Wednesday, November 18, 2009

Backtesting vs Live

It's been a little while since I have posted something. Jupiter is currently running at -270 pips. Trade frequency is very low for this very slow month. Pluto currently at +65.

OK...I want to talk about the importance of backtesting and furthermore the differences one sees between backtesting results and live results.

Back Testing :

Whether you have designed a system or strategy yourself, OR are learning one from a coach, before you even consider putting money down on it, you should backtest the system or strategy. Let's move out of trading for a moment. I'm into archery, so let's use that as an example. You shouldn't take up a new activity and immediately step into world class competitions. You would head down to the archery centre/field and practice, work with your coach to see where you need to improve, learn the importance of stance, anchor, aiming etc. Practice practice practice!

Ideally, like anything in life, you should have goals. Goals set to identify your level of experience and progression within the chosen field. Once you have established your shooting style, and have proven consistency in your practice, eg. maintaining a score of above 280pts at 30m, then should you consider competing. (Trechnically you can enter some competitions as a newb, but let's keep attention on wanting to win!)

Back to have the details of the system. In this example, let's use the age old moving average cross-over with a stochastic movement out of over-sold or over-bought. Simple right? Just like watching a guy tight rope across the distance between two buildings. He steps slowly and uses the pole to maintain balance. Yup, looks easy enough, let's jump in and do that now!

Hello pavement!

Trading is no different. Even the simplest of strategies require you to backtest them AND "train" for them. Backtesting is your training ground. If you don't have the patience to backtest, then I'm sorry trading isn't for you.

Utilising historical data, you would scroll through the charts and observe how the market reacts around your predescribed conditions (MA X-over and Stochs moving out of extremes). Effectively, you are preparing yourself, training your eyes to identify when certain conditions are more favourable than others.

More importantly, you are testing the strategy itself to determine if it truly is profitable BEFORE laying down cold hard cash on it. In your backtesting, how did the strategy work in ranging markets, trending markets? illiquid markets?

We saw in Jupiter how illiquid markets in September could have been improved by utilising 4hr charts over 1 hr. In my own backtesting, I didn't see such thin markets. Sure enough the targets were smaller than normal, but not so small that spread would have eaten any profits!

The Pitfalls of BackTesting :

Backtesting is easy stuff really. The Fog of War is completely removed. You can see everything there at once. It's all static data. It's not moving. The stage after backtesting is called forward testing. More on that later, but once you have backtested and then applied your strategy to live charts you will see the difference.

Live charts move! In backtesting, the Moving Average cross is clear. But in live charts, you will see occaisions where Moving Averages cross, and then uncross as the market ebs and flows.

With static data, results can obviously be produced very much in favour of profitability. The charts aren't moving, so there is no need to react immediately on the signal. You can sit there and ponder all you like as to whether the setup is acceptable or not.

With static data, you can cover months of market history in a matter of hours. In live charts it will take months of live chart movement to cover the same amount of data. I recently had an extremely frustrating trade that after 5 hours of watching price hit my signal entry, move up, come back down to entry level, move up, come back down to entry level, to the point I no longer could see which way price would go. This is what live trading is about. Eventually I left the charts and let the trade go. Trading frustrated is a BIG no no. The trade eventually went the way my strategy dictated, and to be honest there was enough movement for me to have locked in Break Even, but the bouncing would have taken the trade out at break even.  Hopefully this example displays how different back testing is to live trading.

Success Rate 80%+
They say to backtest something til it is at least 80% successful, preferably 90%. Any system that is more than 50% successful is profitable, but given the pitfalls of back testing, you want to create a buffer zone to allow for your own interpretations during live trading.

Even during bad and slow times, Jupiter is running at about 60-70% success rate. This is from 80-90% (depending on pair) from back testing.