Wednesday, June 27, 2012

Trading For a Living, Not Living for Trading (Series): An Introduction a...

The following is an interesting introductory video from Airelon, a commodities trader. I stumbled on his videos last year before he took them down. Thankfully he's put them back up. You'll notice that most of his videos focus on discipline and psychology. Successful trading is 10% technical, 90% psychological.

Why I believe trend-following strategies are less effective in forex markets

One thing that took awhile to notice is how range-bound the forex market is, relative to other markets like stocks or commodities. Range-bound markets tend to be dangerous for trend-following systems, since these systems only work by generating a few giant rewards in exchange for many small losses. Such rewards can only be found in long-term, sustained trends in the market. A choppy, horizontal market will wipe out trend-following traders.

I've tried backtesting a few trend-following strategies, such as using an ATR (Average True Range) trailing stop. A few quick backtests using various multiples of ATR yielded unremarkable results.

Luckily, gold and silver are unoffically considered "currency" and most forex brokers allow you to trade both precious metals. However, gold and silver are also commodities with real-world applications, and can be considered as part of the commodities market as well.

Let us compare gold with the USDCHF and the AUDUSD. The CHF is the Swiss franc and is 20% backed by gold (courtesy of the Swiss central bank). The AUD is a commodity currency, with Australia being a major exporter of gold. Thus you would expect some correlation between the two currency pairs and gold.

Gold chart from mid-2006 to mid-2012:

Mid-2006, gold's low was around $545/oz. It peaked at $1920/oz in 2011. This is a multiple of 3.5. For the most part, there is a clean, crisp trend.

USDCHF chart from mid-2006 to mid-2012:

As you'd expect, there was an (inverse) correlation between the USDCHF and gold. The USDCHF's high was 1.2768 in late 2006, before hitting a low of 0.7061 in 2011. This is a multiple of 0.55. Since we want to compare this multiple to gold, we find the inverse, which is 1.81.

AUDUSD chart from mid-2006 to mid-2012:

And here is the chart for the Aussie dollar. Look at all that chop! The Australian dollar hit its low during late 2008 at 0.6009, before peaking to 1.1079 in mid-2011. This is a multiple of 1.85.

What it all means

Forex markets are more range-bound than other markets, as demonstrated by the lower multiples of 1.81 and 1.85 versus gold's 3.5, even when there is correlation. Lets examine why.

1) Currencies measure the strength of national economies. Barring some catastrophe like war or instability, economies only move a few percentage points per year, GDP-wise. This is unlike commodities, which are more subject to "micro-economic" factors like bad weather, labor disruptions or technology.

Stocks are much more nuanced and volatile, with garage start-ups capable of experiencing explosive growth as they expand or create new market share (e.g. Facebook, Google), or suffering catastrophic collapse (e.g. Enron). Can an entire national economy see the same kind of growth as Facebook did? Nope. Likewise, you won't see an overnight collapse (barring some catastrophe).

2) The forex market is the largest and most liquid on Earth. Informed, liquid markets tend to find equilibrium faster than less-informed markets like penny stocks.

3) Governments and central banks have an interest in maintaining price stability. A central bank or government isn't going to care if Facebook stocks are overvalued. But if their currency becomes overvalued and starts harming local exporters, these two institutions are more likely to manipulate their currency, or even peg it.

4) Currencies are less influenced by inflation than stocks or commodities. Lets assume we have a currency pair, with both currencies experiencing 3%pa inflation. Over time, inflation should not influence the price of the currency pair as both currencies are being devalued at the same rate. On the otherhand, stocks and commodities will be rising 3%pa on inflation alone.

5) Currencies are less likely to be influenced by emotion like fear or greed, especially the latter. Greed or wishful thinking may create bubbles in certain stocks like the dot com bubble. However, no-one is ever going to bet their house on the value of the New Zealand dollar flying to the moon.
That's all I can think of at the moment. Thoughts appreciated.

Tuesday, June 19, 2012

ADX Crossover system now in forward-testing phase

Over the last few weeks I've extended the backtest on my ADX crossover system to the USDCAD (1996-2012) and the EURUSD (1999-2012). I've manually analysed and recorded over 2,000 individual trades. This was insanely long. While backtesting I managed to watch through seasons 3 and 4 of Breaking Bad, and season 1 of Battlestar Galactica. I think that's about 40 hours of pure backtesting in the last few weeks.

I've also performed a basic optimisation of my system. The optimal R:R between the three pairs turned out to be 4:1, which yielded an approximate 20% return per trade. I like to test the AUDUSD in the near-future as well.

Here are the basic equity curves for the EURUSD, USDCAD and GBPJPY at 2% risk with a $10,000 account, from mid-1996 to early 2012 (except the EURUSD, which only came into existence in 1999).

The USDCAD didn't perform well, but I was expecting this. The USDCAD is a real bitch to trade. Even Nial Fuller admits he avoids the USDCAD.

My experience trading the commodity currencies is that the AUDUSD, NZDUSD, gold and silver, are the best to trade, I tend to avoid the USDCAD as I find it fires off many “false” trading signals, this may have something to do with it being heavily influenced by the price of crude oil. Whatever the reason, I typically avoid trading the USCAD and advise my students do the same, perhaps at a point in the future the USDCAD will “behave” more logically, but at the current time I tend to avoid it like the plague.
My next plan is to forward-test this system with the major pairs, in conjunction with a backtest on the AUDUSD.

