Backtest and system is like head and tail of a coin. If there is no system then it's not a real backtest.
Thank you all for all your inputs. It is all very helpful. By the way, I am hoping that my brother will just help me for the fun of it. He retired and doesnt do much anymore, so he has the time . And just in case its works, then he can share the riches along with me lol.... He programs in C or C+ (cant remember which one) and not sure about Python. On a serious note, I do intend to only work with higher volume stocks, so that grabbing 1 or 2 shares should be fast I would think. I know there is no way to know what the volume will be for any given day, but if the stock has a week long or longer history of trading at 500,000+ volume, then I would assume that there should not be a problem of grabbing 1 or 2 shares fairly easily at open. I know that I wont land every one at the open price because of volatility, but some will benefit me and some will hurt me, but in the end, if I am letting the computer buy and sell on specific rules, hopefully it should even out with the sheer amount of trades that are being made. Just thinking out loud here. Of course I am hoping to do this with a broker that doesnt charge fees for stocks ( I will only be trading US stocks), such as TD that I have now, or something like Tradestation, but I dont know if they are able to be setup to do this?
I would love to properly backtest this as a 'live' real time system, but there is no way that I can think. I am hoping that there is a way to do forward paper testing with live data, that I can use for the next 6 months or year. But I do not know if any platform will allow paper trading in the extent that I am attempting to do versus one stock at a time testing.
Given that you do not know how to program and likely have limited capital (i.e. "grabbing 1 to 2 shares"), why not look for ways to trade that do not require the technical overhead of automated trading? I can almost guarantee your idea is not going to be net profitable when accounting for trading and opportunity costs.
I understand your thoughts, but buying 1-2 shares of 3000 different stocks adds up to a substantial amount of money, as they could be any dollar amount really and not just $1 stocks. I think I read that buying one share of each available stock in the market now would cost $100K or something like that. Point is, buying 1 share each of a 3000 different stocks isnt much different then buying 3000 shares of a $20 stock, except that my chances of win/fail are spread out MUCH more. I really don't want this thread to turn into a "your a newbie and your ideas will never work so don't waste your money" type of thread. I really just want you guys experience on what platforms/setups can physically accomplish it if any. And again, can it possibly be done with any of the brokers that charge no fees for stock trading, as that would be the only feasible way to do it due to it's nature. I think TradeStation has API abilities, so can it be done with them or TD? Thanks for everyone's inputs so far.
Your model or rather the slippage in the model was a problem. I agree that trading at the open is pretty horrific, after the opening auction there's about 4-5 minutes of almost complete silence even in names that have decent volume.
Brokers that offer "free" trades sell order flow and your fills will be even more terrible. Your idea sounds DOA but go ahead.
1. Will your brother write you the selection algorithm for which stocks you want to buy at open? 2. For placing orders you need to have an account at some broker that allows API trading and write to the API. Drop me a PM I could direct you or your brother to a way of doing that reasonably painlessly. 3. Placing thousands of orders takes a bit of time, even through the API. If the broker you're working with supports MOO, MOC, LOO and LOC orders then you could pre-place them. Of course, if your selection criteria requires you to do the selection right at the open, then that won't work. 4. TD Ameritrade has no-commission trading but I wonder if they would frown on you flooding a thousand orders into the system at market open.
I have an idea. How can I implement it? I learned that the first step is to test it manually. The advantage is that you see in slow motion what is going good, what is going wrong, when this happens and why this happens. Automation is used to many times as a blackbox: people have an idea and quickly program it to test it. They run it and only see the end result (if there are no errors in the programming). Manually testing shows you the behavior of the "algo" from start till finish. That's what you need, information from start till finish. The fact that you would even buy at the opening in a very volatile market, without even knowing when you will get filled, without checking volume...shows that your "idea, algo or strategy" is not thoroughly thought off and tested. I have seen already a few IT wizards trying to build a trading system, and many make the same mistake. They have an idea and start immediately to program. I worked the opposite way: first develop and manually test (and analyze the weaknesses of the system) till the results satisfy me. I have to because I don't know how to program. But that was my luck.