how to automate a successful manual system
- This topic has 57 replies, 16 voices, and was last updated 3 years ago by
.
-
I have been day trading for about 10 years and in that time I have tried and dismissed countless systems, some of them ridiculously complicated. Since the beginning of this year I have settled on a system that is simplicity itself and, day after day, week after week, it literally never fails. I am posting it here with 2 purposes; one is to share it with anyone who might find it useful – but also to set a challenge to anyone who can successfully automate it.
I live in London and trade the FTSE 100 but there’s no reason this shouldn’t work on other indices if you prefer. I use a 2 minute chart with 2 indicators: HULL 13 and HULL 34 (Fibonacci magic). In a nutshell, when the 13 turns up I go long, and short when it turns down. Simple as that.
The HULL 34 is there as an approximation of what is happening at the 5 minute level and it plays a much lesser role in my decision making. For example, if the 13 reversed but the 34 maintained a strong trend I would close but not reverse my position, just wait for them both to be moving the same way again. But if the 34 were still falling but gradually flattening out, and the 13 turned up, then I would enter long even if the 2 MAs were not in agreement. I think this ambiguity in how I interpret the HULL 34 is what makes it so hard to code.
I know that I also watch for how many bars close above or below the HULL 13 in anticipation of a change. This is also hard to specify in code.
So as a manual system there is some interpretation going on, but very little. Even if I ignored the HULL 34, the 13 on its own would make money. You can easily check this by counting the number of points to be gained by following this method on any given day. The key for me was developing the discipline to always do what the chart was telling me to do. We are not here to predict the future, so forget about what what you think it’s about to do – forget about trendlines, forget about support and resistance, forget about job reports or war in the middle east, the ECB or all that macro-economic stuff that makes your head hurt. The chart already knows everything, so don’t agonize about it, don’t even think – just look at what the MA is truly actually doing right now – and do that.
This is perhaps harder than it sounds and it takes some practice, but what you have in front of you is a graphic distillation of the collective opinions of thousands of traders from all over the world. If you just go with their consensus in real time then, by definition, you can never go wrong. I don’t mean that you will never place a losing trade, but the losses will be small. At the end of the day you will be well ahead, I promise.
I hope that people find this useful. If you can master it, it’s a license to print money (yes, it truly is). If it could be successfully automated, that would be even better. I’ve made a number of attempts and it baffles me how well it works manually and how poorly this on its own works as an algo:
1234567//HULL 13Period= 13inner = 2*weightedaverage[round( Period/2)](typicalprice)-weightedaverage[Period](typicalprice)HULL = weightedaverage[round(sqrt(Period))](inner)c1 = HULL > HULL[1] and HULL[1]<HULL[2]c2 = HULL < HULL[1] and HULL[1]>HULL[2]This I is what I think I am doing, but maybe there’s more interpretation going on than I am aware of ?? visual clues I’m picking up on from experience that lead me to click a bar or 2 in advance of what the algo would do ?? I honestly don’t know, but hey, there’s a challenge for you.
Best of luck! (although frankly, if you just do the above, luck is irrelevant)
Total of 12 users thanked author for this post. Here are last 10 listed.
good evening if we only consider HULL 13 as an indicator with a customization of the buy and sell curve, this gives that. Adapting a strategy to this would be the objective?
1 user thanked author for this post.
An automated strategy can only ever follow a set of fixed rules – it cannot interpret anything.
An automated strategy can only know what an MA in a strong trend is by the programmer exactly defining that in code. The programmer can define varying levels of MA slope and test each one and then choose the one that has worked best in the past but as we all now that is just curve fitting. However if there is a broad range of slopes that are profitable then that is what is generally known as an edge and then running multiple strategies trading within that range of slopes will possibly provide a profit. Not the biggest potential profit but an average of all the profits possible in that range. Pick just one of those slopes and you are gambling that the future will be exactly the same as the past.
Because strategies cannot ‘interpret’, running multiple strategies that all have potential to work within an area that is similar to your own interpretation is as close as you can get to a strategy that thinks like you do…. until that is we have proper machine learning – and even then we have to provide the machine with a pretty damn good set of rules as their first education.
2 users thanked author for this post.
Yes, of course, a program cannot ‘interpret’, it can only follow rules. And if I were that serious about trying to code what I do in manual trading then i probably would have posted in the other forum. What is interesting to me is that I know I have the appearance of a system, and I think I have described it well enough that anyone else should be able to do it. I know that I am following rules and that a very finite number of factors inform my decision making – but if that were all I was doing then it could be coded quite easily. So is it fully possible to even articulate what I am doing? or is there some other factor? will there always be an element of fuzziness to it that remains beyond the scope of coding.
In the case of slope, for example, I don’t have to get out a protractor and measure an angle to intuit the likelihood of a strong trend making a sudden V-shaped reversal. Yes, anything can happen – we’ve all seen it – but sometimes you just know at a glance how unlikely it is. And I think there’s something in those near-instantaneous intuitions – that maybe you only get after 10 years of staring at charts – that will always give manual trading the edge. At least, as you say, until we have real machine learning.
1 user thanked author for this post.
Could you please give us the indicator
For manual trading the HULL MA is included in PRT moving averages, just select it from the dropdown of options (simple, exponential, weighted etc). In an algo you have to use the code I posted above:
1234//HULL MAPeriod= 13inner = 2*weightedaverage[round( Period/2)](typicalprice)-weightedaverage[Period](typicalprice)HULL = weightedaverage[round(sqrt(Period))](inner)1 user thanked author for this post.
yes, that should work except that I have always used (typicalprice) instead of (close)
what error do you get?
1 user thanked author for this post.
With this indicator when I try to insert it it gives me the following error:
Just follow the instructions of the error: write “RETURN HULL” at the end of the code.
1 user thanked author for this post.
Safe is to wait til the candle closes and confirms the change of direction, but I have been known to cheat. If I’m getting in before the MA changes colour then I look for a break above the previous candle (sometimes I win, sometimes i get burned)
I don’t use stops or targets, I close when the 13 changes colour and usually change direction as well. Recently I’ve been experimenting with doubling my position size on the ticket and using net off to simultaneously close and reverse. Obviously some profit is lost by the time the MA turns, but there’s the advantage of always being on market. Requires total focus and conviction!
2 users thanked author for this post.
Find exclusive trading pro-tools on 
