Welcome to NexusFi: the best trading community on the planet, with over 150,000 members Sign Up Now for Free
Genuine reviews from real traders, not fake reviews from stealth vendors
Quality education from leading professional traders
We are a friendly, helpful, and positive community
We do not tolerate rude behavior, trolling, or vendors advertising in posts
We are here to help, just let us know what you need
You'll need to register in order to view the content of the threads and start contributing to our community. It's free for basic access, or support us by becoming an Elite Member -- see if you qualify for a discount below.
-- Big Mike, Site Administrator
(If you already have an account, login at the top of the page)
The Strategy Development Process for Rules Based Trading Systems
I'd like to start a thread, where we'll discuss and cover the Strategy Development Process. I've chosen to post in the Psych/MM section because, many of the concepts and even the process itself is heavily dependent upon a trader's knowledge and discipline and understanding the process will better help you to have a "plan" and execute that plan.
Hopefully we'll get some great discussion and wisdom from some of futures.io (formerly BMT)'s experienced and respected members.
This thread should be lengthy, so here's an initial agenda (which I'll adjust with input).
1) The decision making process.
2) Trading Objectives (where we're trying to go and what we're trying to accomplish).
3) A real life strategy development practical exercise (where we walk through creation of a viable strategy from cradle to grave)
4) Process review and improvement (review our process and methodology and develop feedback to make it better).
"A dumb man never learns. A smart man learns from his own failure and success. But a wise man learns from the failure and success of others."
This thread sounds to my (admittedly tender young) ears to have potential up the wazoo; practical and actionable morsels and meals, strategic argument, tactical efficiencies, market psychology and philosophy..... but reined in by the need to contribute to a plan a system a trade... logic...and at a year old there's..... nothing.
Holy cow... just realized something.... I've been a member now 350-some days, almost a year, as old as the thread.....
.... and this is my first post.
crap, the irony of it all hit me and I forgot what I was going to say (wow, nice 1st post)
Aha, yes, I'm engaged in strategy/system development at the moment. Some new tools have arrived on the market that allow complexity, type, control via WYSIWYG far, far superior to what's been avail to now. I suppose I'll keep you posted, hopefully Monday evening, July 1, with my first thought and experiences.
BTW I'm looking for a hopefully ma but windows will do, flow charting software....
This didn't seem to take off but I'll add my novice thoughts for what they are worth. Don't trade a system if you haven't backtested it authentically and know the set of rules you have previously defined, and do not deviate from, actually work over time, otherwise you're stabbing in the dark and destined to fail.
There's lots of talk about trading psychology and how it's a big part of the battle, which it probably is if you don't have a coherent plan you can have confidence in enacting.
Im currently attempting to develop a robotic system (hopefully many) with no experience of code, limited potential coding ability, but a big desire to try. I find the process of having theories about what might work in terms of trading rules and taking them from concept, to backtest to actually trading an exciting prospect.
Hopefully this post will spur on some further comment.
Everyone will try to crack this a different way, but I can give you a perspective that might save you some time.
Instead of working off of the traditional model of:
Hypothesis: > Testing > Analyzing > Tweaking > Retesting > Tweaking > Giving Up > Trying another Hypothesis
Do it faster like this:
Pick out all the variables that could potentially have any type of correlation and measure and quantify how they move the needle up or down in relation to each out. Don't even worry about trading yet. Just start by getting all the discrete variables in a nice data set you can play with then start to see the relationships. I'll give you a few examples.
Start like this: Take sequential list of every price change along with key statistics that occurred during that price level. Then map out for each price level: Did the bids win or the asks win? You will typically see the full population split around 50/ 50
Then you overlay your level 2 data. Pick out some key ratios: if the first X levels of the bid LOM are Y > than the first X levels of the Ask book do you still get a 50 / 50 split of bids winning vs. ask winning. Go ahead and map this out into 25 to 50 different combinations of different ratios and just look at it.
Then take your transactions: When one side is taking more heat than the other side, quantified by X, Y, Z how does this move or not move your starting 50 / 50 split.
I could go on and on about all the things you could test. Then once you see some things that are moving the 50 / 50 starting split to 60 / 40, start overlaying more than 1 variable at a time.... You will start to see the needle move to 70 / 30. 80 / 20 etc.
Then you will need to figure out if adverse selection / participation bias will prevent you from making bets along with the winning population. Spoiler alert, you will! Tons of mathematical edges won't work in practice because of how the exchanges queue / matching works.
Then you need to figure out what bets you can pull off, and how fast you would need to be to make those bets. etc. etc.
This is the process that will help you crack it exponentially faster IMO.
Best of luck
Ian
In the analytical world there is no such thing as art, there is only the science you know and the science you don't know. Characterizing the science you don't know as "art" is a fools game.
The reason you should build your data set with all the key variables at once is because if you are taking on the endeavor of sourcing the raw data you should go ahead and grab everything. You will get economies of scale by doing a full pas at all the data in first step vs. trying to source every possible variable one by one over time and integrating an ever changing data structure bit by bit. If you design the structure right in the first place you will save a ton of time in the long run.
Regarding testing just one variable as a hypothesis: I am suggesting precisely the opposite of this. This is where people lose most of their time. Instead of working to prove or invalidate a single variable with a specific measuring stick, it is exponentially faster to build a relationship map of all the variables that could potentially influence price movement and measure key relationships from -100% to +100% and see how this impacts everything. Then you can just look at the data and tell that when X relative to Y is 50% or > then 60% of the time Z will occur. You will literally already have the answer before you even think of the question.
If you started out sourcing one variable, and thought of one test to try, and found one key ratio to measure, and then ran your tests you might be going down the wrong rabbit hole completely.... And there are potentially thousands of permutations to this, so why not just build a data relationship map in the first place and start with all the key variables and ratios that hold statistical significance.
Anecdotally I spend around 5 years with the traditional hypothesis testing method and only really kicked the tires on a handful of viable edges. Then I took about a year off of prediction trading based research completely and just worked on sourcing microstructure data and building a comprehensive dataset with every relevant level 1 and level 2 event sequenced so I could analyze it.... And in 1 year I was more productive than in the first 5 by a considerable margin.
But it is not a path for everyone because it is extremely technical and difficult to write an extraction program to source your own microstructure data set to study. But for anyone that wants to find some real edges, this is the best path IMO.
Having said all of this, I will mention that perhaps this is not the appropriate suggestion on this topic considering many people getting into strategy building are just trying to crack "Hello World / The MA Crossover", and I don't mean to discourage new traders by suggesting this level of complexity is the only path. For advanced traders that are still struggling to find edges, maybe it is time for this path, but for new traders that still have wide eyed optimism.... Try hypothesis testing at first perhaps. It's not bad, it's just like playing whack a mole and eventually you will want to break something, so down the road.... There is a better path.
Best luck!
Ian
In the analytical world there is no such thing as art, there is only the science you know and the science you don't know. Characterizing the science you don't know as "art" is a fools game.
thats exactly my last insight after years of statistical system data analysis. You are wasting time, if you test one variable after another, because there are always strong correlation between all of the variables. I can't find the edges if you look only one-dimensional in a multi-dimensional data set.
I'm currently on the track to change my analysis to use a correlation map and yes, it's very technical and it's not easy. One main point is, that you can test new system variables easily.
no, i meant the dependency between a set of variables and a positive outcome (fixed min. target at the first step). After that you can test if the single variables are influencing the outcome in a statistical significant manner. Then you can strip down the whole variables set and run the process again. The whole analysis sould be running without user interaction.
