Algo Closing None-existing Positions

I am a newb, first and foremost, and starting out fresh. Currently I am trying to build out an Algo that simply follows a 30,60 MAVG.

Right out of the gate, my algo tries dumping position I do not (or should not) have; if my buy criteria is 30MAVG > 60MAVG, and the backtest starts out with just the opposite, I should NOT be sitting on any open positions. However, sell my close criteria is 30MAVG < 60MAVG, it tries to dumping my order amount, which in this case is +/-100%.

What is even odder is that I am getting some returns before any positions are actually opened, aka 30MAVG < 60MAVG (about 75-days into the backtest).

Any I missing something or is this something to do with the IDE?

Additionally, I am trying to figure out how to restricting the number of open orders to X so I am not constantly buying—even if X is a fraction of my overall portfolio—and the buy amount to the current value of said stock:

current_value = data.current(context.stock, "price")

order(context.stock, current_value*1)

Thanks for any and all help!

Michael,

Welcome.

Your algo seems to be doing exactly what it's coded to do, but maybe you didn't intend.

First, you questioned "my algo tries dumping position I do not (or should not) have". Your code has exactly two order methods:

order(context.stock, 1)  
order(context.stock, -1)

Those methods buy and sell, respectively, exactly one share of context.sid (in this case sid(14001)). Selling a share of a stock you don't hold assumes you want to short that stock. The algo isn't 'dumping' stocks you don't own it's simply taking a short position. You probably want to use one of the other order methods like order_target_percentage and then couple it with a check like

if context.sid in context.portfolio.positions:  
    order logic....

Also note that the 'handle_data' function is automatically called every minute. Your algo tries to buy or sell exactly one share of context.sid every minute (depending on moving_avg30 > or < moving_avg60) . Look at the transactions tab when you run a full backtest and you can see better what your algorithm is doing.

Another note. The poor performance (almost a straight line down) is probably because of commissions. Unless you specifically set a preferred commission the backtest defaults to $0.0075 per share with a $1 minimum cost per trade. Trading 1 share at a time costs you $1.00. Sometimes for initial backtesting it's helpful to set the commission to $0 (just remember to set it back to something realistic unless you are trading on Robinhood).

Good luck,

Dan