I ran a mainly automated system. Helped a ton in increasing size from what I had been doing manually. Track your trades with what your system would have predicted. This helps gauge correct slippage and quicker feedback on whether you should intervene or not. I usually beat my automated.
I run MC64 in the cloud and have made over 30% YTD. This will be my first year I'm expecting to have automation for the entire year so I'll keep you all posted.
I've run semi-automated (meaning automated signal with manual entry due to amp cqg mc agreements during 2013 prevented full automation through cqg's execution gateway) since PFG July 2012 bankruptcy and what I can say about pitfalls is that intraday trading can assume 2 ticks of slippage with fairly high roundturn commissions but trailing stops must have high targets to be accurate. In 2014 I will be completely automated with an E-Signal data feed, CQG execution gateway and Multicharts64 platform trading in a cloud at Vision Financial Markets. Results have been stellar so far at 30+% YTD and no more than 2 ticks slippage for live fills meaning it will match my assumptions if that were to continue. Overfitting occurs from trailing stops typically and I only use market orders. For stops, checking the close of a bar is the best way to have them built into the system. Checking each bar's closing price automatically sends market orders to act as a stop and that's what I feel the best way to do it is. Also done for profit targets and can also be done for synthetic trailing stops.
Thanks bwolinsky. I also run Esignal - Multicharts - Interactive brokers. Why don't you use IB as your broker and therefore there will be zero problem with complete automation. Couple of questions if you would be so kind to answer them: 1. Do you use Inter Order Generation on/off with you multicharts backtests? 2. Do you perform day trading or swing trading with majority of your systems? 3. Do you run strategies on Futures / Stocks / Forex? 4. Do you believe each market has its own characteristics so it would be impossible to find holy grail to trade them all? 5. Would you say your strategies are simple or very complex. 6. Do you have any method to turning or halting strategies which you find are not working? 7. What Profit Factor did you average in 2013? 8. If you rely on close on bars then i assume your stops are pretty much dynamic. Do they rely on price on the chart or are they often just levels such atr stops, pivot stops, support/resistance stops, percentage stops. 9. Do you believe in strategies which are buy on open of a day and sell on close of day or do you feel these give too much scope for draw down and have less predictive power? 10. Are you equally hedged short/long in your strategies? Thanks. Only answer what you feel comfortable with answering.
Per 1) It is always off. Basically, I run indicators off the completed bars then place market orders. Per 2) I swing trade stock indexes and day trade commodities always out before EOD. Per 3) Futures Only and ETF for stock index futures swing trades. Per 4) Each one of my system can arbitrage every market, but the best results I've found must have some analysis for efficiency so I use the efficient market frontier to determine what's best. For commodities most investors are more comfortable with commodities being out before the end of the day. Per 5) I would say the strategies are World Class Complex Genetic Algorithms. Per 6) This can be measured through the Efficient Frontier so there is a mathematical definition I use to analyze when to use them. Per 7) I don't know, but the APD in backtestests is around 0.7 with profit factors greater than 1.6. Per 8) They are percentage stops and the indicators as with the swing index systems have a factor to play in closing trades. Time stops and a combination of the two also work well for day trading. Per 9) I've not seen strategies that would buy the open are necessarily that bad, but EOD is more determined by the day trading strategies I use particularly in Corn and Gas markets where EOD is found through genetic optimization and so do not get out EOD without also being before the times where they might close in their Open Outcry Pit Session Times. Per 10) I've never understood the need to be both long and short in a strategy but primarily for swing trading indexes I find one system overrides what might be conflicting directional hedges and for daytrading I'm purely directional with time based exits and percentage stops but also could be out based on the same indicator.
Hi, What makes you doubt your system? Are the measures that are being generated in line with what your back tests/paper trades suggested or are they off of that? Just wondering as another system guy struggling with this stuff.
I have been testing my system live and at the same time running parallel testing in Sim to see if entries, exits and slippage all line up. 99% of the time things are in line with back-testing. What makes me doubt my methods are back-testing and my testing process. I still think its too easy to curve fit and sometimes i wonder if markets are random walk. I still think some of my systems are too simple are chart based which can be random noise in intra-day trading. So far i have now been running 11 strategies: - I have decreased position size on two strategies so they have a little less weighting overall. - Removed one strategy for under-performance and being to correlated to another strategy. Results so far after about 3 months is: -6 strategies are positive. -1 strategy is destroying the market right now but has only had 14 trades with 12 pretty decent sized winners. PF way over 6 right now. - 3 of winning strats are pretty much flat. - 4 losing strategies all less than $1000 dollars - 1 of those strats discontinued - 1 strat i have complete faith in turning it around as its a strategy i traded as discretionary trader for years and i am positive it will work. - 1 losing strategy which is my biggest loser is my DISCRETIONARY TRADING Basically i suck at trading discretionary trading and averaged down into a couple of trades. My conclusion so far would be to say so far without discretionary trading i am at about PF 1.17 which is not good. I always wanted to have many strategies to run live and basically pull the losers if they exceed or perform worse than they should based on back-testing. I will keep adding more strategies as time goes on. Because money management is at the center of automated trading then i think it will be pretty hard to be a losing automated trader if you are running a series of strategies. Odds of them all going tits up at same time when they are uncorrelated should be very rare (especially as many time i am hedged long and short). So far every month has been winning month. 9 winning weeks 5 losing weeks. Bottom line, i will run these live for a year and then see if data is inline with backtesting. I think odds of being flat on the year maybe a reality for me. Odds of losing more than 10% of my account should be rare based on how strats are trading right now.
Here are the performance figures for my live trading of CL AlwaysIn since I switched to v46 on October 15: - 137 trades - win% : 59% - avg W/avg L: 0.92 - avg net/trade : $86 - P/F : 1.38 Now, the global view since I started trading that system, all versions combined, is the following: - 560 trades - win% : 53% - avg W/avg L: 0.93 - avg net/trade : $19 - P/F : 1.08 I am now convinced the versions prior to v45/46 had a number of over-fit patterns, still they had done extremely well over the 1st half of 2013 (beg. of July, I had 200 trades, 57% win, avg net/trade $106, P/F 1.47, net account return +45%), before doing equally bad over August..October ... those versions are making some money since then, not as much as v46 and certainly far from their 1H13 performance. If anything, one takeaway is that 200 live trades is no more guarantee than 5000+ backtest trades.