Tuesday, 9 September 2014

Seven deadly sins of trading


I don't believe that trading and investing in the financial markets requires a huge amount of skill to do reasonably well. It's mostly avoiding making a series of typical mistakes. Here then are my Seven Deadly sins of trading to stay away from. Some of these will be familiar but others are often overlooked.


No system


True skill in trading is very rare in my opinion. Most traders would be better off sticking to a simple rule based trading strategy.

I've also noticed that the trading 'systems' in many books are vague and/or highly subjective and not really a set of rules but rather a set of guidelines. This could be because:

  • The authors have a system but are unable to articulate exactly what it is.
  • They do have a system, can articulate it, but choose not to so they can make some more money flogging expensive seminars or newsletters.
  • They don't really have a system at all, but have tried to write down some rules ex-post on what they think they do.
In any case it doesn't set a very good example to the amateurs reading their books. If a system can't be written down in a precise algorithm that a computer could execute without any human input or judgement, it isn't a system.

Of course there is no point in having a system if you don't stick to it.



Not having, or not sticking to, stop losses


Even if you are a skilled trader who uses their own judgement rather than some strict rule based method for entering positions you must have strict rules for exiting your positions in the event of a loss occurring. Otherwise you won't know how much capital you have at risk on each trade.

Suppose you set up your stop loss so you would lose £10,000 if it was hit. The price then moves through the stop loss. Is your response:

  • The fundamentals are good so I will hang on. Probably a good time to average down actually.
  • Only an idiot would sell out here, this is just a short term blip
  • The original stop loss made sense but now the Fibonacci retracements have shifted so a new lower level is in order.
  • The short sellers are trying to take out the longs, once they've done that the price will rebound nicely.
  • I can't afford to take the loss!
  • I've already sold out. Why are we even discussing this?
If you picked any of the first five options, you are a danger to yourself and your bank account. You thought you had £10,000 at risk on the trade but now its actually much more than that.



Risking too much capital 

 

There are fancy mathematical models like the Kelly criteria and much simpler ways to decide how much to put on each trade.

(If you are interested in this subject then Fortunes Formula is a must read)

Behind the equations is a simple truth. No matter how good your trading is if you bet too much you will stand a chance of blowing up before your skill is translated into good performance.

Consider the genius who develops a trading system with a back tested Sharpe Ratio ratio of 2.0. Ignore the hype of anyone peddling anything better; this would be a really great system.  

To maximise your expected wealth if you had twenty thousand quid (bucks or euros....) of capital you should aim for an expected annualised standard deviation of returns of around forty thousand pounds a year. Or if you like around 95% of your daily profits should be between minus and plus 5,000.

The genius will do pretty well. On average if all goes well they will earn £80,000 a year pretax from their trading system, and more in later years if they can keep money in and compound their returns.

(Note to readers who still work in the financial markets. In case you didn't know to a normal person £80,000 is a lot of money)

If you adjust your risk according to any gains and losses then at that level you have around a 96% chance of keeping at least half your capital over a ten year period.

But a 2.0 Sharpe ratio in back test rarely materialises into reality. If you only managed a Sharpe Ratio of 1.0 (which is still very good indeed) then you only have a 50% chance of still having half your capital intact one decade later. And if you've really screwed up and get a SR of 0.5 then there's a very good chance (88%) that you will have wiped out at least 90% of your assets!

Most traders would do very well to expect a SR of 0.5. This means downsizing your bets to a quarter of what the 'genius' did. On average you'll make £5,000 a year at that level. Not enough? Then put up more cash. But only what you can afford to lose 90% of if things go wrong. Better still stick to trading part time job.



Setting stop losses according to money management rules


95% of trading books give lousy advice on position sizing by telling you to set your stop loss according to your capital.

So if you are trading crude oil futures and you have £10,000 or around $16,000, then you risk 3% or around $500 on your trade. Fine so far. Since NYMEX crude oil contracts are for 1000 barrels, your stop needs to be 50 cents per barrel below your entry. That means if you buy in at say $100 you will set your stop loss at $99.50.

On the other hand if you are a much wealthier person with £1,000,000 in trading capital then with $50,000 at risk your stop loss would be $50 below your entry, or at $50.

