The concepts you presented are not new: splitting a position is features in many backtesting packages, reversing position is an old trick used by some, scaling in and out is a very old technique and used pretty much by anyone with some size. I don't know what "flatdetect" and "switch ticker" are though. Here's the thing, you come from a disciplined background and very likely are a quite proficient programmer (writing C+ isn't for everyone). The problem arises when your approach is systematic but then you start talking about maybe stopping it when it's down -15% or perhaps the puts would help the strategy, there's a lot of if's and but's as it's all conjecture. You need to quantify and test your assumptions and that's where the real data comes into play. Random data has no place in real life scenarios. Testing on random data series is nothing new, it's been around for over a decade at least and all conclusions I've seen point to it not being relevant - not inferior or better, just different to the point it doesn't help you design something that works. But yes, we all have opinions. Look at it from an investor's viewpoint, you present this alpha producing strategy with incredible returns but don't have access to even 20k anywhere? Logic would say you'd sell your car/mortgage your house/borrow from family/use a credit card if you truly believed it will produce these great numbers. Now if you aren't sure about the performance, then the logical choice is to sell the system because you have no risk.
Given that you are selling the system, the burden of proof is on you to prove it's valid for the real world.
You should have read more carefully what the requirements of the system are. Trading $20k won't be enough. The system is for big accounts, ideally $600k or higher. With smaller accounts (>= $100k) it works too, but the MDD is for my own taste a little bit too high, ie. risky. The higher the acct the less riskier is it, ie. smaller MDD = smaller risk. If I had the play-money myself I never in my life would offer this system here or anywhere else; I would keep it my secret. But the crude realities cause me to find an alternative for seed-funding. I'm not selling this system, nor disclose its inners: it is offered only for lease on an annual basis (negotiable), and only as a black-box system. I could also be interested to join a company / firm, instead of starting my own in some years. Regarding the algos: I cannot go much into detail. Flatdetect just detects flat markets (important because of time-decay with options), switch-ticker might be specific to this system, it just closes the pos and "continues" with a different ticker due to flat market with the first ticker, etc. ("market" in this context means the single ticker (aka the underlying) each bot trades). Regarding market crash: my statement does not imply that the system will behave bad or very bad in such a scenario. The -15% trigger is just a generic example that could be used by any system. Since this system trades both Calls and Puts, in a market crash situation things could simply cancel each out on average, meaning: no difference in the result as Calls lose, Puts win...
I have read the conversations carefully. My suggestion for the 20k was not for trading but for buying real data. Leasing the system is not that different from selling it as anyone using it will affect edge and liquidity. I'm only poking holes in your method because I think there might be something there but you're only 25%, maybe 50% there.
Out of curiosity, which country are you from? Regardless, now things are getting clearer: so something around $100/month for data is outside of your reach right now. I can respect that. That obviously also rules out $15-20K for backtesting data. (Options are incredibly expensive! I had no idea.) In my opinion then, you should first seek a sponsor for test data, then with that history to boost your credibility*, worry about finding capital. Whether that sponsor supplies just a forward-testing paper account or back-testing data is up to whoever will take this chance. *Credibility: it doesn't matter whether, in your opinion, live testing is useful or not... so long as it does matter in the eyes of your eventual investor(s). Once you get actual market tests accumulated, you might also want to build a more formal prospectus-type letter, proof-read by a third-party. On something this important, you can't afford an imperfect image. If I had the $600K to spare, I'd definitely throw $100/month for a few months to help you iron out the kinks live... Hopefully someone around here will have a similar level of interest. Good luck!
...and on this point I forgot to add: I'd refrain from ad hominem attacks, if I was in your position. Calling one critic "stupid", telling another he "should read more carefully", those are very abrasive responses. You're a seller right now, marketing your system, and thus you should be composed as such. Just ignore @Chubbly if you're out of polite answers for him; that's what that button's for.
Ok, the source code of my GBM routine can be found here: http://www.elitetrader.com/et/index...-code-in-c-for-generating-time-series.297370/ Let me know if you find any bugs. cu
I'm a non-native German, ie. foreigner, living in Germany. You can imagine how my chances are to convince local people, especially in these times... I am afraid I've the fate of pople like Louis Bachelier and Vinzenz Bronzin, living in wrong times and country... And English is not my native language, but that's not a big deal, I have no problem to read and understand any academic paper/book or news articles. I'm more like a researcher, not a business man, so self-marketing is unfortunately not one of my strengths. I was hoping to convince people with the result of my work, not with any fine-printed prospectus etc. Thx for the tips and suggestions, I'll try.
I checked you GBM code and it looks good. Now we need to review your option pricing code and verify that you applied the same parameters (volatility, drift, time structure dt, etc.) that you used to in GBM. Would you mind publishing your option pricing algorithm? If you used the standard black-scholes formula there should be no proprietary IP in it.