Systems Workshop: "Buttonwood" E-mini daytrading system

[earik]

That Murphy's Law piece appears *everywhere* when it comes to systems. Check out the equity curve for the merged system from the MTS book itself, and the correlation with the release date:

MTS Equity Curve

mts_equity.png (49.27 KiB) Viewed 81074 times

A nice drawdown right after the release date to get everyone to stop trading, right before it broke to new equity highs.

I've dealt with some institutional guys recently through my fund, and they seem to be acutely aware of this as well. Their solution seems to be to leg into systems. So rather than committing 100% of the trading capital all at once, they'll do a percentage over a period of months which avoids buying in right at the high and having to weather the storm. Kind of like dollar-cost averaging, but for a completely different reason. After it happens to you once or twice, you just assume that the market is out to mess with you personally, and you take actions to make it harder...

Earik

[hjelmstade]

After it happens to you once or twice, you just assume that the market is out to mess with you personally, and you take actions to make it harder...

Yeah, certainly seems to be the case sometimes

And legging into systems sounds like a good idea. I'll probably start doing that in the future as well so thank you.

[NDscorpini]

Hi Earik,
Hopefully I am not out of line for asking this question(s) but...
1. How many systems in your fund?
2. How many different time frames?
3. How many different commodities, forex pairs, stocks does it trade?
If by changing the variables could use reduce the drawdowns and generate a smoother equity curve? Just asking
Best wishes,
NDscorpini

[earik]

Hi ND,

It's always a balancing act with this stuff. You can get a smooth equity curve, but often at the price of some other variable (like annual returns). In my case, I chose to shoot for high annual returns, but I have to accept some volatility as a result. So there are definitely some deviations from the ideal, similar to the plot I just posted of the MTS system. If you went to the opposite end of the spectrum, you could compare the returns of a CD at your bank. A perfectly smooth equity curve (every month is always up), but you only make like 1% per year. If you could get the smoothness of the bank with the annual returns of the fund, that would obviously be the ideal. Easier said than done though...

Regards,

Earik

[NDscorpini]

Hi Earik,
The only reason why I brought it up was an article I read about Toby Crabel. He manages billions of dollars but yet he can allegedly make 1 to 2 percentage per month fairly regular basis. He achieves this but running 8 to 10 systems at a time on a whole host of different instruments that are negatively correlated as often as possible. It doesn't hit the homerun but it is steady. He was running an employment ad at one point looking for people who could develop systems.
Best wishes,
NDscorpini

p.s.
The problem with this approach is, you have to be rich to get rich. How do you start out unless you have a insane amount of money to start with?

[hjelmstade]

Hi Earik,

I finally had time to come back to Buttonwood. I turned off my previous version several months ago as I suspected something may have changed with the weather filters as Buttonwood was trading every day without any gaps which was greatly impacting its performance. I just now had time to investigate.

I loaded the buttonwood_index script you shared earlier in the thread on a daily chart and discovered that the spherical filter has been above its threshold for nearly all of 2017. The first screenshot below is showing buttonwood_index on 2016/17 daily chart and the second is from 2015 as a comparison.

buttonwood_index_2016-17.png (26.54 KiB) Viewed 80999 times

buttonwood_index_2015.png (27.15 KiB) Viewed 80999 times

I then created a new version of Buttonwood that allowed me to turn each weather filter on and off and confirmed that the spherical filter is causing the issue. Even after reoptimizing the threshold and playing around with the other spherical settings, it is still not ideal.

Now that it has been a while since you first built Buttonwood, do you have any thoughts or new ideas on the filters, especially our spherical filter?


Now the progressive filter is holding up very well. In fact, reoptimizing only resulted in the progressive_orb being changed 0.95. Threshold of 3 is still the best after several years. The charts below are a new version of Buttonwood that is only using the progressive filter in addition to a couple other slight changes since the last version I shared. Over 5 years it still has an avg trade of $70. If I run for just 2017, its avg trade is only $40 so it may still need some work but it is a whole lot better than using both spherical and progressive.

What are your thoughts on just using the progressive filter? Would you consider that too risky? Do you have any other ideas or hints on where to proceed from here?

buttonwood_6-25-17_summary.png (31.86 KiB) Viewed 80999 times

buttonwood_6-25-17_equity.png (22.46 KiB) Viewed 80999 times

Thanks!

[earik]

Hi hjelmstade,

Sorry for the delay in getting back to you... It sounds like in the original build, I didn't go back far enough with the Spherical filter to test those threshold levels. It sounds like some larger planets must have moved into play, and pushed the whole index up. Once you re-optimize the filter, things would work until those planets move out of orb, and then you'll have the opposite issue where you aren't getting any trades at all, since the threshold is now too high. Kind of a catch-22 there... One workaround for that might be getting rid of static thresholds in general, and use something more dynamic. The Adaptive Zones indicator might be useful here, or something along those lines. The idea being to measure the highest and lowest points in the index over a specific lookback period, then set the threshold based on those numbers. That way you can "float" along with the tide and would look for periods where the index is at relatively higher points than recently, rather than higher than some arbitrary value. That's probably what I'd investigate next.

Or, you can just ditch the Spherical filter altogether, if you think the other one alone is fine. One thing that bothered me a little about Buttonwood is that by the time we were done with it, it was kind of complicated, with lots of moving parts. That's always a bit of a red flag, so throwing stuff out doesn't necessarily feel like a bad thing...

Regards,

Earik