Subject: (To newer traders):
It is interesting to watch several systems that were flying high until the 2/27 plunge, struggling to regain their footing even on trades a couple of weeks later. It would seem that they should almost immediately get back on course.
A similar thing happened in the mid-May 2006 plunge.
There is something called “market structure” (or “market geometry” among other terms, used to describe the overall behavior of a market such as stock indices). After an event like the few days following 2/26, and continuing, sometimes the underlying market behavior becomes nearly unpredictable to certain types of systems.
This is one reason (among many) it is so difficult to make a system that is profitable over the LONG-term. It also can cause a serious hole in your brokerage account.
Some potential solutions:
– stay away from systems with a 4 week equity curve. That is NOT a tested system!!!
– trade more than one system at the same time. This can smooth out your performance, instead of having huge losses or profits. Perhaps a couple of decent systems at $50 a month, is better than the one at $350 a month, that doesn’t necessarily offer better returns…
– trade unrelated instruments (some forex, futures, physical commodities, individual stocks, …). At least Interactive Brokers enables you to trade multiple instruments out of an account. There are likely others
– trade different things within a class. Rather than just trading emini S&P futures, also trade T-Bonds, and/or currencies, and/or metals, meats, grains, energies, etc. Within stocks, don’t just do high-tech stocks, do transportation, food, utility, REIT or other types…
– don’t overleverage. Remember the old saying, that the best way to make a small fortune, is to start with a large fortune…
– learn about money management Stop losses, profit targets, controlling the amount of leverage, This subject is where everyone will give you their opinion, but that does not mean they are correct. Personally, without proof, it is worthless to me. There are many trading beliefs that are not backed up by solid evidence.
For example, someone might say “profit targets are useless.” My answer is, provide me a representative spreadsheet with a large number of trades generated by the system. I will test this theory myself. Of course, like a system’s ability to predict, the profit target must ALSO hold up under walk-forward testing.
Good post. If I may add to this:
Start small and increase gradually.
Perhaps the last two points that you mention each have two components:
1. Choose a system that doesn’t overleverage and uses apropiate money management, and
2. Know in advance what you will do if the system suddenly deviates from this.
I suppose that you do not mean that new traders should impose their own untested stop loss rules to a system, because that is dangerous as well.
> It is interesting to watch several systems that were flying high until the 2/27 plunge, struggling to regain their footing even on trades a couple of weeks later. It would seem that they should almost immediately get back on course.
> A similar thing happened in the mid-May 2006 plunge.
> There is something called “market structure” (or “market geometry” among other terms, used to describe the overall behavior of a market such as stock indices). After an event like the few days following 2/26, and continuing, sometimes the underlying market behavior becomes nearly unpredictable to certain types of systems.
I agree, but sometimes it’s simpler than this. Often systems or traders find a trend and with little concern average down into every counter trend
move. As long as the trend persists they make money hand over fist. Indeed, trading the trend is the right thing to do, but systems need to give up a little to protect themselves. I.e. “the trend is your friend until it ends”.
Check the S&P. Since August nearly every 5 day decline was substantially recovered in the next several days (until 2-27). Likewise
the EURUSD has had multiple multi month trends and some forex systems have gone wild when in sync and lost big when out.
So: check to see if the systems success is simply coincident to the underlying market trend. Did the system blindly hang on until a recovery? Or did it control drawdowns when it was wrong? You may have to actually get in there and look at the trades but it’s worth the time.
Good feedback, Jules & Sam
I thought it might be nice to have a "newer traders" thread, and try to keep it alive with advice for people who are new to trading and/or Collective2…
Also a place they could ask questions freely.
May I add: Understand the concept of regression to the mean.
It is very tempting to subscribe to a system with a sharpe ratio of 4 over a 4 month history. Looking at systems with a 2 year track record however, it becomes clear that no systems can maintain such a super sharpe ratio in the long run. This is equally true for any other statistic. Thus, it should not come as a surprise if after 4 months of sharpe=4, the next 4 months will bring you a sharpe of 0.
If I may add: Understand regression to the mean, because it is a nasty subject.
For example, it does not work the other way around in the example that Science Trader gave. If the Sharpe ratio is 0 in a four months period, you should not expect that it will become 4 in the next four months.
About a year ago there was a discussion on this forum where people suggested to “take advantage” of regression to the mean by subscribing to systems in a drawdown. The participants did not agree in that discussion. My opinion is that this is very dangerous and that it is generally impossible to take advantage of regression to the mean in this way. I can think of an exception to this in some special circumstances, but it won’t work as a general method to select trading systems. (Extreme-os says that it uses reversion to the mean and I cannot deny that this system has predictive power, but there it is applied to stocks, not to trading systems, and probably the system does much more than this.)
I think that this attempt to “use” regression to the mean is another version of the “gamblers fallacy”, which is the belief that after a series of losses it will become more probable to win. This is not true for independent events. It can be true for curves with a negative autocorrelation, but you won’t know that because the autocorrelations are not reported (yet) in C2. Autocorrelations are often small, so even with negative autocorrelations I would not recommend it.
I think that it is generally much wiser to view a serious drawdown as a warning sign and to stay away instead of trying to take advantage of it.
I agree. The notion of regression to the mean is helpful in getting a realistic expectation of a system’s performance in the longer run. It shouldn’t be “traded” by newbies.
Great idea, Ross. Can I refine it by suggesting a dedicated thread to newbies that is easily located. For example, when I first started with C2 the first thing I looked for was a “forum for dummies” like myself and was, obviously, unsuccessful. Additionally, the forum should be highlighted amongst the forums that currently appear at the beginning of the list (New features, etc). It’s of no use if the newbie can’t find the material easily.
FWIW
John
