Well, this is something we hoped to avoid, but cannot. Some of you have noticed differences between the RSI calculations in FastWays versus those of FastTools and other programs. These differences are intentional, and the purpose of this note is to explain why.
Note that I do not use Trade, but Roland Schnable, one of my beta testers, has duplicated the results in Trade by using Buy = price up through 30 or 70 and sell = price down through 30 or 70.
One of the implicit assumptions of most programs is that there can be no whipsaws when using RSI. The following screen shot from FastTrack illustrates what this means.
Let's expand the section from slightly before 1/1/99 through mid-2000.
Observe that:
a) just about at the beginning of 1999 there is a sell signal generated
because the RSI falls below the top line,
b) the RSI goes back above the top line for a brief period near mid-year and
rises sharply above the top line in the third quarter,
c) the RSI moves below/above the top line several more times, but
d) the RSI does not go below the bottom line until the first quarter of 2000 and
then rises again shortly thereafter.
The question is whether or not those trips back above/below the top line
should be treated as buy/sell signals or not.
The view taken by FastWays is that these are in essence corrections of a prior
bad signal. FastWays would generate buy/sell signals for these
"retracing" movements.
In contrast, other programs would not generate a buy signal until early in the second quarter of 2000 when the RSI dropped below the bottom line and subsequently rose above it.
Note that by allowing for retracing FastWays will capture most of the sharp increase during the last quarter of 1999 and in the early part of 2000. Others will not capture this time period.
Which is right? Our choice is clear, we want to capture that sharp period, or in contrast avoid a sharp drop should the picture be inverted. But ultimately you must decide which approach is right for you.