I'm currently working on a graybox. At this stage, it is more of a conceptual project. Remember, I built a clone that allowed me to historically optimize my discretionary trading. The building of a graybox is a natural development based on my goal of combining the best of my discretion with technology. A graybox is a trading interface that is partially automated but allows for human input. There is a great deal of freedom and creativity that one can explore when designing a graybox. There are some fundamental differences in approach, as well.
We need to distinguish between a genuine trading edge, and what I like to think of as automaton trader. In other words, we can build our graybox to trade in such a way that if given the correct direction that it would trade profitably: think buying pullbacks in an uptrend. Again, I think of this sort of notion as an automaton or mechatron. The system would only be profitable if given the correct direction or "an edge" from us.
The second way a graybox can be created is that we have a genuine system that has an edge. In other words, it would be profitable if traded automatically and it provides signals which we deny or allow.
We're basically thinking about the different ways that we can approach this topic. The last topic really provides the foundation where we should focusing on automating. Basically, we want to automate and systemize the aspects of our methodology that we're not doing good at. And for me, I'm very strong at calling direction. I'm extremely strong at predicting the direction of the market. I'm pretty strong in getting good entries but certainly that could be improved upon. I'm fairly good at getting exits. I'm very poor at finding a good stop out level.
And this is tangent but if you have an edge then a stop loss will almost always cost performance because you will have some false positives. Think about: you have an edge and that edge is worth 7 winners out of 10. If you have an edge then the more risk you take up to a certain level then the greater return you'll get. There are a few exceptions. First, if you're trading a single instrument, single position, then holding onto a loser for too long would hurt performance because you're not able to take the new profitable trades that come up.
Back to the topic, I bring up the stop loss because it is the most simple form of an automaton or automation. We choose a set dollar amount for our stop and now we've automated that process. We've freed our mind to focus on what we do best.
Why would we trade using an automatonic graybox versus an edge based graybox? Well, the edge based graybox may be somewhat of a limitation for a discretionary trader, like myself. You would only be able to take trades when the edge signaled. You'd be more of a filtering mechanism where you deny trades. The graybox I envision building: I'm the driver. I'm pushing it to take trades or lean on a direction. Its the mechanism that gets me into a trade and out of a trade.
Again, I want to get back to this "why". And the "why" is that the computer may be able to get a slightly better entry or take a better exit. It might take the form of combining the automaton with some genuine edge for a mixed model. This is an experimental project, and I don't know where its going. It could take the form of a scalping engine that scalps in the direction that I give it. I envision a somewhat rich experience with the graybox. I envision the following inputs:
Direction: The direction the box should trade.
Key levels: Key levels where the box should take profits or do something
Confidence: Trading size.
Urgency: Does it use limit or market? How close are the limits?
I do know that I want to be "driving" the trading versus "filtering" the trades. In other words, the way I envision this is that the graybox isn't providing suggestions to me but I'm providing suggestions to it. Of course, a more complex model would be a 2-way street with me driving it and then getting feedback and approving/denying its recommendations. Yet, I'm looking at this as a 1-way street. The way I'm envisioning it is that the box is micromanaging the trade entry and exit and taking the big picture information from me.
On a related note, I'm going to be working on a custom HUD for my discretionary trading. Most charting applications are not really suited to displaying information in the optimal form for my trading. Of course, charts are useful. Even so, I do not even use real-time charting for most of my performance this year (or years prior for that matter). Indeed, sometimes our limitations provide the mechanism for novel inspirations.
Tags:
commentary
Posted by
Market Predictor
at
2:38 AM