As I'm managing one discretionary position right now (short JKS) - looked at the market before close. Interesting day.. Such a strong up day in SPY with a gap, and yet, nearly every short I had from Friday is at profit comparing with Friday EOD levels. That is 7 positions carried over the weekend. I noticed if leading stocks had a big rally and then there is an unusually big move for the overall market - that marks an end of the rally in those leaders and they take a break. No predictions here, just observations. My guess is - that is what is happening once again. Quite a few stocks had absolutely insane up moves over the past 2 weeks. Val
I'd be having a similarly profitable day except for one crapola day-trade short in NET. Makes me want to find a way to automate news retrieval. They had some kind of happy announcement at 9 AM that made the prior overbought condition irrelevant.
Val and I each have different apps for automation. Neither is publicly available at this time. Maybe some day (after I finish this RealTest release). Basically I use RealTest to test strategies and generate daily scan output, with symbol and trigger price for each strategy. The automation app reads that file, monitors price action, and sends orders when appropriate. It also has all my exit rules built in (hard-coded with a few simple parameters -- one of several reasons it's not yet a general-purpose app). I think Val described his solution in a bit more detail earlier in this thread.
In addition to what @mhparker mentioned: For EOD data updates and RT market scans I have VM that runs a batch every 30 mins and uploads everything my execution software needs to AWS S3 bucket. Here is how that batch configuration looks like to give you an idea. RT's output is CSV files. Val
@ValeryN Thank you that's helpful. I have been trading with stockbee since 2008 but I am a novice when it comes to automation 2 of my siblings are programmers and have done lots of projects for me in the past using aws lambda functions. How many months or years did it take to create the automation and strategies you have in place? Any other resources outside of the books recommended to develop a level of comfort with the systems currently running? PS I am in my 30s as well and run a tech company full time rdetech.
Glad it helps. I like what Pradeep from Stockbee does and been a member in 2014 or so but found myself a bit too distracted by all the chatter going on. So many styles mixed in. Though, on a few occasions I have considered coming back for the social element, and quality of people is certainly higher since it is behind a paywall. Maybe I will at some point Looking back - connecting with Marsten back then was the biggest value for me. It felt like he was the only trader who actually traded purely mechanically without cherry picking and knowing he has done this for 15 years gave me extra confidence that it is possible. To be honest, it is a hard question to answer. This is like the story with an experienced ship mechanic who has been invited to fix a problem no one else could fix. He went around the ship, looked around, listened to noises, then knocked with a hammer on something and got the engine running in under an hour. The price he charged was 15k. Owner got angry and said - it took you just one "knock" with a hammer to fix it! And the mechanic said - knowing what to knock on is what you are paying for. The point is, once you know what to do - it is not that difficult. But knowing what to do and having great confidence in it takes the proverbial 10,000 hours. Confidence is extremely important as in trading, where knocking on the wrong part not only doesn't solve your problem but also cost you $$$$. Didn't mean to get all philosophical on you, but this is really why answers will vary. My first backtests were bad. Data issues, coding issues, unrealistic cost estimates, over-optimization, too little trades sample. Going thru that stage took about 1-2 years. To get over the basics and how to backtest and do that for basically every strategy in a book I've read. By the end of that stage I was disappointed in pretty much all the books I've read and pretty much nothing there could show even remotely tradable systematic edge Was pretty confident of my ability to say if something does have potential or not Still not entirely confident if what I saw as having potential will actually be tradable Went thru expectations adjustments in regard to potential performance and where it will be coming from Had a couple of strategies that seemed to be most promising While that was happening I was trading discretionary, and pretty much 100% shorting stocks. Screen time required for that was crazy and I decided to actually launch my first automation based on purely systematic strategies that showed a promise after those 2 years. Initial version didn't take long, might have been overnight or over the weekend, as all I did was creating a Java program to read AmiBroker's CSV output and submit orders to IB + a bit of logic to monitor them during a day and close on a certain condition. Then took about 6 months to learn on what's feasible to fill and what not, incorporate those into backtest, come up with new versions of strategies. Those started to look promising. Couple of strategies were running at that point. I think it was in 2014. Then I had a life changing injury and eventually made a come back in tech after taking a bit of a break. Started building a business. Years after I found some time on weekends to re-launch that trading project. I started with testing how my old strategies did and to my amazement - they performed fairly well.. That gave me motivation and confidence to continue. Ended up rewriting my whole execution environment over couple of weekends, making slight adjustments to strategies and relaunched everything. Once it started making money I put more effort in it to automate everything including daily routine with data updates, signals generation etc. Probably another year to get to where it was just running for weeks at the time. If I would start everything from scratch today - Creating strategies takes the longest. But it wouldn't take much as I know where to look and my expectations are very realistic. Maybe a week to code something and test properly. If you get into something entirely new for you then probably another 1-2 years. Creating execution environment (I'd still do it from scratch) - ~1 months Automating all things around (reporting, data updates etc) - ~2 weeks A lot of backtesting. Ultimately you want to come up with some at least remotely decent strategies and see how they perform in the future. Maybe put just a bit of automation around the best ones which doesn't take too long. Then see if you're getting the kind of fills you expected, see life commissions / borrowing availability and costs if shorting etc. Maybe start with MR ones. The way I see them today - it is almost difficult to find the ones that don't work. But most will have a very tiny edge that is degrading over many decades and will likely continue to do so. But is still very tradable. On that note, in systematic trading - not looking for a huge unrealistic edge is an edge. Knowing what to look for when backtesting is huge. Even on just "results" level. Maybe I'll write my next post on that. Val
I guess there are no shortcuts to developing strategies. Since I’m not a mathematician I would need to piggyback on other resources for ideas and strategies. Is there a library or community out there with strategies to play with? Examples included in real test are a great start wondering if it’s possible to find a larger library of ideas to play with or is it just read books? Trying to avoid the endless search in the wrong direction.
There are several currently profitable strategies in the RT examples. The book Val mentioned near the start of this thread outlines a good approach too. You're not looking for the holy grail. Just a collection of good-enough strategies that are not all correlated. No advanced math is required. (I've studied neither calculus nor statistics.)