Here's the disconnect. You don't want to obtain any live data for it, not even 1yr for those 67 underlyings. $15k or even $150k isn't too much if your system will return 1000% easily, everything requires an investment of some sort. Also, having a backtest with various market periods (like 2008) is much more valuable than a backtest on generated data. I'd be interested but I think in the real world, you won't do well.
When will you show a backtest against real data? I have shown much more proof that my system works than you have. I have shown every transaction for 11 years which includes the crash of 2008. I can only conclude you won't ever test against real data A) You can't because you know your blackbox will fail B) You don't know how to do it Either way you don't look very smart. It is obvious that your posts are getting more abusive because you have been revealed for the fraud that you are. I will continue enjoying my real money that my system is making.
Could you just forward test it with data. I mean IQFeed that I use is like 70$ a month plus exchange fees for realtime. If you don't mind delayed, 70$ a month. They have a 500 realtime symbol limit but that is just realtime. I use amibroker and it's able to change my realtime watches and adjust itself(removes oldest watched and adds new). They don't provide their api for free, its 300$ a year. 300$ a year is just for yourself to use API. I use Amibroker and they provide me the plugin they developped for IQFeed. I don't pay for the API as the API is usually given to the commercial software developpers as its in IQFeeds benefit to gain customers. I don't normally use tick data but it has 180 days of tick data. Historical data ranges depending on markets. I haven't tested but I have downloaded 5 years of minute data with amibroker, might be able to push more, just never tried or needed. They also have access to OPRA prices. This can get a bit complicated as I think they report each strike and expiry as its own symbol. Not sure(not an option guy). So for 70$ you get some basically endless data. I've built databases that are 26k symbols back 20 years with daily data(just to try) I keep the ones I use smaller, 1k-2k symbols. I don't trade realtime as I use daily signals at the end of the day. Their exchange fees are pretty much like anywheres that transfers over exchange fees to the client. NYSE is like 6$, Nasdaq is 6$, OPRA is 6$, the futures exchange fees can get pricey but you still get access to delayed and historical data. Key part is you need a software they support(see page) ,make sure x software has a way to export as I'm guessing you need to read this data in whatever you built. Amibroker is capable of exporting csv files of any data I get, I'm sure others do this aswell. Or you can buy their API for the year and go to town with it. I watched a video Adam Grimes did and it was a chart produced by a random number generator simulating a coin flip, this was charted as a lign graph. It is truly amazing how random can produce some very lopsided results and yes the chart could be seen as possible price data. It made some very very large runs in both directions for a 50/50 possibility. If markets are truly random, and they are completely efficient, in my mind any trader who makes money is basically a monkey throwing a dart and getting lucky. Their has to be some sort of inefficiency, a imbalance of selling/buying to make money with some sort of system. I just can't see how a system on random numbers can work. And thats going to be hard to sell to anyone. It seems you believe in this system but its going to be hard to sell if theirs way better options out there, from traders with proven statements of live trading. Especially when dealing with a large sum like that, theirs many many options.
Q Trading on Algos * Johannes A. Skjeltorp Norges Bank Elvira Sojli Erasmus University Rotterdam Wing Wah Tham Erasmus University Rotterdam and Tinbergen Institute Preliminary Draft Comments welcome Abstract This paper studies the impact of algorithmic trading (AT) on asset prices. We nd that the heterogeneity of algorithmic traders across stocks generates predictable patterns in stock returns. A trading strategy that exploits the AT return predictability generates a monthly risk-adjusted performance between 50-130 basis points for the period 1999 to 2012. We nd that stocks with lower AT have higher returns, after controlling for standard market-, size-, book-to-market-, momentum, and liquidity risk factors. This eect survives the inclusion of many cross-sectional return predictors and is statistically and economically signicant. Return predictability is stronger among stocks with higher impediments to trade and higher predatory/opportunistic