Hello Mav, Could you talk a bit about the heuristics of OR, A levels and the number line values with respect to using the number lines as an indicator of strength/weakness? Here's what I'm getting at: depending on what you use for an OR has an impact on A-level as does the value you use for the delta above/below the OR used to compute the A level. As the number line is calculated based on confirmations around the A-level the width of the OR and A-level deltas directly impact the number line so based on this the number line may show more or less strength/weakness for the same series of bars. I.e., a tight A-level may result in a confirmed number line (+9, +9) whereas a wider A-level may show less strength. This is obvious but how does this play into the interpretation of the number line? Thanks
Hello Mav, Would you elaborate a bit when you say that "weak trending days produce choppy spreads"? It would seem just the opposite to me especially if you a trading two divergent markets. E.g. a strongly trending long minus a strongly trending or maybe even less strongly trending short would, at least to me, seem to me to be an excellent smoothly trending spread candidate.
Take out the word weak. I just mean strong trends in either direction. Spreads tend to trade counter to the structure of flat price. When flat price is trending, spreads tend to be choppy and vice versa.
I'm a long-term trend follower. I trade a simple mechanical trend trading system that I have developed and fully automated. It is diversified across 25 markets (commodities, bonds, currencies, indexes). It requires action only once a day (to recalculate risk, modify orders and adjust stop losses). It is done fully automatically and I only receive a daily report for review and a sanity check. Average duration of a winning trade is about 60 trading days (3 months) and average duration of a losing trade is 30 days (1.5 months). Now it is no secret that long term trend following suffers from long and deep drawdowns and is very difficult to follow psychologically (who wants to trade through a drawdown that lasts 1.5 years?). Nevertheless I'm pretty confident in the approach because: 1) I have convinced myself with extensive research and testing that markets exhibit trending behavior (this is the edge of the system). 2) When developing the system I made sure there are only a few parameters (degrees of freedom) to optimize while using a lot of data (many years x many markets), to avoid overfitting. In the end it feels boring and a waste of my abilities to just sit there and not try to improve the reward/risk profile of my trading. So I have a question to the veterans of ACD: Is taking the long-term trading system as means to define the long term trend and then using shorter-term discretionary ACD to trade in and out within the longer term trend a viable way to improve the reward/risk of the original slow system? How can one answer such a question when ACD is deemed untestable?
Hi, do you mind if I ask what's your return like? Also, would appreciate the drawdowns and winning rate details. I am trying to run a long term system myself. Just wanted to have a rough idea what to expect. Thanks.
I think ACD is very testable. You have to get the number lines first but once you have that data it becomes a data analyst's dream.
Returns depend on risk per trade. My risk per trade is 2%, CAGR 20% and max drawdown 22%. That makes for MAR of 0.9. Win rate is 40%, but the result distribution is very fat tailed. If I recall correctly 20% of the profits come from top 5% of the winning trades and 50% of the profits come from top 20% of the winning trades (which is about 8% of all trades). Monthly stats: 55% of months are positive. Average winning month is 1.6 times larger than average losing month. Longest drawdown is about 18 months. These stats were estimated from the test starting from 2005, trading costs included. Arguably, the last decade was very tough for trend followers.
Determining the number for the day requires intraday data. Unfortunately I only have EOD data for the futures that go back to the 80s and even 60s for some products. I have access to the short time frames for Forex pairs though, I could give it a shot there.
Hello Mav, Would this apply to both convergent and divergent spreads? It seems to me that a divergent spread where both markets were trending would be the ideal trending spread.
Yes, but this is very rare and also in a way is not what you want. It's almost like trading flat price at that point. If they both reverse at the same time you will get killed and they only need to reverse a little bit for you to get killed. Ideally you want to track things that are highly correlated and move together. That will keep the volatility of the spread to a minimum.