
Thanks!
I liked the look of the equity curve on the JY 1440min Break Out with High Volatility so I tweaked it with separate lookbacks for long and short along with separate profit targets and came up with a "too good to be true" result. It uses stop orders marginally above and below the highs and lows and I've set the backtest to fill once order price is exceeded and I've included in slippage and commission but I'm still not convinced the result can be as good as the back test shows. Below is the code, the inputs aren't hugely different from your original code.
input: lookback_L(50), lookback_S(35), profit_L(2250), profit_S(2750);
if AvgTrueRange(10) > AvgTrueRange(50) then Buy next bar at highest(high, lookback_L) +.001 stop;
if AvgTrueRange(10) > AvgTrueRange(50) then Sellshort next bar at lowest(low, lookback_S) -.001 stop;
SetStopposition;
if Marketposition < 0 then Setprofittarget(profit_L);
if marketposition > 0 then Setprofittarget(profit_S);
Are you using Look Inside Bar Backtesting? You might be folling the strategy backtester with small stops and/or targets.
@@AlgoTradingWithKevinDavey I tried both Look Inside Bar set to 5min and the full 1440min, it only impacted the performance by a small amount. It doesn't trade often, I guess I'll sit on it for a year in incubation to see what happens with new data.
@@garybooth3130 - I would also test with 1 minute LIBB. Are your results from in-sample optimization, or are they from walkforward testing? If the former, I would not trust the results at all.
@@AlgoTradingWithKevinDavey Originally I used 8yrs of in sample data and then tried it on 2yrs (most recent) of unseen data. Last night I added an extra 2 yrs out of sample on the front end of the timeline as well so now there is 2yrs in the beginning and 2yrs at the end out of sample and 8yrs in the middle in sample. The performance is still consistent.
@@garybooth3130 I would think some live testing would be in order at this point. Good Luck!