Sunday, July 29, 2012

Historical trade - 9th May 2003 - EURUSD

This pinbar confirmed my theory about Friday closes. This particular one again formed at the end of Friday.

The only things in its favour were:

+ size
+ off a recent resistance / support level
+ off 8 EMA

However, it did not form at a 50%-62.8% fib level, nor did it form along a trend line. As you can see, price gapped up on Monday's open before reversing.

This trade would've provided a 1:1 R:R, but because of the gap up, I would not have been able to open an order on the break of the pinbar.

- on intra-day charts (4H and below), beware of reversal signals towards Friday close.

Historical trade - 11th April 2003 - EURUSD

A very large pinbar formed on the 11th April. 2003, on the EURUSD. This setup looks very similar to my previous entry.

Factors in favour of the pinbar:

+ size
+ off a significant support zone
+ off a 50% fibonacci retracement level
+ off 21 EMA (barely)

What happens next...

Some thoughts:

- this setup isn't as strongly ranging as the previous setup
- it took awhile for the pinbar to break and trigger an entry (16 hours). Setup a pending order until pinbar is broken in either direction?
- this particular pinbar also formed just before Friday's close. I'm suspecting that short-term reversal signals will be prevalent towards the end of Friday as traders close their positions for the weekend.

Historical trade - 14th February 2003 - EURUSD

This trade would've taken place on the 14th Feb, 2003. I identified a pinbar with the following confluence:

- rejection off 21 EMA
- rejection off trend line
- rejection of 50% fibonacci retracement level
- rejection off a secondary support level

What happens next...

What can we learn from this trade?

- firstly. the pinbar is located in the middle of the longer-term range
- the pinbar formed late Friday just before the weekend close
- the breakout candle was a bearish doji, so close trade if the breakout turns bearish?
- the support level that the pinbar formed off wasn't significant
- while the pinbar formed off the 21 EMA, it formed below the 8 EMA, suggesting that momentum is turning bearish at the time of the pinbar's formation

Zones of repulsion

One of the ideas I've wanted to focus on is trying to define zones of repulsion on a chart. If you can identify likely zones of repulsion, you have a good method of identifying places to enter and exit a trade.

So far I've identified the following factors that indicate a likely zone of repulsion. If you have a confluence of the following factors, my belief is that any price reversal is more likely to be strong and persistent. If you have a price action signal like a pinbar or engulfing bar, I would look for a minimum of three of the following factors before I would consider a trade.

- support and resistance levels
- trend lines and channels
- rejection off moving averages
- big round numbers for price levels (1.0000, 1.5000 etc)
- fibonacci retracement levels (especially 50-62.8%)
- chart patterns
- fakeouts / upthrusts out of a range
- oscillator divergence

Volume spread analysis, return to price action

I spent a few days learning volume spread analysis (VSA). This is probably one of the most arcane ways of reading charts. The basic principle behind VSA is if you see a candle with low tick volume, the smart money is staying out. If tick volume is high, then the smart money is moving in.

You want to be moving in when the smart money is. If the smart money is staying out, it's because they want a better price before they move in for the kill, e.g. if you see a bullish candle with low tick volume, it suggests the smart money is staying out and letting price rise before they move in to short. However, if you see a bullish candle and tick volume is EXTREMELY high, then it suggests the smart money is actually shorting and trying to hide their selling in a bullish candle.

VSA is much more complex than that. The main difficulty with VSA is subjectivity. How do you define "low" and "high" volume? VSA is its own eco-system, complete with its own chart patterns and terminology. They call pinbars "upthrusts" or "downthrusts"". Dojis are called "no demand" or "no supply" bars. It's a different way of seeing things, but it seems to work for alot of people. However, there seems to be alot of scope for second-guessing and is far too arcane to me.

Price Action

I think my flirtation with mechanical indicator-based trading is over. After testing 30+ systems and their variants, I don't think I will find a highly-profitable system, or at least one that can be traded by a human. I suppose the profit factor from a mechanical system will always be low. This is not a bad thing per se if you can trade with high frequency (i.e. a black box). I'm probably years away from becoming an algo-trader, though.

I spent a few hours today reviewing price action material and assessing historical trades using Forex Tester 2 (highly recommended for backtesting). I'll start posting my historical trades from tomorrow. I have a few ideas up my sleeve which I want to test. 

Wednesday, July 25, 2012

Update on FRAME backtesting and data bias

Yesterday I sampled a new set of 100 trades from random dates between 2001 and 2011 and saw my expectancy plummet. The possibility of data bias has always lingered in my mind, and yesterday's backtest confirmed its existence.

Grabbing a large sample size isn't enough. For sampling to be as representative of a population as much as possible, you want your selection criteria to be strictly neutral.

The 100 samples I grabbed from the previous backtest originated in early to mid-2001. I've always intended to grab a total of 1000 sample trades, with the other 900 coming from other years. However, I envisioned grabbing those 900 samples in "chunks", in the same manner as my first 100 samples.

The problem with grabbing data in a "chunk" is that each sample influences the next one. You want this influence to be minimal to avoid bias. If I pull a 100 sample-size chunk from a strongly-trending period, those 100 samples will give a giant green tick for a trend-following system. But it's not representative of the entire population, which will most likely be a mix of trending and ranging periods.

Lesson learned.

I now use to randomly choose dates and times for sampling.

As for the backtest results from random dates, the optimal R:R looks to be 7:1, providing an expectancy of 29%. That expectancy is not too bad, but reward-to-risk is too high for my liking. My win% is 17%, so expect alot of drawdown between wins. You'll have to trade small for this system to be profitable. But trading the 15M chart provides so many opportunities. Realistically, you can trade 20 times per month on the EURUSD alone.

However, this isn't something I want to continue backtesting, for now anyway.

Friday, July 20, 2012

Introducing the FRAME system

Yes, I'm still alive. Just because my posts have slowed does not mean I've stopped focused on trading. It simply means I'm focusing my energy on actual trading rather than the blog.

This is the system I've been developing and backtesting for the last month. It's based on Bill Williams' fractals and a 50 SMA (simple moving average), which I shall call the FRAME system. FRActal + Moving averagE = FRAME. Geddit? :)


So far I've tested 100 samples from the EURUSD on the 15M timeframe.


Most of my backtesting has been conducted on the 4H timeframe, with unremarkable results. However, I've spent today testing on the 15M timeframe, with much greater returns.


Activate the fractal and 50SMA indicators on your trading platform.

When price crosses over the 50SMA, look for the first fractal to form opposite the SMA. i.e. If price crosses the 50SMA upwards, look for a fractal at the top of a candle. If price makes a downward cross of the 50SMA, look for a fractal at the bottom of the candle.

Once the fractal forms, that means a retracement is underway. If the retracement does not cross the 50SMA, then you have the setup of a potential trend. Enter on the break of the fractal.

Stop Loss

For the purpose of my backtest, SL = 20 pips. I determined that the average candle on the 15M timeframe is 5 pips, so this SL gives you space to cope with a 3-4 candle retracement after entry.

Take Profit

I'm still determining how to optimise my TP targets, but after 100 sample trades, it seems that a TP of 90 pips seems best. It provides an expectancy of 63.5%, which blows away everything I've tested so far.

Example setup


So far I've only tested 100 trades on the EURUSD from Jan 2001 to March 2001. I'd like to gather 1000 by next week, but so far expectancy is looking real juicy.

Wednesday, July 11, 2012

Saturday, July 7, 2012

Lower timeframes = trend-friendly?

I've been spending the last few weeks doing more backtesting, per usual. In the last week I've focused on Bill Williams' fractals, moving averages and trendlines.

Anyway, I decided to focus my backtests on the 4HR chart instead of the daily and asked myself if greater profit can be found trading trends on lower timeframes?

To answer this question, I decided to find the average candle size for the AUDUSD on the 4HR and daily timeframe from 2001 to mid-2012, which were approximately 47 pips and 107 pips. I then found the low and high for the AUDUSD during this period, which were 0.4818 and 1.1078. The range between these two numbers is 6260 pips.

This is just an elementary exercise. On the daily timeframe, this range is equal to 58.5 average-sized daily candles (6260 / 107 = 58.5). On the 4HR timeframe, this is equal to 133.2 average-size 4HR candles (6260 / 47 = 133.2).

So if you managed to perfectly pick the trend from bottom to top, you'd make over double on the 4HR chart than the daily chart. The same logic will apply to the 1HR and lower timecharts. Of course, this will be balanced by the relatively higher proportion of spread that you will pay as you trade the lower timeframes, plus the "noise" that'll become more prevalent.

So if you are trading trends, you might examine the possibility of trading the 4HR or 1HR charts, and to stick to the major currency pairs to keep your spread to a minimum. The daily charts seem to be more suited for mean-reverting or range-friendly systems.