How did the guy get caught? Did they recognize the fact that someone downloaded the software and back trace it to him or did he just talk too much?
Or do you just program it to search for certain conditions like stocks that fall with certain moving averages or within certain points on an indicator? Anyone ?
I just googled on that and it looks like he was arrested/sent to jail and appealed and got out... http://www.aleynikov.org/ https://en.wikipedia.org/wiki/Sergey_Aleynikov http://www.forbes.com/sites/billsin...onvicted-for-trading-code-theft/#772f613e3d59 The Last Day at Goldman Sachs Beginning at approximately 5:20 p.m. on June 5, 2009, Aleynikov’s last day at Goldman Sachs, he transferred substantial portions of Goldman Sachs’s proprietary computer code for its trading platform to an outside computer server in Germany (all of which was accomplished while working at his desk at Goldman Sachs). Without informing Goldman Sachs, Aleynikov encrypted the files and transferred them over the Internet. After transferring the files, he deleted the program he used to encrypt the files and deleted his computer’s “bash history,” which records the most recent commands executed on his computer.
Too big to explain. To put it simple, 1) You should have good/strict entry/exit rules, 2) Program/logic to scan/look for the entry parameters and order executed with stop/take profit levels predefined or calculated based on your account/order size etc. 3) Exit when the conditions met. (moving average based tradings systems never worked (based on my testing) at least in the long run).
I'm a bit surprised and amused by this statement. Firstly trading competitions prove nothing. They are normally a very small sample - very short periods of time. They don't give you statistically significant results. Secondly languages aren't profitable. The trading systems they run are what make profits. Most systems can be run on any language. The reason you use one language over another will depend on which of these factors is important to you: ease of learning or coding, robustness, availability of libraries, speed, support, ... and so on. Arguably C is the most profitable language, or perhaps assembler / machine code ,since you can't run an HFT system on anything slower, and HFT is the most profitable method of trading. But anything slower can easily be run on python, java, R, ... you name it, as long as its turing complete. Python, java, R, C++, Matlab are probably the main languages that people use (not in any particular order). Each are good for different purposes. Now I wouldn't personally use R to run an automated trading system. I have in the past, running a portfolio with a few hundred million. But I wouldn't do it again. It's absolutely fine for developing a system, ie. backtesting. It's pretty popular amongst the AI, DL, ML community, probably equal with python: https://opensource.com/article/16/11/python-vs-r-machine-learning-data-analysis But the type conversion, error handling and memory handling make it insufficiently robust for looking after my money. You can write robust R code, but it's much harder than writing robust python. The way it does objects is ugly as well, but that is personal taste. GAT
Are the checks and balances that broad at trading firms and other large organizations like Exxon or Lockheed Martin? Do they check computers to determine if something has been downloaded or do they check outgoing emails and investigate suspicious addresses? I wonder if GS is more of an IT firm as much as it is a brokerage firm? I wonder how much money (as a percentage) they make from HFT and using IT, as opposed to fees earned from private stock sales like they did 50 years ago. Do your programs ever make buys and sells on their own or just search your parameters? How effective are the programs you buy? Are they significantly better than doing your own searches on Schwab or TD Ameirtrade? Thanks
1. I would guess big companies will/should have good checks/balances. 2. I would guess so. Nowadays anything you use in office or where ever be logged/tracked. 3. No idea. 4. It (programs) can be done to trade or just give alerts after sensing the opportunity. 5. I have not used Schwab or TD. Not bought any system but have developed for my own use and for few clients.
"They don't give you statistically significant results" - R systems made more profit on an equal basis, some would argue that is statistically significant. "Secondly languages aren't profitable" - fundamentally incorrect, if the language does not allow you to deploy business rules in a seamless manner for your model then that impacts your profitability. You are looking at it from a technical perspective, explains the dislike of R and remember you are running automated systems at 2%/mth. Some people use the languages on a fundamental level, others combine technical and fundamental models. The more code you write the more 'robust' a language you need, Buffet doesn't use technicals but is one of the wealthiest on the planet. Last I heard one of the big international banks was still running my code five years later for their group consolidated trade analysis, because I minimized the code base to the absolute minimum. Less code means more stability, but very few know how to do that, you actually have to use some intelligence and combine fundamentals, a lost art these days. Which is why you now have two sets of people, the quants using 4GLs like R and python and the programmers using 3GLs like Java and C, then everyone wonders why we have flash crashes and never the twain shall meet. The technicals will 'always' leave a hole that the fundamentalists can exploit, it might be microscopic but it's there, but then I only made 20%/mth in Q3 exploiting those holes. Java and C do not allow time to market for new business rules, you are discounting that the OP doesn't have an army of programmers, which means 4GL is preferable to 3GL on a business level until they have a stable methodology. If they're a technologist it's the other way around, but as they said Python it's smart to assume it's the former. One more trying to generate a different conclusion to fit their subset view of the summary, what a waste of time. It would be fine if people reading actually learnt, but given the last Random 50/50 thread where provided a complete matching summary-detail-conclusion, everyone's still blank because it doesn't match their incomplete (lack of Theory of Mind) view on the world. Doesn't mean they don't make money, but it's hard work and unstable when volatility increases as the fundamentalists exploit the holes in their believed to be perfect tech solutions.
There is also this guy. https://www.ft.com/content/ea0438e0-2186-11e5-aa5a-398b2169cf79 G-Research do not fuck around when it comes to these matters.