I am reading more & more posts / articles that mention "event-driven" strategies. I have found a "generally accepted" definition: "Event Driven Trading is any strategy that seeks to exploit pricing inefficiencies, occurring when companies are involved in corporate events such as mergers, takeovers, restructures (including share buy-backs, spin-offs and capital returns), de-mergers and lock-up expiries. Traders, who follow event driven strategies, attempt to predict the outcome of a particular corporate event on the security price as well as the optimal time to commit capital. For example the recent failed private equity bid for Qantas would have provided opportunities for events driven traders." (http://www.thebull.com.au/experts/a/249-does-trading-based-on-news-and-events-work.html) BUT ... I have also read a number of posts / articles that mention "event-based" strategies as opposed to "time-based" strategies, and I suspect in those cases people do refer to something totally different (for example, a CEP - Complex-Event Processing function in a HFT system). I am curious & interested in understanding better that angle ... at a high level, what can be those events ? Here is a tentative start to a list to get the ball rolling: - purchased order flow information (not for the retail traders!) - bid/ask pattern - price pattern - would an indicator-based signal (say, the cross of to MAs) qualify as an event ? - ... I am looking forward to your inputs. Cheers D.
The second part of your discussion began at about the time Dow Theory was constructed. By reviewing the last 400 years of trading innovation , you will find about 100 developments that contributed to your quest. Before information grouping, a "tape reading strategy" was fairly well refined on the basis of your quest. The PC and the printer made it possible to formalize record keeping of information groups. There is a misunderstanding of the cart and horse regarding displays. The information group comes before the display. From the 1840's onward, algebras for differnent bases made it possible to handle the lack of continuity of information groups. So paradigm theory made it possible to completely describe markets and then, in greater detail, use logic to flesh out market variable cycles. Cycles have two halves now defined as trends. Trends, in turn, have three parts. The groups of parts are defined in two ways and they are 100% correlated. Thus everything that is needed to have a complete system is scientifically and mathematically defined. Depending upon how things are viewed the market could be opaque. ET has the most outstanding examples of this phenomena. Good luck.
Minor follow up. A very large collection of indicators has emerged. Each has a set of associated "signals" that serve to increase the degrees of freedom. Corrolation among the indicators is various. A misconception among non-mathematicians makes serious mistaken claims that have become beliefs instead of myths. For processing complete market descriptive systems about 70 degrees of freedom are sufficient. There are 56 elements in the event spectrum. Each is arranged as part of sets and subsets to uniquely define the events of the market operating system. Time series mathematical development is used all over the place. In each aplication the results do correlate and the reuslts do not describe the full operation or opportunity to take the market's full offer. A vast body of study and regulation has failed to develop your quest so far. No one, so far, has been able to understand the consequences of using such methods. In five years everything will shift from the status quo to the topic of your quest.
Jack, I understand & respect (more & more I should say) what you say - FYI, I now have a fully automated system that trades Crude-Oil by staying always in the market, just reversing position when deemed appropriate - a 24/7 form of SCT, although using a different approach to read that particular market. Backtesting performance since 2007 is more than outstanding (can't backtest prior to that, as CL went all electronic in the fall of 2006, and that system works off low-level volume chart). This has me convinced now of the power of this type of trading (always-in), and if there was a known path to your SCT I would happily take it - but that information appears so scattered that I gave up already a couple times on that. But yes, I understand SCT is event-based, so I should add price+volume patterns to the list. Cheers D.
When a fund says they're event-driven, they're talking about corporate events.... Earnings, news, spinoffs, restructuring, etc. I think the other case you're referring to is more microstructure based, like measuring time via trades or volume rather than clock time. I think in that context they're calling times of the day that humans trade by convention (open/close/option expiration) to be events that machines can game. I'll defer to the algo guys on here on that.
A great example of event trading happened when Deckers outdoor corp. (DECK) jumped from 40 bucks to 50 bucks after it reported its Q4 earnings per share. The EPS was 2.77 and the consensus was 2.61 so it was better than expected and that brought a lot of buyers in in a very short amount of time. The really exciting part of this move was the short ratio being really high-6.53- combined with the price action. Sellers had moved price from 45 down to 40 so most stops were outside 45. 40 was also very well defined S/R so lots of latent buyers waiting to scale in all the way up and lots of latent sellers waiting to scale in all the way down. When the earnings came out better than expected price gapped up as longs rushed into the market and shorts had to scramble to get out. Then price blew threw all the stops that were in place from the shorts that had pushed price down to 40 from 45 causing the market to move higher all the way up to the very nice big round number of 50. That was one of the better examples of a predictable event (earnings report ) that had a slightly better than expected outcome ( 16 cents per share above consensus ) and the relationship between price action and market micro structure. Hope that helps.
You make a strong point about getting to and then switching to always in trading. The support, confidence and comfort required to trust yourself is a significant bump to get over. Added experience, then becomes iterative refinement. As I begin my journal I will resume posting complete charts and parallel logs so you can see the pieces and then see the relationship of the pieces. I will throw in the chapters of my book so the context of the ideas and math is made a little clearer. When looking at the illustrations, please notice that the kinds of short comings of the platform are just worked around mentally. I can't debug the platform since it eats up too much coding and exceeds the processing capacity of the platform platform slows then can freeze and it is difficult to repair since the platform is no longer working. Do you noitce what the contract capacity limit is on the CL? My lower limit on contracts for ES will be 50 and I'll go to 5 times the ES capacity for given times of the day. You may see me trading the ES on a slower fractal as well during the lower capacity middays.
This site has a Python 'project' for doing event studies: http://wiki.quantsoftware.org/index.php?title=QuantSoftware_ToolKit I took a coursera course that taught it: https://www.coursera.org/course/compinvesting1 I really like the project, it can do a lot, it can search for an event in many securities in one run, and simulate a trade portfolio while at it. The downside is it is all Python source code, and you are pretty much on your own to refactor it. If you like coding and know (or want to learn) Python, I think it would be useful. The course presentation unfortunately needed more work, it was a rough 'first-time attempt' (lectures posted late, did not finish syllabus, lots of students in the course complained). I pushed through with the coding because I wanted to work with and learn some Python. One 'event' in the course was: 'the stock drops below $5'. I think it was a simple example to start with, and I could see how it would be possible to use TA signals as 'events'. It can also do events if you have a list of dates; I ran one that found stocks (on average) went up from 20 days before div payment to 20 days after (note, data was from 2008 - 2012). I think I did that by getting the div payment dates from yahoo, and told the code to use those as events.
I am looking forward to learning from your journal! When are you planning on starting it? WRT CL contract capacity, this is nowhere close from being a concern for me at this time ... however, given CL trades about 1/10 the volume of ES, I would suspect that 1/10 ratio applicable to contract capacity. But then, CL average daily range varies from 140-ticks to over 300-ticks on the last 10 years, which is about 2.5 times the ES average daily range (in ticks).