It has been a long time since i updated on any of my systems. I have currently been trading my systems for nearly 8 months now and i thought i would update with some results and commentary. I started with more strats, but i have removed 4 due to either flat performance or due to being to much as % of my portfolio. I have added some new strategies which were much more rigorously backtested and all of them are performing very well. The strats that trade the most have the least risk per trade. Ones which trade infrequently go for the bigger wins and hence have more risk on any given trade. My aim was to obtain smooth equity curve by combining strategies. Here is current equity curve (i started only with $10,000 allocated towards trading these) - these results do not include my discretionary trading. Out of these 8 strats: 1 has performed amazingly, 1 has good gains, 5 have positive gains but many of these are near to flat. One is losing strategy which has to be reworked because i know it can make money but it might have been curve fitted as it was the first automated strat i developed (was not wise about the errors in back testing). My next step is now to re-work two of these strategies by breaking them down to their simplest form and seeing if they still have an edge over 3 samples of data. So far i am reasonably happy with the results, all the later strategies developed have performed the best and i think that may have to do with less curve fitting and better ideas. I think the theme of the market is the most important factor when looking at the results. During the exceptionally low volatility we have had at certain periods over the last 8 months my strats performed the worst (as to be expected). During the volatility spikes they seemed to do very well. Traders need volatility to make money, cannot underestimate this point enough.
Trade all strats currently on: YM, ES, NQ, TF. Stopped trading the CL strat as it was too big % of portfolio. - if automated strats keep performing i will trade portfolio of stocks - system will probably scan 100-200 stocks and trade same strategies over multiple stocks (waiting for right software to do this, Multicharts is introducing portfolio trading soon).
Hi! What timeframe(s) do you trade? What are your strategies' timing edge based on? Price patterns? Statistical analysis? Do you derive your edge from observation and testing or do you use some type of machine learning? How complicated or simple would you say your rules, components and parameters are as far as sheer quantity and complexity? Just curious...
I have posted my automation journey in a few other threads. I was a discretionary trader for years so i try to take some of those ideas and program them (sometimes this is very difficult to do). All my strats have to make sense and can't just be a random idea. Basically like any other types of strats i use momentum, mean reversion and trend following strategies. I like to use different types of themes to my strategies so they become less correlated. Also, because my automated account started at a puny $10k i did automate hedges if i was long or short too much in one direction - i wanted to limit drawdown as much as possible and eek out small consistent gains. I don't do any super short term strats e..g HFT. My least exposure to market strategy has on average 45 bars in market (1 min bars). Some of my strats start on the open and some trade after the opening range is in place. I don't think i have any type of arbitrage, 100% guaranteed edge or any insider knowledge. I am simply trading what i perceive to be price action (every trader has a different definition of price action). I do not use indicators to form any of my strategies (however i do use a moving average in two strats as a filter). I figure market is mostly random noise (intraday) which small time traders can take advantage of - we will never know the inside news or secrets but what we do have is the ability to be nimble and move in and out of positions like a ghost. Big traders can never have this ability and i think that's where our edge lies. I trade mostly 1 min charts and 5 min charts. I would love to automate swing trading positions across multiple markets and stocks and i really think that would be way more profitable and easier to do so, however i have limited account size for my automated trading. I don't use any machine learning or genetic algorithms. I simple take a discretionary idea and begin a test on it on a sample piece of data. If it beats the test and shows positive expectation then i begin work on strategy modifying entries, targets and stops. Test it on various samples of data and if it shows a positive expectation and has Profit factor > 1.9, more than 300 trades over 6 years of data and a good sharpe ratio then i will trade this live and track it compared to how sim entries would have performed. I find far more slippage on my market orders (sometimes 2-3 ticks so bear that in mind). My rules don't seem very complex, most of my code is 600 - 1000 lines. Nothing exceptional, nothing super complex or mathematical. There is no guarantee my automated success will continue but i now know that it is possible to be profitable trading in this manner. I simple now have to re-modify some existing strats and retest to see if they were curve fitted. Then i need to increase this account size to start taking advantage of trading portfolio of stocks. I have about 15 strategies which i have created which i don't use (i still continue to monitor and log sim trades in them however). Some i don't use because ones that trade Gold and Oil simply become to large a percentage of portfolio, others are too volatile. Also, i only trade during RTH and no holding overnight. Want to leave strats on 24hour but my experience something can always go wrong when you are away from computers. If i have to be away from computers i connect remotely via my Samsung Note 3 or my tablet and access my desktop easily. I have email and text alerts from any positions i have on. The biggest problem i have found is when data drops and my multicharts then has missing data which has affected some trades in real-time. Also desktop had habit of freezing in the summer (i found due to overheating as my desktop was overclocked significantly). If i had the money and an assistant i would trade multiple markets 24 hours a day (but someone needs to monitor just in case of error).
Thanks Dhalsim, I eventually want to trade in a similar fashion as you described, but sometimes it's hard to estimate what's actually doable and reasonable to expect, so it's encouraging to know someone else is doing it. Btw, Op and other people reading this thread might find this ama interesting http://www.reddit.com/r/IAmA/comments/9s9d7/iama_100_automated_independent_retail_trader_i/ Even though he's more of a quant I personally picked up some stuff from it
That is my philosophy as well. I think us little guys can do quite well for ourselves if we dont try to take down the giants, but just feed on their scraps. So many people try to fight unwinable battles by taking on giants, when alls they have to do is be smart and selective in their battles. Sounds like you've done that, kudos to you my friend! I haven't launched my system yet (getting close!), but it is promising to see that someone has been able to make an automated strat work with so little capital. Great job!! I plan on starting in the 20-30k range with my strats, which do use machine learning, and do make an attempt at trading as fast as my non-colocated, NxCore Data Feed, and IB commission structure will permit me to profitably.
Great progress Dhalsim. You should consider leasing a VPS in Chicago. Your automated trading would be less susceptible to power outages, Internet outages, and ... this summer heat you were mentioning. Cheers, Dominique