Let's say you took your own % performance and made a chart of it. Let's say your chart breaks major support, would you then sell yourself short? Or start fading yourself? Or what if your chart is oversold, would you then long yourself? Obviously the ideal system will have a steady edge and your chart will steadily go up. But I've come up with systems that fluctuate, and I can't help but wonder what would happen if I traded these systems just like a stock. I suppose this really examines what TA is all about. Or this could be possibly the most asinine misuse of TA ever. Just food for thought. Systems that trade systems... By the way Natalie, if your reading this, this is the beginning of the deluge
IMO it only works when your system shows streaks, thus there is an element of dependency between consecutive trading days. If you have such streaks you are perfectly right and can imbed that within the strategy. But it is necessary that you make sure that you do not overfit the basic strategic idea. Most people that I know who claim they can trade equity curves by buy draw downs and selling equity highs, have not done the necessary homework to check for streaks. Simply looking at the chart and identifying five draw downs where you think you should increase size, will most likely by a trap. This is because of big standard error due to the small number of observations. Eager to hear other's opinion on that ...
Mark Douglas discusses this in "The Disciplined Trader" about how prop firms used to chart the performance of traders and then allocate capital accordingly when they knew the individual was coming up against resistance so to speak. interesting................................ I think it definitely has merit. Everything is mental about trading. You will only give yourself as much as you value yourself. You only give yourself a limited amount each winning day, then thats certainly going to show up as some sort of resistance.
As man mentioned, on surface it may look like a good idea to trade your equity chart, but I believe you're adding another layer of complexity to your trading which can skew your results. It opens up a whole new pandora's box of issues and problems.. The old rule of KISS, has a lot of merit when it comes to systems trading. Rather than trading just one system on one mkt, why not diversify among different mkts or systems. I trade three different mkts using a similar system. I chart equity for each of the mkts individually and also combined equity. I try to select mkts which are non-correlated, so to smooth out the combined equity chart.
The idea has merit, but also has quite a few dangers associated. For the sake of argument a system is being traded all the time, except when the system is Long (as per the equity curve chart) it is traded for real, and when the system is short it is 'ghost' traded instead and no money behind it. Running an essentially mechanical system I would expect to see a good hit rate and a good ratio between profit/loss. keeping a track of the systems performance is also necessary (that's obvious). Moving on in my thoughts. Market conditions do vary and can go through periods of time when systems that are very profitable in one set of market conditions can become unprofitable in others, and obviously I have no desire to trade uprofitable systems. So knowing what type of market conditions generally prevail will dictate what type of system is used to trade it. Analysing the equity curve from a system can be an early warning of a system starting to fail (albeit temporarily) and might save some losses. But there are other methods of ascertaining where a particular system is still valid and within the acceptable range of profitability. On the other hand there is always the possibility that the set of drawdowns was temporary and factored into the system performance, and that the system goes short (i.e. from actualy trading to ghosting) just as it becomes profitable again and long just as the losers show up. Even moving between different systems can be subject to this problem, although it can be argued that if you have the right system running in the right conditions at the right time for maximum profitability you would be increasing your overall results and reducing your risk. So even Ghosting it can lead to always being in the wrong place at the wrong time (where choice of system is concerned). And it's a problem that gets compounded the more systems you have. Now that is not to say that it is all bad news, because systems that are very heavily curve fitted tend to have a limited shelf life, or a limited number of times that they can be applied. They can be good in one set of market conditions but not in others. So keeping track of both market conditions and the results of the 'ghost' equity curve can identify longer periods where the system has greater profitability. My thoughts so far are that it could be a good tool if used in combination with other tools, and with great care, but on its own carries quite a large element of additional risk associated. Natalie
Aha! So the argument behind doing TA on your own equity curve is that you are predicting your own emotions? Just as the argument behind TA in general is that it predicts emotions? That's interesting. So it follows that this 'derivative' TA would only work on discretionary traders, because with a system there is no emotion.
Natalie doesn't it all come down to the question if the performance of a system can be forecasted at above random probability? if the answer is no, than there is no possibility in trading the equity curve. The system will do fine in the long run but it cannot be predicted when in the short run. if the answer is yes, than all you said is applicable. how to find out whether there is nonrandom dependencies? autoregression for the equity curve itself and mulitregression for exogene events come into my mind. actually we are running a system where we found small but significant autocorrelation in the equity curve and now we are constantly changing the leverage according to this analysis. Seems good so far. At other systems we found no significance, thus no trade in there. For your thoughts on multisystem: I think this is the future. Have a dozen systems at hand and try to predict if the environment will be favourable for them. I assume it might be easier to forecast environements than price itself. For instance in a sharpe down trend systems that profit most from breakouts out of quiet sidemoves will not find their best trades in the near future. Or systems that profit most from big overnight gaps as characteristical for a strong bull market, will most likely not do to well within the next months. You see the point. Easy to add VIX or whatever have you. Then those systems best for that environment get more money than the others. I think the future is permanently shifting funds from one strategy to the other. Not sharply but according to continuously changing odds for the stratey set. peace