There have been a lot of changes in the markets in past few years. Do these changes have any impact on the earlier working of the Trading Systems?
There have been a lot of changes in the markets in past few years. Do these changes have any impact on the earlier working of the Trading Systems?
Yes they have. The electronic markets have changed things. In some markets like the stock indexes. We can still get signals off of the pit and trade the electronic versions. In other market there is no longer a pit or a viable pit so we need to change our entries since we can't use the electronic opens which are not during times of high volume. One trick I used was to enter on the close price using a limit order. This will give us an entry which performs almost as well as the old "Market" enter at tomorrow open when we had a pit.
Murray, You've got about 40% of a great thread here. I would love to see it run to completion! Things did get sidetracked but I know how a customer facing life is; we can't just tell people to go to hell even when it's for the best. There was a thread in the same vein a few months back by TalonTrader about how to create a system which really stayed focused because he had the benefit of not having anything to sell, so he could tell people to go to hell in order for the thread to not get derailed. If it's worth anything, what I would like to learn to do is figure out how we know to pursue a strategy to disregard it. I'm working from 4.5 years of intraday forex data, and I can tweak and tune and optimize a strat that works just fine for the first 4 years, then fails spectacularly for the most recent 6 months. It would be awesome if you could show us how your product can point us towards the honey and help us steer clear of the bees.
What about if you use a degradation test to stop trading it when it starts to fail. or you could try to develop a system over a number of different markets with simple rules to try and make it more robust. or trade more than 1 system/market and hopefully they wont all degrade at once and you can stop trading the bad performing one and hopefully the others will make enough so that the losing strategy doesnt effect your account too much.
These are exactly the things I do in the trading systems I actually trade. I add equity curve filters in some of the systems and also combined into a tradeplan in TradersStudio multiple systems based on different methodologies. Also except for intermarket based system or systems which use other special market data or fundementals I always develop systems on a basket of markets.
Hi Murray. Thank you for posting and keeping the thread alive. I'll be interested in you expanding your thoughts on a couple of your conclusions. You arrive at different places then I have. First I always use the same signal parameters for buys and sells. I believe if you have uncovered some true edge in the market, then you do not need to start optimizing the components. Further if you have a true edge it really doesn't help to any significant degree anyway. Second while I do have some edges that are very transferable between markets, when I take the concept between markets it always benefits from adjusting to the particular market. I have watched this play out for years of watching after the systems were built and it bears out. For example if it is a momentum model that signals a trend, it may work across a wide varirty of markets, but each with different optimal thresholds to get the signal. Third using the exact same parameters on all markets just uneccesarily punishes performance. All markets have there own personalities. These personalities don't seem to change. The only thing I may put a qualifier on there is the switch from pit sessions to 24 hour electronic 24 hour sessions has in SOME cases efected some market characteristics. Last, I never use moving averages because they are just number crunching and real edges are IMO more to do with market personsaility.
I always start with using mirror logic on the long and short side , but I will tell you long and short is not the same. Short side price action is normally faster so as long as you have enough trades and a robust optimization surface using different parameters for the long and short side is acceptable. In terms of using the same parameters on all market, I am still strongly in that camp. I would say that in a adaptive system these things are not an issue, for example I would use the same percentages of the dominant cycle for all markets, the cycle software then can adjust the parameters. So for all markets I use 1.25 time to dominant cycle for example. So in cotton with a cycle of 20, that 24. In Crude with a cycle of 40, that 50. This address this issue and gives me the best of both.
If I understand your view, all markets behave similar so therefore they should all use the same system parameters. As well all markets behave different when going up then going down, so different parameters for long and short. Would it be fair to say in your experience that the SP, Euro, Cotton, Corn and Heating Oil all have more in common during rally phases then any one of them has with its own rally and own decline phases? I perhaps have not really grasped what you are saying? Optimization is by definition part of strategy design. We all have to pick what to do to be comfortable we are avoiding over optimization. At the same time we do not want to punish ourselves into using strategies that don't really work as well as they could.