How B# Works

HOW B# WORKS - OPTIMIZED PSAR

B# uses the Parabolic Stop and Reverse (PSAR) algorithm developed by J. Wells Wilder Jr. in 1978. PSAR generates a long/short reversal line based on stock price movement using two mathematical parameters – a STEP ratio between points and a MAX, if incremental steps have not triggered a reversal (ref: New Concepts in Technical Trading Systems, 1978, J. Wells Wilder Jr., Trend Research, pp 9-20).

As per the original levels recommended by Wilder in the reference above, PSAR calculations, STEP is set at 0.02 and MAX is set at 0.2. Unfortunately, although PSAR curves developed with these fixed STEP and MAX levels are visually appealing; in practice, they rarely provide profitable trading direction. Wilder suggests that PSAR is best suited to trading within a defined trend (perhaps as a safeguard if fixed STEP and MAX values are used).
 

B# implements a very different approach in that in order to determine the optimum STEP and MAX values, B# implements a brute-force look-back virtual trading algorithm that calculates the P/L of 491,469 scenarios (STEP-MAX pairs) over one year, by gradually incrementing STEP and MAX values for each scenario. The most profitable back-tested scenario is assumed to have the optimum STEP and MAX values going forward. 

The virtual internal B# "trader" starts with an initial “investment” of $1000 and "pays" a 0.25% “fee” for each virtual transaction.
 

B# presents these optimum values as well as the results of the most profitable scenario, along with the timing of the virtual trader’s buys and sells. When taking real world trading positions, it is particularly valuable to know the most recent Long/Short position of the virtual trader, the P/L of the scenario and the PSAR trend.
 

The quality of this optimized scenario is provided in the [ScoreCard] where each Buy/Sell transaction pair can be studied in more detail. The quality of the Buys and Sells are also presented in the Charts where a profitable Buy/Sell is labelled as a Good Buy/Good Sell. A good Buy is one where the execution price is lower than the previous Sell (this is almost always the case). If the transaction has been loss making, it is labelled as a Bad Sell in the charts. The effectiveness of the B# algorithm is also given in the [Results] screen where the number of transactions and the percentages of Good/Bad Buys and Sells is listed.