Excellent article and I appreciate the openess of your approach and portfolio construction. I am on the same journey, but only 5 systems in, so great to see someone who is a couple of steps ahead. Some questions for you -
1. Did you build your systems from scratch and/or purchase them?
2. Are you fully automated?
3. How do you manage your capital allocation to each portfolio?
4. I notice the last system has a 65% drawdown - is this a protection for a black Swan event? (Have you read Laurens bensdorp book?).
1. Mixture of systems I made or found online or worked together with others to make or purchased
2. I am about 90% automated. I just have to press a few buttons
3. Capital allocation is more based on what the system can handle to get over commission drag / slippage issue. It’s pretty close to equal weight plus or minus a bit
4. It’s a hedging system. Buys VXX when the markets seem like they are crashing. I have not read his book but am familiar with his work
Hey! Currently it’s only stocks, one day I may branch out. I do not use leverage often, only times is when offensive systems shift to defensive systems and they have a little bit of time of overlap and share capital.
Excellent article! Thanks for sharing your journey. It's inspiring. I've observed that mean reversion systems tend to lose their edge over time more frequently than momentum ones. Momentum systems work for a wide set of rules and parameters, whereas mean reversion ones work under a narrower set of rules and parameters and frequently lose their edge. How do you increase the chance of the mean reversion edge to persist out of sample? How do you split your test between in sample and out of sample? What do you do when OOS test is bad?
Thank you! I definitely prefer momentum and trend. I’m not the biggest fan of MR because they are more fragile. The best way to do it is to trade the whole parameter space. I can write an article on this. And my IS may be pre 2018 and oOS from 2018 to today maybe. When OOS is bad I try to figure out why. If it’s an obvious risk oversight or something I’ll fix it (maybe a handful of stocks went to 0 in the OOS and that didn’t happen IS, so I add a stop). Other than that you have to throw it away.
1. So far so good in terms of live OOS, will continue to monitor. My systems are very long term so it will take awhile to be 100% certain live OOS matches.
2. I do miss some shorts due to not getting short shares. I need to revisit this as I do have issues with this sometimes. IBKR isnt the best broker for short locates.
Awesome article. Really cool to see the portfolio level returns vs the individual systems. Sounds a lot like Basso's all weather approach with a commitment to filling the potholes. Great stuff!
Thank you for sharing this level of details. If I may ask, do you use any type of porfolio optimization (referring to your portfolio of strategies) or simpler approaches such as equal weight or scale by strategy volatility?
Love this. I can't automate my trading strategies due to tax issues. I mostly trade one monthly strategy now. Any ideas of systems that I can trade manually without taking to much time?
Can you define the constraint a little more? Sounds like you can only get in and out of the markets pretty infrequently? If that’s the case long term trend following systems are probably your best bet. Weekly or monthly timeframes
Hi there, thank you for sharing!! Quick question: (a) what is the period you apply to optimize your parameters (eg 7 years), and how often do you re-optimize (eg every 6 months)? (b) do you use walk forward testing to limitate hindsight bias?
Excellent article and I appreciate the openess of your approach and portfolio construction. I am on the same journey, but only 5 systems in, so great to see someone who is a couple of steps ahead. Some questions for you -
1. Did you build your systems from scratch and/or purchase them?
2. Are you fully automated?
3. How do you manage your capital allocation to each portfolio?
4. I notice the last system has a 65% drawdown - is this a protection for a black Swan event? (Have you read Laurens bensdorp book?).
Thanks again Darc
1. Mixture of systems I made or found online or worked together with others to make or purchased
2. I am about 90% automated. I just have to press a few buttons
3. Capital allocation is more based on what the system can handle to get over commission drag / slippage issue. It’s pretty close to equal weight plus or minus a bit
4. It’s a hedging system. Buys VXX when the markets seem like they are crashing. I have not read his book but am familiar with his work
Hello. Interesting articles. Do you trade stocks? Or options and futures? Do you use any leverage?
Hey! Currently it’s only stocks, one day I may branch out. I do not use leverage often, only times is when offensive systems shift to defensive systems and they have a little bit of time of overlap and share capital.
Thank you for re-clarifying (while it was in article). Nice backtest!
No problem, and thanks! We pray it continues to look nice into the future too!
Excellent article! Thanks for sharing your journey. It's inspiring. I've observed that mean reversion systems tend to lose their edge over time more frequently than momentum ones. Momentum systems work for a wide set of rules and parameters, whereas mean reversion ones work under a narrower set of rules and parameters and frequently lose their edge. How do you increase the chance of the mean reversion edge to persist out of sample? How do you split your test between in sample and out of sample? What do you do when OOS test is bad?
Thank you! I definitely prefer momentum and trend. I’m not the biggest fan of MR because they are more fragile. The best way to do it is to trade the whole parameter space. I can write an article on this. And my IS may be pre 2018 and oOS from 2018 to today maybe. When OOS is bad I try to figure out why. If it’s an obvious risk oversight or something I’ll fix it (maybe a handful of stocks went to 0 in the OOS and that didn’t happen IS, so I add a stop). Other than that you have to throw it away.
Thank for the insights! Some questions from my side:
1. how is the live performance vs backtest?
2. like question 1 but for the shorts. How is your system affected for not getting short locates?
Thanks!
1. So far so good in terms of live OOS, will continue to monitor. My systems are very long term so it will take awhile to be 100% certain live OOS matches.
2. I do miss some shorts due to not getting short shares. I need to revisit this as I do have issues with this sometimes. IBKR isnt the best broker for short locates.
Awesome article. Really cool to see the portfolio level returns vs the individual systems. Sounds a lot like Basso's all weather approach with a commitment to filling the potholes. Great stuff!
Thank you! I’m a big fan of tom basso
Thank you for sharing this level of details. If I may ask, do you use any type of porfolio optimization (referring to your portfolio of strategies) or simpler approaches such as equal weight or scale by strategy volatility?
I take a simpler approach like the two you mentioned. I’m currently writing a portfolio development series of articles to answer that exact question!
Love this. I can't automate my trading strategies due to tax issues. I mostly trade one monthly strategy now. Any ideas of systems that I can trade manually without taking to much time?
Can you define the constraint a little more? Sounds like you can only get in and out of the markets pretty infrequently? If that’s the case long term trend following systems are probably your best bet. Weekly or monthly timeframes
Hi there, thank you for sharing!! Quick question: (a) what is the period you apply to optimize your parameters (eg 7 years), and how often do you re-optimize (eg every 6 months)? (b) do you use walk forward testing to limitate hindsight bias?
Great post! Like the detail.
I understand that you do bactesty in RealTest?
If I buy a subscription, will I have access to the code of those strategies you use?
I use RealTest yes. There are example strategies that come with RealTest but everything I trade are made by me
I understand, I use RealTest myself.
What I mean is, are the strategies you use yourself possible to see? If so, on what terms?
Very good work my friend!