Hey Thomas,
Nice :). Your algorithm is great as you have it, and I have a few suggestions that you might be interested in. First, it you may find it easier to use an order_target function like order_target_percent instead of just order. This way you don't have to check if you own the position in the if, you can just set the target for .1 of your total portfolio. But you are obviously doing some other stuff here, so the way you are doing it may be ideal. You can also remove the lines about the notional — those are left over from the example algo, and aren't currently doing anything in your algo besides being defined.

Another thing, the way you have it now you don't actually need to import talib. This is because you are using the built-in talib functions, not the ones imported from the module. Although we are deprecating the ones you are currently using, they will still work, but they can be very buggy at times (which is why they are being deprecated). See my example here using RSI in talib. There is also a 26 day warm-up period in your strategy that you will have to account for if you use this.

You might also find the record feature interesting. But again, nice strategy!

Gus

Thomas,
Welcome to Quantopian, I took some time go through your algorithm. I got it using minute data and talib rather than the ta transforms. I also found an accounting error that was allowing you to buy more securities than the imposed limit of 10.

When handle_data is called, context.portfolio contains the current account values as of the time when the function was called. The portfolio variables will not be updated until handle data is called again. That means context.portfolio.cash will not change within that bar, even if you place several orders. You have to do a little extra work to track any money spent within handle_data calls. Here is an example of how you can do that.

# Managing cash spent within a call to handle_data

cash = context.portfolio.cash # Cash as of the beginning of the handle_data call

for stock in data:  
    price = data[stock].price  
    if some_buy_signal:  
        order(stock, 10)  
        cash -= 10 * price

    if some_sell_signal:  
        order(stock, -10)  
        cash += 10 * price  

The algo respects its 10 stock limit now, so the rest of your logic was good. It trades less often now also, only when a new slot opens up. Maybe you can add something to select the 'more ideal' stocks to fill those open slots, whatever that might mean : )

happy tinkering,
David

Thanks Gus & Dave! There's a lot more I want to do with this algo but just haven't been able to get back to it in the last few days. I'll definitely incorporate your suggestions and I will definitely be filtering based on the best stocks fitting MACD crossover criteria. Hopefully I can reach out to you when I get stuck again!

Hi Calhoun, I am new to Quantopian. I cloned your algorithm and trying to understand it. This might be a very naive question :) You have "9.99" in the code, is this the trading fee? In the comments, you mentioned $5 as trading fee.