Creating software to predict stock markets

Hello,

I am new to this thread, and i am fairly good with tech analysis. I am in the process of developing a software, currently in requirement gathering stage. I am taking conservative approach.

My intention is to store US stock data for 1 to 2 years. I will be storing 100-1000 stocks, probably in 15 min, 30 min, 1 hour, 1 day, and 1 week time intervals. From database perspective, we are looking under 50,000 rows. As for columns, there would be additional fields that could capture Technical indicator details ( 40 to 50 technical indicators ). Some of these indicators are custom TA indicators.

I would like to hear what programming language would you use? I also would need suggestions on databases i could use. Ideally i would want to create into software, which could be shared with others - either in app form, or executable form.

I appreciate your responses.

Thanks,
Nara

 

What's the plan for this? The market is crowded with trading software.

Is this like some sort of signal alert or are you building like a simple backtester?

The software that usually wins in this area is the one with the most freedom to let you do what you want to do and provides the best tools to make certain tasks easier.

Didn't really get the post. Theirs a lot of competition and plenty of great well developed products already.

 

This had been done countless many times in various forms and shapes. What's new?

 

"The software that usually wins in this area is the one with the most freedom to let you do what you want to do and provides the best tools to make certain tasks easier. "

I am working on predicting behaviour of certain stocks, based on certain algorithms derived from technical indicators which i have created, considering best big-O analysis.

The intention of developing the one from scratch, is because i find there is lack of freedom, for the indicators i work with. If i do input a stock, the outcome of my program would let me know the best TA combination that has worked in the past, and would use this to send alerts. The software is not being designed to buy sell shares. Again, the primary function includes backtesting, prediction of future movements, based on past movements. Makes sense?

 

You have a unique idea...if only it was 1987.

 

Honestly I don't even think this would be unique in 1987. LOL. Maybe next he is going to tell us he found a unique way to connect to the internet via dial up modem.

 

I still fail to see the novelty here. There are plenty of existing open source and commercial backtesting/optimization frameworks out there (Quantpedia, NinjaTrader, MetaTrader, WealthLab, TradeStation, MultiCharts, OpenQuant, AmiBroker, TradingBlox, RightEdge, just to name a few) which allow you to plug in your own (custom) indicators, backtest, optimize, send alerts, and much more. If you want to write one from scratch, no one will stop you, but it's a serious undertaking, if the intent is to compete with what's out there. On the other hand, if this is just a programming project for fun/research, then sure, go for it!

With regards to your questions about the programming language and the database, you will not find the consensus here. There are too many choices, and too many entrenched developers to argue about the superiority of a particular choice. I'd suggest using free and/or open source technologies, since you are not building an industrial strength platform.

What's the "big-O analysis"?

 

Another Green = Buy and Red= Sell.

No such thing as predicting anything.

 

You are predicting that the "green" follows green and that the "red" follows red.

 

You need to investigate Data Mining software and advanced statistics to determine which set of indicators may be reliable.

So you wish to out-perform the quants with PhD degrees at Point72 Asset Management ?
These guys are paid $300k per year plus a huge bonus for performance.