Monday, June 11, 2012

Further update on ADX scalping system

I finally finished the backtest on the GBPJPY from 1996 to 2012. My sample size is 731 trades. The final optimal R:R was 2.5:1, which provided an approximate return of 13.57% per trade. So, on the surface, this looks like a usable trading system.

The next step will be to optimise this system and improve my return. Once I finish, I hope to provide an in-depth post detailing my findings.

Sunday, June 10, 2012

Update on ADX scalping system

I've managed to compile 500 sample trades using the ADX(3) crossover. My final expectancy was around 0.05R per trade at a 1.67R reward-to-risk. Not an ideal result. A 5% return per trade is better than a loss, but I was hoping for something better.

Tuesday, June 5, 2012

ADX scalping system

I've spent this weekend backtesting a new "dumb" ADX scalping system. You basically set the ADX to 3 periods, based on median price. You enter on the break of the candle whenever +DI and -DI cross, with your SL placed on the opposite end of the candle. No regard is paid towards things like resistance or trend. Like I said, dumb.

My sample size is only 200 but so far the results are quite good. I've tested the GBPJPY from 1996 to 2000 on the  daily timeframe so far. Optimal reward-to-risk seems to be 1.67R, which offers an expectancy of 0.15R per trade.

What I really like about this system, though, is the abundance of opportunities. On the GBPJPY alone, I'm averaging about one trade per week. Trades don't usually last for very long either. Throw in some more currency pairs, and you can probably trade ten times per month on the daily timeframe. I'll update my findings once I finish backtesting the GBPJPY and, if expectancy is still positive, optimise the system.

Monday, June 4, 2012

Possibly the most important metric for any trading system - "expectunity"

Many newbie traders like to define their success by the amount of pips they win per trade. For others, it's the win rate (%). Both are important, but by themselves, tell an incomplete story.

Measuring your success by the amount of pips you win seems simple. In fact, I would use this metric as a way of measuring your ability to predict market movement. Obviously the higher the better. However, it ignores your position size (the amount, or number of lots, you decided to buy or short). Example: I decide to buy a single lot of AUDUSD and win 100 pips. Feeling super-confident about a trend reversal, I decide to short the AUDUSD with three lots, with a stop loss of 50 pips. The trend continues and hits my stop loss.

According to my metric, I'm still up 50 pips (100 pips won - 50 pips lost) and should be in profit, right? Nope. If you take position size into account, you are in negative territory.

100 pips * 1 lot won - 50 pips * 3 lots lost = - 50 pips lot

Disregarding pips altogether:

1 lot won - 3 lots lost = - 2 lots

Suppose your win rate is 90%. This looks marvellous on the surface, but tells nothing about the size of your wins or losses. Hitting a 90% win rate is extremely easy. Just set a tiny profit target per trade, or worse yet, use no stop loss. In fact, with no stop loss or leverage, it's conceptually possible to win 100% of the time (disregarding something catastrophic like deregisteration of a currency). Just keep your trades open when they turn against you and wait, wait, wait until the market turns back in your favour.

Of course, the process may take years if you bought at the absolute top or shorted at the absolute bottom. But as long as you don't close your losers, your win rate remains at a solid 100%. Yep, impressive, but not realistic nor optimal. If you take opportunity cost into account, you end up as a strong loser.


A better metric to use is expectancy. Expectancy is average win rate (%) * average amount won. Amount won can either be absolute pips, dollars or a risk:reward ratio (my preference).

The trouble with using pips is that not only is the average daily movement for each currency pair different, but they shift over time. An expectancy of 50 pips isn't so useful if it was based on data with an average daily movement of 120 pips, and today's average daily movement has dropped to 80 pips. Some currency pairs are also much more volatile, such as the GBPJPY, making an expectancy from a less-volatile pair like the USDCAD incomparable. Apples and oranges. You can try standardising your expectancy across different currencies, but it's a tedious exercise and not needed, as I'll explain a few paragraphs below.

Using dollars is acceptable if you're risking the same amount of dollars per trade. But if you're profitable, this shouldn't be happening as you'll be increasing your dollars per trade. Not really useful.

My preference is using R:R. It's unitless, thus allowing comparison with other trading systems and markets. It also describes your system in terms of risk, and risk management is what trading is all about.


And that brings me to my next point. A high expectancy is good, but what's the point if your system only allows you trade once a year? Thus we arrive at the most important metric for any system: "expectunity".

Expectunity = expectancy * opportunities to trade

Decreasing our expectancy is acceptable if we compensate by increasing our opportunities to trade(e.g. trading on lower timeframes, loosening our requirements). High-frequency traders understand the concept of expectunity by trading with extremely low expectancy, thousands of times per day.

The role of a trader is to optimise their system in terms of expectunity. At this point, it becomes psychological. Everybody wants to be right 100% of the time. There's comfort to be found in predictability, but this will hurt your profitability. To maximise your expectunity, you must be willing to accept loss and being wrong. But in the end, it's all about seeing your bottom line growing.