This is a follow up from the 'Options Backtesting Software' thread, I'm thinking on a business plan and would like your opinions. Please don't hesitate to tell me if it sounds overly naive or optimistic Resuming the plan in a one-liner, I would call it "the make money button" 1) Put up a web interface where individual traders can try do what I do, only on a simulated virtual account rather than the real thing. Within the constraint of a $5000 account, they can choose: - Automatic trading: let the machine pick the optimal trades. - Manual trading: they can choose from a pre-filtered list of trades, then for each trade they get a risk/reward screen. The key component of this first phase is a button labelled "Make Money!". Question is, do they press it or not? 2) The goal is to have 1000 users satisfied with their portfolio performance (>= 30% annualized). If they are, then they have the option of investing those money (or more) in real life. The key component of this second phase is a button labelled "Invest Money!" If 1000 users press it, the process moves to next phase. 3) I don't have the software and infrastructure for handling a capital of $5M (1000 users x $5000), but I'm pretty sure that if I get to this phase, I'll find financing for it. Revenue source is trough commission only: 20% of the profits, no admin fee. If I get financing, then I'll put up the required hedge fund and the software backing it. 4) The initial 1000 users can invest for real using the same auto or manual system. Scaling up occurs: more invested money means more resources to improve the system (software, models, strategies), which means more invested money and so on. Let the bashing begin!
Noone replied, so I'm gonna continue myself. Sorry for lorbing The web frontend idea was specific to my trade flow, but I just had an idea. What about MAKE THE FRONTEND AGNOSTIC OF THE BACKEND? It could connect either to my options trading engine any other engine. Like, Stepandfetchit seems to have something too. And we might add more implementations. A thing I want to achieve is statistical independence of the trades: that is, they should be uncorrelated. This way, as long as the variance is finite, the mean profit adds up and eventually exceeds the variance. Think you make trades with a normally distributed PNL. Average profit is $1 and standard deviation is $3. This means, with 95% probability you'll lose at most $6 (2 standard devs), and with 99.7% probability you'll lose at most $9 (3 standard devs). The $1 you make on average doesn't seem overly attractive compared to the risk. But now, think you make 100 such trades and they're independent of each other. The average of the total PNL is now $1 * 100 = $100. And the variance of the total PNL is now (3 * 3) * 100 = 900. The standard dev is then sqrt(900) = 30. So now you make an average of $100 and with 99.7% probability you'll lose at most $90. Almost surely you'll make at least $10 on *every batch of 100 trades*. That's known wisdom: to profit in this business you gotta do many winning trades, even if small. But they need to be independent. One way is to do them serially in time, but that takes time. I'd rather do them parallel in time, but I don't have enough capital for that - therefore the need to associate with others. And if they have their own system, the better - chances their trades are independent of mine are pretty high.
Such a frontend could be entirely open-source and needs to be so - because it would require some degree of standardization / protocol agreement between providers.
Just running the improved version of my software on some 300 randomly selected stocks that my laptop can finnish today. I've got 9000 stocks though and need to make this happen faster. Think, try, win or lose - either way, get feedback. Finance or Pharma, same thing.
good luck with your project i hope you can make a friendlier interface than some automatic options backtesting software out there