Hey guys,
I would love to know if you guys do some kind of event study when researching for possible edges - and if so, how do you do / implement that?
I am not talking about classical event studies, like studying "abnormal returns after earnings" or other discrete events. Better phrased: I am not talking about typical events like earnings or M&As at all, but about events in a more "technical" sense. I.e. I might have a idea in mind like (simply put): price move of at least X% over n-m days, followed by a tightening consolidation and a breakout with an explosive increase in volume. <- That's my event, I want to investigate what happend after that "event"occurred.
*(Let's please not discuss the exmaple itself but let's keep it as just what it is - an example to illustrate my case).*
I can define technical rules to identify that situation. But I never really liked to simply backtest it but wanted to gain a better understanding of that situation. So I love to see how the monitored instruments performed after such an "event" occurred.
* How did the instruments performed subsequently after that event?
* I want to see several (different types of) plots of the subsequent returns to gain better understanding of the distrbution of subsequent performance
* I want to control for other factors (e.g. like benchmarking against different indices and/or sector averages in the same time frame etc.)
Yes I know I also have to care about quite a range of different biases when performing such a test. The same is true for backtesting. => But I never liked to simlpy backtest an idea and receive a bunch of entries (and exits), price charts showing those entries/exits, aggregated performance numbers.
In that stage of researching it just feels way too early to perform a backtest that includes way more parameters that you can tune (and misuse). I do not want to incporporate things like exit logic, stop loss orders, money management rules in this stage, but only research to get a better understanding of that situation at all (to see if there might be a true edge in that idea). I feel like that kind of event study I usually perform gets me deeper insights - and I feel more confident in really trusting my results compared to "trusting" the results of a backtest..
So I would like to know:
* Do you guys research your trading ideas in any way comparable to my type of "event study" research?
* If not - why not? Why do you prefer backtests (or other alternatives)?
* If yes - how do perform that research? I usually code that stuff myself, using 1d historical OHLCV data, because I never found any tool that matches my requirements.
Would be happy to hear your ideas :)