In our eternal search for the trading Holy Grail it is often tempting to try and find the “ultimate” signal (indicator) and apply it to as many instruments we can. This single solution approach for the most part fails miserably, think of a carpenter with only a hammer in his (or her; QF is an advocate for equal opportunity) toolbox. While making signals adaptive is definitely an improvement however, I think we sometimes miss the point.
Instead of harassing and optimizing a signal ad absurdum to improve the backtest, one would be better served by looking at the big picture. One signal in itself only contains so much information. While there are a lot of good indicators that perform very well by themselves available (the blogosphere is a really rich ecosystem in that regard); their power is only magnified when combined with other signals containing different information. The secret is in the signal aggregation. In other words, in how we form and use an ensemble of signals isolating different pieces of information to build a profitable strategy (note the use of the word strategy as opposed to system, careful wordsmithing aside, the difference is paramount). This is a topic I have been taking a close look at recently and I think the blogosphere is a perfect tribune to share some of my findings. As a starter, here are some points I will be touching on in the upcoming series of post.
1. What are the basic intuitions behind ensembles and why can they help in building trading strategies?
2. How do we isolate and quantify specific pieces of information and then observe their effect on the instruments we trade.
3. How to we evaluate current pertinence of the signals.
4. Finally, how do we aggregate all the useful information and build a strategy from the ground up.
The mechanics are going to be explained using a simplified example for readers to follow along but the intuition will be the same that the one behind the first QF strategy to be tracked real-time on the blog. I still don’t have a fancy name for it but it’ll get one for its official launch.
QF
Quantum Financier 
One way that people have done this is to assemble a “swarm” of several hundred individual trading systems. Each trading system reports its position (long, short, or flat), and then you tally these positions. (In effect, you give each swarm member a vote.) If the net position of all these systems is above the go-long threshold, why then you go long. If net position of all systems is below the go-short threshold, you go short. Same for going flat. Here’s a pair of research papers that describe the approach, and how it’s being traded, today, with actual customer money, in a CTA fund http://goo.gl/jNCtw http://goo.gl/T8r52 Naturally, their selection criteria for including individual trading systems in the swarm, incorporates a preference for low-correlation (or negative-correlation) swarm members. Interestingly, each trading system in the swarm trades the same portfolio. Thus diversity comes from timing, direction, and duration – but not from diverse tradeables. If the URLs above get destroyed by your blog software, have the moderator email me & I’ll send them thataway.
Thank for the interesting links and good comment. Swarm is akin to the ensemble concept in the sense that they aggregate different signals. While I won’t use a kitchen sink approach and throw everything in the mix and optimize, similar to particle swarm optimization, there will be similarities. To me its all about balancing quantity and quality while also using the market intuition traders bring to the table, machines just can’t reproduce yet.
Best,
QF
Hi,
Being a novice, ideally an effective QF strategy should incorporate sensitives to the third/fourth moment and probably include 3 signals (2 for confirmation – time varied).
Wrt to trading strategies, the basic intuitions are:
1) Is price going up or down and
2) What is the trend: Likelihood of this continuing?
From there, the leading questions on mean/ sensitives / volatility are formed or reached.
I personally would like to see a forward projection model strategy using something like a Brownian Motion(see below link) coupled with a Binomial series, appropriated scaled by drift.
link: http://www.youtube.com/watch?v=e79OtCamxD0&NR=1&feature=fvwp
Peace.
Hi Kez,
Good comment. I’ll address the comments in order. First 3 signals is very arbitrary, I can hardly find a combination of 3 indicators that contain the amount of information we can obtain by forming an ensemble. The point here is that we want to gather as much information as possible and find the grouping of signals that give a measurable edge in the market. These are also time sensitive, by fixing the indicator we completely ignore this dynamic aspect of the relationship.
Basic intuitions seem allrigth.
wrt the brownian motion and binomial series forecasting model, I am not aware of any literature that show the validity of the approach, as such it is low on my list but since you ask, I will be looking into it.
Best,
QF
Brilliant, although I’d also be interested in investigating should you be generous enough to assist. Thanks
Kez,
Fire me an email: quantumfinancier at gmail dot com, if you want, we’ll see what makes sense to do.
Best,
QF
Hi QF,
I hope the finals went well, and congratulations on picking up an offer for your quantly services.
Just writing to let you know that I for one (I’m sure there are plenty more) am very keen to hear more of your thoughts on ensembles and combining strategies. I’m currently investigating methods to combine mean, and variance forecasts from multiple strats into a single forecast for position sizing calcs (a la optimal ‘f’/Kelly).
I’ve been a fan of your blog for some time now, and will continue to visit and re-read old posts until you get some time to (hopefully) continue on this interesting path ;).
Regards,
Mark.
Thank Mark, I am drafting the post in between moving internationally and making sure I am ready to start work so it will come along and hopefully will be worth your wait.
Best,
QF