Tuesday 18 May 2010

After a good trading day, inevitably a really bad one

Just when you think trading is relatively straightforward . . . . A good start to trading today. Initial trades in crude oil and the FTSE put me up 75 points in the first hour. But then a series of unavoidable whipsaw trades whittle down the gains. A mistake in the FTSE applying the entry signal rules costs me 25 pts. A trade goes from a 50 pt profit to a 15 pt loss in 10 minutes - a far faster reversal that the systems were set up to cope with. But then around 2:00pm, a decent sell in both the Dow Jones and crude move me from an 80pt loss to a 100 pt profit for the day. Then things go really pear shaped Around 18:00 I miscalculate trade entry points three times in a row and enter trades that I shouldn't have because I have the wrong entry point programmed. These three errors cost a quite staggering 280pts - inevitably it seems that the market swings are so much bigger when you have a trading error in place. The rest of the day is a small negative as some "real" whipsaws wipe out gains from the downward market moves. So I finish down around 350 pts knocking out most of yesterday's great gain. But at least there is the knowledge that much of the loss was due to my errors rather than the trading systems - in theory this should be fixable Clearly I now have to spend some time sorting out a better entry methodology - some initial ideas in late trading, but I really have to sleep on this overnight. Addendum - the next morning All systematic trading systems have an Achilles heal - a precise configuration of market moves that seems almost uncannily impossible for the trading system to cope with i.e. a configuration of prices that cause the system to almost precisely zig when it should zag and vice versa Some weeks back, using some simulated data, I was able to get a rough idea of what this worst case situation is. It occurs when the market oscillates at a certain fairly-precise frequency both in respect of price changes and times between highs and lows. The systems are fairly robust either side of these frequencies. Quicker moves are ignored, longer moves are profitably tradeable. But moves of exactly the wrong size and period are loss-making. And it doesn't matter if these oscillations are superimposed on an up- or down-trend or occur in sideways markets Also psychologically such moves are quite tough. I sell a market short and within five minutes it is rallying strongly. I buy it back and go long some time later and almost at that precise moment, the market reverses and falls all the way back. It is almost as if the market knows precisely what I am doing and then does the opposite. This may also explain the errors in setting the precise entry points. These should have been wider than I actually set - but narrower entries stop the pain of being the wrong way round sooner. Psychologically, it is sometimes difficult to make the right decision in the heat of battle when everything is going wrong at once (and in such an annoying manner) True, it would be much better if the trade set up rules were followed as they give a chance to break the cycle, e.g. , by not going long at a high but remaining short as the market reverses and falls. But more generally, it just has to be accepted that sometimes this will happen. A few weeks back, crude oil traded lower all day, losing over 250 pts while I made 4. And that was precisely for these same reasons So what can be done. Firstly, the effect can be spotted only after it has occured a number of times. This is pretty easy - trades come relatively quickly, and every trade makes a decent sized loss. But this is little help. At some point the cycle will break and a buy will be followed by a good run upwards and vice versa - possibly even good enough to wipe out the previous losses. So other than correctly following the trade-entry methodology, I should continue to trade per the system Another alternative is to use stop losses, but they are notoriously unreliable for my type of trading system. While they might work in the precise conditions of yesterday, overall they are a bad thing. Finally, this is really an argument for blending in other systems that work on different frequencies. For instance, if I apply the same systems to data on a different period - e.g. 10 minute data instead of 5 - then the systems are not as affected as they were yesterday. They might not have minded a day like yesterday as much as the systems I am using did. But I am not yet set up to manage more systems. So just grin and bear it - yesterday was just "one of those days"

No comments: