Mark Thomas — Trade On Track
Continuing on with our price action study, in this article we’ll put together a plan for analyzing a particular currency pair, on a particular time frame, and see if we can work out a way to get some probabilities on price continuation.
In the last article we talked about a study that was published a few years ago that presented some probabilities of price continuing in the same direction. For instance, the study said that if the EUR/USD moved 15 pips on a 1hr chart, then it was 72.4% likely to continue to 40 pips in the same direction, without hitting a 20 pip stop loss. Those figures make it look dead-easy to trade based on probabilities. Just keep trading consistently using a 15pip-break rule, set a stop at 20 pips and a target at 25 pips, and you’ll win 72% of the time!
We’re here to test this little theory out. While I’m a strong believer in consistent, smart trading being able to win the war in the long term, I would like to do some more testing, with more precise entry, stop and exit points. So, in this article, I’ll map out a plan for doing this testing.
There are obviously lots of possible currency pairs to trade when trading the forex. For our new price action study, I thought we should limit the analysis to just one pair, the EUR/USD. Analyzing too many pairs just makes it more complex and confusing than it needs to be. Basically, if we can get some nice probabilities and a workable, profitable strategy on one currency pair, then it’s likely we can apply it to other currency pairs later if we wanted to. Therefore, we will stick to just the EUR/USD figures for now.
To do our analysis, we’ll need quite a bit of price history so that we get a decent sample to do our backtesting. There’s a fine line between “curve-fitting” a strategy and performing a thorough analysis on historical data. What this means is, we want to make sure our probabilities are as accurate as possible, given what has happened to price in the past. But, we don’t want to tweak things in our strategy so far that it ONLY works on the past results. We want decent statistics based on historical information, that is likely to carry through into the future as well.
Also, we need to define how fine-grained our historical data will be for our backtesting. We could take hourly data (high, low, open close figures) – but that leaves a lot of room for error and doesn’t allow us to do any price predictions for anything less than an hour. Or, we could take tick data – which is price information every time we get a new bid/ask data from the broker (could be several per second). This could result in a huge amount of data though, and is probably overkill.
Taking these considerations into account, I think a decent sample of data would be the last 4 years of price information, taken at 1 minute intervals. So, we’ll have open, high, low and close figures for every minute during trading hours, over the last 4 years to work with.
To keep this study as useful and relevant as possible, I’d like to analyze different bar periods. Some traders like to trade 5 minute bars, others are most comfortable with daily bars, and probably the majority of us like to trade the periods in between. So, I will analyze and present statistics on the following bar periods: 5min, 15min, 30min, 1hr, 4hr and daily. I have a feeling that we’ll get the best results on the higher time frames, but we’ll include all of these just to be sure.
With our price study, what we’re hoping to do is just pick a starting point, wait for price to move a certain number of pips in one direction, then measure how far it continues to go in that direction. To keep things simple, we’ll make our starting point the close of a candle (or bar). This could be ANY candle, it doesn’t have to be at the bottom or top of a swing, it doesn’t have to be a long-bodied candle or a doji, it doesn’t have to be reliant on already being in a trade or not. So, for each of our chart periods (5min, 15min, 30min, 30min, 1hr, 4h and daily), we will randomly pick 5000 candle closes over the 4 year history that we have. These candles could be during peak trading times, just before or after news announcements, we don’t care. If we can keep the starting point as random as possible, it means it should be easier for traders to follow any strategy that we can derive from this study.
The idea is that we take our starting point, then wait for price to move to a certain trigger point before we enter. We could go crazy and analyze 20 different trigger points, but that would just provide unnecessary information and overload our brains. I’ll keep things relatively simple and just analyze 3 different trigger points: 10, 15 and 20 pips from our candle close. I don’t really want to go any lower than 10 pips because spread becomes more of a factor and depending on your broker’s spread – traders will obtain different results.
Once we’ve done our ‘virtual’ trade entry at the trigger point, our study will calculate how likely it is that price will continue to various price levels. I will analyze and report probabilities to the following levels PAST the trigger point: 10 pips, 20 pips, 40 pips, 60 pips, 80 pips and 100 pips. So, those pip levels are target levels or the number of pips we would make from the trade if those target levels are hit.
We also need to consider a stop loss level. Once again we could go into a gazillion permutations by analyzing 20 different stop levels with the above targets. That would be madness, so we’ll just consider 2 different stop levels: 20 pips and 40 pips. The stop level would be where the stop is placed from our entry point (which is the trigger price).
In summary, the plan that we’ve mapped out in this article is to:
Take EUR/USD 1 minute data over the last 4 years
Pick 5000 random candle closes for the 5min, 15min, 30min, 1hr, 4hr and daily chart periods
For each of the above samples, test entry levels at 10, 15 and 20 pips past the candle close
Using stop levels of 20 pips and 40 pips
and calculate some probabilities of price continuing on to 10, 20 , 40, 60, 80 and 100 pip levels
That’s quite a bit to get through and we’ll have multiple pages of backtested statistics from this study, which I’ll present in the next article.
Winner’s Edge Trading, as seen on: