Luke wrote:I'm sure most have already noticed this but for those who haven't:
Each FPI number can only be calculated by the corresponding Bid/Ask price or the FPI number is meaningless.
So a BBS FPI is calculated by Ask * Ask * Bid & a SSB FPI by Bid * Bid * Ask.
This makes it impossible to test on historical data unless one has data with Bid/Ask.
I have found that using the price median ((bid+ask)/2) is actually giving me more realistic results for the fpi. When you use the b,a,a or the a,b,b, you'll notice that most pairs with high spreads will hover away from 1. If you'll take a look at the FPI REAL TIME, you'll see how close to 1 the fpi stays, especially in quiet times! That's because it uses the price median. Now, understandably it would be hard to understand that this can provide value. The value is in knowing how far off the fpi is - in the real world. One way to use this would be to calculate the minimum amount of fpi travel need to break even ... etc. We have to remember that when dealing with fpi - it is a way to make sure that the quotes are within a reasonable distribution. However when it comes down to bottom line profits, a 5 point fpi move in any given ring will probably not equal the p&l of a 5 point move for that same ring under different circumstances. Thus, we need to create a model to be able to estimate the minimum net profit with (x) amount of FPI travel. Unfortunately, just by calculating fpi using bid ask ask or visa versa, it doesn't mean that you'll breakeven if it returns to 1... heck, you won't even know where  really is!