2018 - 8 Future Basket Trading Journal

[southall]

Anyone want to share their thoughts with making minor discretionary tweaks to their trading systems / strategies for taking single day trades or 1 – 3 day swing trades? How long did it take for you to be “in tune” with an individual market to make stop loss level tweaks or swap instruments you were trading to something else, but still trading your core premise / strategy?

When i make any tweak to my systems the change has to generally improve the system's metrics (eg Sharpre/Mar ratio or profit factor) without reducing robustness. By robustness i mean the ability of the system not to break down badly under unfavourable market conditions.

Now your system, in back testing at least, is a holy grail system with a 20:1 profit to drawdown ratio (200K a year profits, 10K drawdowns) with hardly any losing months. I don't know why you would want to tweak such a system?
Also you didnt want to worry about making this system robust as your plan was to switch to a different approach if market conditions change to the point of breaking it.

 

[MidwesternTrader]

Now your system, in back testing at least, is a holy grail system with a 20:1 profit to drawdown ratio (200K a year profits, 10K drawdowns) with hardly any losing months. I don't know why you would want to tweak such a system?
Also you didnt want to worry about making this system robust as your plan was to switch to a different approach if market conditions change to the point of breaking it.

In the thread intro I tried to allude tongue-in-cheek that any back tested system that looks this good is inherently flawed in some way as no holy grail exist. Going live is the only way to expose the flaws. The journal is a method to present real life challenges of system and the trader as well.

I don't think a small tweak here and there to plug an obvious leak takes away from the core design of the system. Like KC where real life trading proved the -1.5 cent stop was unrealistic to fill consistently due to intra day volatility spikes and thin trading volumes compared to other instruments.

Its akin to building your own sailboat and finding out whether it can sail across the Atlantic. If a leak appears in the hull, you should probably try to plug it. It's still the same boat. Now if I was looking at making a major systematic change, that would be like trying to change the hull design in the middle of the ocean.

Getting feedback from the board provides nice perspective. I like throwing out the thoughts this experience is bringing to mind for the crowd to kick around. Doesn't necessarily mean I'll implement any of the feedback.

 

[kevinkdog]

In the thread intro I tried to allude tongue-in-cheek that any back tested system that looks this good is inherently flawed in some way as no holy grail exist. Going live is the only way to expose the flaws. The journal is a method to present real life challenges of system and the trader as well.

I don't think a small tweak here and there to plug an obvious leak takes away from the core design of the system. Like KC where real life trading proved the -1.5 cent stop was unrealistic to fill consistently due to intra day volatility spikes and thin trading volumes compared to other instruments.

Its akin to building your own sailboat and finding out whether it can sail across the Atlantic. If a leak appears in the hull, you should probably try to plug it. It's still the same boat. Now if I was looking at making a major systematic change, that would be like trying to change the hull design in the middle of the ocean.

Getting feedback from the board provides nice perspective. I like throwing out the thoughts this experience is bringing to mind for the crowd to kick around. Doesn't necessarily mean I'll implement any of the feedback.

Be careful. Maybe you think there is a leak in the hull, because you found water in your boat. But maybe the water in the bottom of your boat just came from a freak wave, and your hull is fine.

Hard to tell after only one month.

 

[HobbyTrading]

@MidwesternTrader Do you already take volatility into account in your trading system? You do mention three ways of defining a stop loss, but it is not clear to me when/how you decide to switch over from one to the other.
I don't do daytrading, but my automated system calculates the volatility over the most recent 25 trading days. So this volatility gets adapted every day.

 

[MidwesternTrader]

@MidwesternTrader Do you already take volatility into account in your trading system? You do mention three ways of defining a stop loss, but it is not clear to me when/how you decide to switch over from one to the other.
I don't do daytrading, but my automated system calculates the volatility over the most recent 25 trading days. So this volatility gets adapted every day.

Currently I stick the SL levels that have shown the best P & L performance over the past 2 years, except with the change I made in KC. The analysis post was hypothesizing whether I should investigate switching SL levels or not in instruments like NQ and CL as they appear to have different performance for different SLs depending on current volatility of the instrument.

Based on the feedback from others, I wouldn't think about changing the SL levels from week to week or day to day unless I can develop a back test proven method to incorporate a volatility measured adjusted SL level that improves long term P & L.

 

[HobbyTrading]

So you have three alternative solutions, and you base your choice on a 2 year back test. Why did you decide for a 2 year back test? And not for example a one year, or three year back test?

 

[MidwesternTrader]

So you have three alternative solutions, and you base your choice on a 2 year back test. Why did you decide for a 2 year back test? And not for example a one year, or three year back test?

I wanted to back test with a large enough sample size that also had recency relevance. 1 year entails about 260 trades. 2 years entails 520 trades. I felt the larger sample size of 520 would help smooth randomness.

But why not then increase sample size to 3, 5, 10 or 20 years? At some point I believe one loses relevance of today's market by using data from years too far in the past. For instance, the CL market from 2007 - 09 was vastly different than 2016 - 2017. Same can be said of NQ from 1999 - 2001. I don't see value in bringing those years into the sample set to determine how to trade today's markets.

I'm sure not everyone has the same idea for this as I and others have surely used 10 year sample sizes as effective back tests for their successful methods. But you gotta pick something to go with at some point.

The flip side to this discussion is how do you manage future market change? Today's market characteristics are not guaranteed to continue tomorrow. Therefore yesterday's successful approach may not continue to be successful in the future. Thus the need to possibly develop an approach to set stops in each instrument on a volatility based variable equation going forward.

 

[southall]

But why not then increase sample size to 3, 5, 10 or 20 years? At some point I believe one loses relevance of today's market by using data from years too far in the past. For instance, the CL market from 2007 - 09 was vastly different than 2016 - 2017. Same can be said of NQ from 1999 - 2001. I don't see value in bringing those years into the sample set to determine how to trade today's markets.

Seems you have a trading belief that any particular year will be more like the preceding year than one say from 10 years ago.

That seems to be a naive belief that most traders start out with, i certainly believed the same.

In reality, for my type of trading at least, i have found that you cant predict the type of market and system performance you will have in any given year, it could be completely different from last year or it could be the same.

 

[MidwesternTrader]

In reality, for my type of trading at least, i have found that you cant predict the type of market and system performance you will have in any given year, it could be completely different from last year or it could be the same.

Yes, I concur. I think continuing to my last paragraph in the post echoes this thought. One cannot predict the future.

 

[themickey]

@MidwesternTrader Do you already take volatility into account in your trading system? You do mention three ways of defining a stop loss, but it is not clear to me when/how you decide to switch over from one to the other.
I don't do daytrading, but my automated system calculates the volatility over the most recent 25 trading days. So this volatility gets adapted every day.

May I ask which volatility formula you use?