This doesn't make any sense. The oil market doesn't know or care how much trading capital you have. Since right now it moves roughly $2 a day or less two thirds of the time (one standard deviation of returns) the poor guy is probably going to get stopped out in hours, whilst the rich guy will be waiting months or even years.

Some books improve on things slightly by pointing out that the richer guy can trade more contracts. True, but how many more? And should he change his stop loss depending on that?

Fact is you should set your stop loss independently  of your wealth based only on the dynamics of the market and how long you want to hold positions for. Assuming the poor and the rich guy have the same opinion about these things their stops should be in the same place. Probably somewhere between $0.50 and $50... but lets call it $5 just for the sake of argument.

That means that both guys will have the same $5,000 per contract at risk for each trade.

You should then buy contracts according to how much money you want to risk on the trade, which of course is dependent on your wealth. That means the rich guy should pony up for 10 contracts. The poor guy shouldn't be in the market at all unless he can trade fractional contracts maybe through CFD's, or if he is going to day trade for which a much smaller stop loss might make sense.




Having profit targets and/or not using trailing stop losses


Again this is an error made time and time again by many books by 'experts'. Profit targets are pointless.

Why should you care about how much money you make on each trade? Isn't it better just to make as much as possible by using a trailing stop loss? Worst still if a profit target like the stop loss is set according to your capital.

Profit targets only make sense when they aren't profit targets, but part of your trade logic. So if you believe that a market is oscillating between £1 and £2 a share then of course it makes sense to sell at around £1.90. If you have evidence that trends collapse when they become overextended, and that equates to a level £5 above where you are now, then by all means sell out at this level.

This is all good but is completely different from a blanket 'Sell once you have a made a profit of £1000' rule which you apply to any market, regardless of it's dynamics or the reason why you put the trade on to begin with.




Trading too much

As well as risking too much capital nearly everyone out there is probably trading too much. I've seen it done by highly sophisticated institutional investors with complex models to optimise trading patterns versus costs. I've also seen it done by naive characters scalping in trading arcades being exploited to create commission flow for brokers.

From the maths above someone with £100,000 of capital can probably take a position in one crude oil future. It costs me around 100 cents in commission to trade one crude oil contract. But worst still the bid to ask on that crude oil future is 1 cent, or $10 of real money. So that's $12 per open and close trade if I have to pay the spread. If do ten of those a day I need to make $30,000 a year just to break even. On £100,000 of capital that means a reasonable Sharpe ratio of 0.5 will be pretty much halved.

To an extent the financial markets are a zero sum game. You can make reasonable pre-cost returns by not making mistakes. But only if you have a great deal of skill will you make enough to overcome the costs of trading this much (or another advantage like fast connections and a bent market).


Overfitting

 

This one only applies to the really smart people out there who have developed trading strategies through backtesting, data mining, call it what you will. As I said above a 2.0 Sharpe ratio in backtest really materialises into reality. Unless you have a really good grasp of statistical significance then don't even try to fit your trading models. Pick some really simple trading strategies, average them out, run them and stick to them; without going near any backtesting software. I almost guarantee you'll outperform any fancy machine learning / genetic algorithm / econometric magic you care to mention.



To conclude

 

We can't all be Lewis Hamilton (insert name of favourite professional driver). But we can drive safely and competently without having an accident. Only if you can do that should you even think about trying to go faster.

The same is true in trading. Yes it might be that you do have the rare ability to make significantly more than average traders do, but you will never find out if you don't get the basics right. On the other hand if you're just an ordinary person then you can steer clear of blowing up and do okay.

Be careful out there.


3 comments:

  1. Yes I have chomped through a couple of spread betting accounts committing all of these sins...

    In the end I just packed it in...

    ReplyDelete
  2. You mention machine learning algorithms as if they are benign.

    do some quant shops not use machine learning...?

    ReplyDelete
    Replies
    1. Scott
      I think you mean malign. Yes probably some quant shops use machine learning. My issue with it is that it's very easy for novices to use it to overfit a horribly overfitted non linear system with way too many parameters. That doesn't mean it won't be useful in the hands of someone who knows what they are doing. But my personal opinion is that at best machine learning will reproduce similar results to a simpler technique, with more work.

      Delete