EDIT: 7/20/2014: The trading tools and dashboards are no longer part of the site.
Now that I’ve finished the series on the Relative Buying Power tool, let’s go back to the Weekly Projections tool and take a look at some recent upgrades. First, I’ve added four more currency pairs to the projections; USD/CAD, USD/CHF, AUD/USD and NZD/USD. Yay!
Now for the “mathy” part of the post. Double yay! Here’s how I calculate the weekly accuracy scores for the projections. On the Free Dashboard, the prior week’s projections are shown as dotted lines on the charts, with the actual price movements shown in red. So the question is: how do we put a numerical value on how closely the red line matched the dotted line?
First, I look at the extreme high and low values on the chart, whether they’re part of the projection, the actual price movement, or a combination of both. This gives me the overall range of the chart in pips. Then I look at the discrepancies between the projected and actual events. Note that there are four events on each chart; the open, the first extreme (either the high or the low), the second extreme (if the first extreme was the high, this is the low and vice-versa), and the close. I then divide each of the latter three discrepancies by the total range of the chart to get a percentage error. Finally, I subtract the average of these three errors from 100% to get the final accuracy score.
Take note of a couple of things here. First of all, I don’t bother looking at the discrepancy between the projected and actual value of the week’s opening price because there usually isn’t any difference. Since the open is usually equal to last week’s close, it’s obviously pretty easy to project, so that would be cheating!
Also, note that I’m not always comparing the projected high to the actual high or the projected low to the actual low. Instead, I’m comparing the projected first extreme to the actual first extreme, and the projected second extreme to the actual second extreme. So if I projected that the low would happen first, and it turns out that the high actually happens first, then the discrepancy is between my projected low and the actual high.
Since I’m The Capitalist Trader, you can be sure that there’s an example of all this in our future somewhere. And here it is!
Let’s suppose that my projection for the EUR/USD is:
Open – 1.4000
High – 1.4300
Low – 1.3950
Close – 1.4050
The week goes along, and the actual price behavior is:
Open – 1.4000
Low – 1.3850
High – 1.4250
Close – 1.4100
First, the range of values goes from 1.3850 (the actual low) to 1.4300 (the projected high). This is a range of 450 pips which becomes the denominator for our three error calculations.
The error for the first extreme event is the difference between my projected high (which I projected to happen first) and the actual low (which actually happened first). As a percentage of total chart range, this error is:
(14300 – 13850)/450 = 450/450 = 100%
This is as bad as it can possibly get! My projection for the first extreme and the actual price at the first extreme are at opposite ends of the chart’s range for this event. Bad Cap!
The second extreme and the close errors are calculated in the same way (note that we’re always using the absolute value of the difference between the projection and the actual price; we don’t use negative numbers):
(14250 – 13950)/450 = 300/450 = 67%
(14100 – 14050)/450 = 50/450 = 11%
So the error in the second extreme was 2/3 of the total chart range while the error in the close was only 11% of the range. Subtracting the average error from 100% gives:
100% – (100% + 67% + 11%)/3 = 100% – 59% = 41%
Voila! The accuracy for this projection was 41%. And that’s where the accuracy numbers that you see each week come from.
That’s it for now.
As always, stay tuned and…keep pipping up!by