Bollinger Bands #3 Some Improvement

I elected to run the Bollinger Band simulation again, this time not allowing either balance (USD or coins) to go negative. If the balance was too small to accommodate the recommended trade, the number of coins is reduced to an amount that will zero the respective balance. This resulted in the elimination of 87 trades because the required balance was zero. An additional 9 trades were adjusted downward to use the remaining small balance.

So, the question becomes, how did this change the profit/loss over the previous 7 quarters. The original run resulted in the initial $10,000 ending with a balance of $3,796. The new approach still results in a loss. The $10,000 here dropped to a value of $5,744. This is an improvement. There are 2 likely reasons for the improvement. First, the trades in general lose money, so it makes sense that if you make fewer trades, you’ll lose less money.

The second is more complicated. BTC prices tend to move in a single direction for a while. For example, there were 9 buy trades signaled in January 2019 as the price fell from $3,683 to $3,414. Since the price was falling, the best approach would be to only make the last purchase. But that would require hindsight. The second approach only made 4 buys since the USD balance reached $0. This effectively saved 5 questionable trades.

Going forward, I will focus on the 2^nd simulation. All simulated trades will consider the required balance (either cash or coins) before executing the trade. The next effort will begin the search for a profitable Bollinger model by varying the model parameters.  It will be out soon. I’ll use Twitter, @billlanke, when I post.