I think 20 minutes delayed data is fine. If your results are still good, you might have to spend a few months testing with real data (and you might have to make some investments for that), but it will be worth it because you know the data being delayed wasn't an issue. If the results are bad with the delayed data, then you know that the strategy never really worked and you can move on.
The problem with backtesting is, even if it starts from arbitrary time points, that one in advance knows the data, and can be accused of having made use of that knowledge by doing curve-fitting... It is better if the system gets tested with realtime data. For this I prefer GBM (mathematically simulating stock behaviour; https://en.wikipedia.org/wiki/Geometric_Brownian_motion) because one does not have any prior knowledge of the data, and a test can be performed very quickly, and one can do as many tests as one wishes. But many people don't know what GBM is, they think it's fake or made up data, or some even think the random number generator GBM uses could have been manipulated... ie. leading to the same accusations... Btw, in my testing framework GBM data is not generated as a time-series vector a priori, rather as single data, ie. a single bar gets generated and immediately processed by the system, and then that repeats... Ie. no chance for the system to know the time-series data in advance. People interested in GBM can study and try the source code here (use the last version posted): http://www.elitetrader.com/et/index...-code-in-c-for-generating-time-series.297370/
If one tries hard, one can do; I don't see any problems, it's possible. One just needs to memorize all the peak and through points... Maybe a solution would be to shomehow "mask" the data, ie. add a constant value or multiply by a constant factor... But I'm afraid even that is "decipherable"...
You're abusing the weiner process. A through understanding of brownian motion can be used to legitimately exploit a non-casually occurring tiny artifact within real data but BM is an insignificant happenstance byproduct of the algos which create real data. Despite your change to a more agreeable and marketable demeanor it's obvious you don't understand what you're saying and doing. This doesn't mean you don't have advanced skills and knowledge because you obviously do but the premise behind all your knowledge is tainted by this flaw which will forever prevent you from covering commissions and slippage when your ideas are tested on real data. You're attempts to obfuscate the reality of slippage by introducing limit orders within a highly illiquid market adds credence to my theory your ignorance is a cover for evil intent. It's absurd for anyone with even basic knowledge to be replying to you.
One possibility is actually he is promoting here the collective2 service, as he claimed/confirmed he is currently paying ET for his system in a fairly subtle approach! It seems he said he knows options well because he traded options and his data is derived from his experience of trading options. He offered comments on options to other ET posters. However, it also appears that he at the same time he said he do not have any access of actual options data for forward test. Generating random data for 50 stocks in order to closely simulate the actual market data therefore $600,000 funding is required as initial capital is a funny idea for the purpose of minimising MaxDD, imo. It's totally unrealistic as all stocks will have high correlation in a strong trend. An unimaginable loss then could be incurred to his investors!!!!! The $600,000 or even multiple times could be easily gone in the first days of trading! He is actually providing a trading signals service through auto-trading, however charging a profit sharing fee. Besides any potential fees he could gain/get from a preferred broker he later choose. His excuses have been his worries of accusations from others. Hence he did not answer others' questions including mine, but he was angry at other posters if anyone did not answer his questions! Perhaps not an average snail oil salesman!
It's not about being "fake", it's the fact that any generated data does not reflect the market forces. You do understand that market data isn't just any generated data series but reflects real supply & demand and the panic selloffs due to fundamental shifts in the economy? Add to that the situations where correlation in equities is very high and therefore all your positions will be concentrated...and the inevitable harsh drawdown that fill follow.
Someone here has accused me I would make ad for C2 in this thread b/c of the thread title. Somehow that posting vanished. Too bad, there were some other points therein I would have responded to, but I unfortunately had to do something else, and when I came back the posting wasn't here anymore.... I'm in the following dilemma: I cannot talk freely about my system in other threads than the original thread, because then I get accused of making ad for my system. OTOH the original thread has been locked by one of the admins because he says it has been discussed enough. So, the only option left is discussion via PM. Just drop me a PM if you have any further questions regarding my system. I can give you also my email if you wish a direct contact. Thx.
Funny: if I'm logged in then I can't see the posting of OddTrader, but when I'm logged out then it shows up! I should check my Ignore-list, but I think I didn't add him or anybody else... will check... As said: my Ignore list is empty, so there must be bug in this web-system...