algorithmic traders. Our paper is the rst to study and establish a strong link between algorithmic trading and asset prices. http://www2.warwick.ac.uk/fac/soc/wbs/subjects/finance/fof2014/programme/elvira_sojli.pdf Table 3 Algorithmic trading and stock returns: Uni-variate comparisons The table shows the average monthly returns for stocks cross-sorted by AT and dierent characteristics. Each month t we divide the sample in terciles based on end of month characteristic (size, book-to-market, relative spread, trading volume in USD, past month returns, and past 12 month return). Within each characteristic we sort stocks into ve AT portfolios, where the AT1 portfolio contains stocks with the lowest AT and AT5 stocks with the highest AT. We then compute the equal-weighted average return over for month t + 1 for the ve AT portfolios within each characteristic and the return dierence between the low and high AT portfolios. All returns are calculated using bid-ask midpoint prices (adjusted for splits and cash distributions) and corrected for delisting bias, -30% return for stocks with delisting codes 500 and 520-584. t-stat shows the t-statistic for the dierence in returns test for AT1-AT5. Average monthly (%) returns (t+1 AT (t) AT1 AT2 AT3 AT4 AT5 AT1-AT5 t-stat Panel A: by MCAP Low MCAP 2.67 2.17 1.67 1.39 1.21 1.46 5.20 Med MCAP 1.39 1.03 0.69 0.86 0.66 0.73 1.46 High MCAP 0.55 0.42 0.63 0.54 0.57 -0.01 -0.03 Panel B: by BM Low BM 1.06 0.89 0.68 0.63 0.61 0.44 0.95 Med BM 1.54 1.17 0.81 0.79 0.74 0.80 2.20 High BM 1.96 1.74 1.34 1.38 1.02 0.94 3.40 Panel C: by SPREAD Low SPREAD 0.76 0.72 0.77 0.68 0.44 0.31 0.67 Med SPREAD 1.69 1.25 0.77 0.65 0.67 1.03 2.49 High SPREAD 2.58 1.98 1.24 1.26 0.81 1.76 6.04 Panel D: by USDVOL Low USDVOL 2.44 1.91 1.33 1.25 0.86 1.58 7.28 Med USDVOL 2.08 1.27 0.83 0.93 0.72 1.36 3.09 High USDVOL 0.61 0.64 0.59 0.61 0.43 0.19 0.35 Panel E: by R1 Low R1 1.36 1.30 0.94 0.91 0.86 0.49 1.15 Med R1 1.51 1.14 0.98 0.94 0.68 0.83 2.77 High R1 1.51 1.19 1.00 0.89 1.11 0.40 1.09 Panel F: by R212 Low R212 1.59 1.48 1.02 0.99 0.80 0.79 2.09 Med R212 1.30 0.99 0.84 0.87 0.82 0.47 1.85 High R212 1.36 1.35 1.14 0.97 1.16 0.20 0.46 UQ
As the name of this thread suggests, this thread is about the system named "sys13". Everything else is off-topic here, and should get out of here.
off-topic stuff snipped by Thread Opener (TO). Sorry, just a simple question: What is the above table data supposed to tell us? Has it anything todo with the topic of this thread? Please post such stuff somewhere else, it is off-topic for this thread.
And My investment so far is mainly developing and testing the system with its underlying algorithms. As said there are some novel ideas in system development. It took me more than 6 months. Some examples of the algos I can mention: - scale-in - scale-out - switch ticker - reverse - split - distribute - flatdetect ... Some of them I have seen nowhere else nor read in any article/paper/book, ie. I assume they are new ideas by me. Regarding your anger about market crash scenarios: You can close all positions and stop the system anytime. For example if the MDD gets say below -15% or so. And, as already mentioned by me: a market crash can also be an advantage for some systems, especially if they use Puts... Regarding backtesting or forwardtesting with real data: I myself cannot afford the price for the data. As said, I'm using and even preferring GBM data because it is from statistical point of view more accurate than real market data. Meaning: it is more realistic, even if it might sound maybe silly, but it is true because one can generate as much data as one likes (and by it create countless market situations) than real data can only dream of. The game of this system is beating random processes if market rules are applied to the stochastic process. This question has been discussed, everyone has their own stand on that, so let's stop this useless point and continue with other questions.
The conclusion from your "tests" is that the black scholes theoretical framework does not work inside it's own framework. If your results are true, you will upend the entire options market (which is several trillion dollars in size) and probably win you a Nobel prize for Economics. In that process you will earn billions arbing the options markets around the world. Surely that's worth scraping together 15k to test on real data. Think bigger than earning 20% on a 600k trading account. That is if you believe that your results are accurate and reflective of real life...
Can you explain this a little bit? Yes, you could be right with this assumption. Thanks, this question has already been answered many times. 20% is peanuts, sys13 can make more than 1000% p.a.