Good points from Mickey, that is certainly a voice of experience speaking. The bigger the better, in general. Unless you are specifically designing a system for some index. With bigger universe you will start seeing bigger outliers and more vicious DD, so your position sizes will likely need to be smaller to mitigate. Which is actually a good thing. It is better to get used to large DD in your backtest and suspect something is wrong if they don't ever happen. Over the time you will get a sense on what DD should look like for what position sizes / hold periods / exposure. With Nasdaq 100, tickets sample is quite small, so you will see more systems looking like they have small DD and it is easier to overfit. If something is built just based on NS100, I suspect you real life expected DD will probably be 3-4 times of what backtest showed. It all depends, but the more filtered, the less trade count, the more deviation from backtest you need to expect in the future. RealTest supports breaking test results into baskets (bins), which is amazing for slicing/dicing your universe and seeing if there are clusters / tendencies you can take advantage of or filter out. It takes some time to understand how that works, but it is totally worth it. Here is an example. I grouped 3k trades from one system into equal range bins using some measurement of stocks' liquidity. Y-axis is that "bin"'s average trade's PL in %, X-axis - different ranges of liquidity. First bin is for liquidity <0.48, average trade PL is ~0.6%, second 0.48-0.73 with avg. PL ~0.25% etc. You can clearly see that edge deteriorates with a greater stock's liquidity. Which highlights, why, in general, a successful individual retail trading is possible. Val
Hi Val, what slippage estimates do you test with or have noticed in live trading while using LMT orders?
In my opinion, in systematic trading, right answer to this + right commissions structure, is literally worth a million $ as it makes or breaks a strategy. I can say this much - if you can make your strategy perform with 0.1% slippage on both sides of a trade, there will likely be a way to execute it close to your model with <500k account. The larger is your account the harder it will get. Finding a good balance between slippage and stock's liquidity working for your account size is the key that will open many right doors. Val
A pre-ordered "Unknown Market Wizards" finally came in this morning on my Audible. Barely had time to listen to it yet but went over Marsten Parker's chapter and really enjoyed it. As Jack mentioned, Marsten have been the only purely systematic trader with such a long track record he interviewed, which makes him quite unique. I known Marsten for some time now and feel incredibly lucky we've met. Very grateful for his guidance. Glad to see his story finally getting attention it deserves. Val
KXIN today.. insane outlier in a recently added penny stocks strategy. +80% in <24h. This might be a record for me on % return in a single stock for such a short period of time. I feel like a lottery ticket winner on this one. PS. Sometimes I look at those little companies and crazy price action and think to myself - what a garbage little company, then I realize - well, that must be how other people would be thinking about my business, especially if me & my partners would decide to IPO. Funny thing is, I started seeing companies in some of my trades with a market cap less than my own business.. Val
Thanks, as always, for your kind words Val. As you will see (or, in your case, hear), one of the other guys in the book (the one with the most extreme stats) made a lot of his money trading penny stocks.
As I was thinking about ideas for new posts, I realized that I don't need to go far and as the name of this journal suggest - might as well write about Automation. Also, there seems to be more questions about it lately on ET in general. Why don't I post about Automation here much? I believe it can become a rabbit hole very fast. Idea of solving trading thru Automation is very appealing and it gives a false sense of hope. Once we start automating, it's hard to stop, as this is one thing which can actually provide a solid sense of progress and achievement. It can be measured and it actually can be done with 100% certainty. In my personal opinion, the biggest return on time spent on Automation is in Automating very routine and error prone parts of your trading once you sure they work without Automation. Example - automatically sending orders to your broker instead of typing them, or closing position once it hits exit criteria. Key here is - you need to be already certain your method has an edge and works. Automating things that don't make money in a first place will not help Automating edge validation / backtesting. Once you have an idea or maybe already traded it and got trades sample that looks promising, there is a point where you know what trades you are looking for but it takes just too long to go thru past data to dig out more samples and record some basic trades starts. Then sure, automation of such test is the answer Automating something that you have high confidence in working but wouldn't be able to do without Automation. Who knows, but maybe you want to scan premarket for a very high relative volume 5 mins before the open and stock gapping up more than 2%, then go long in up to 20 positions betting 2% capital on each. This just example of something that will be incredibly difficult to do manually and likely impossible to execute well consistently doing by hands Automating the rest. Basically once trading became pretty boring and it's all about executing your system now - why not. Might as well automate the whole thing including data updates, daily prep, executions, positions management, reporting etc Re #2 - good quality backtest is difficult to do. So if you're just starting out your time is better spent on validating your idea manually first. Going thru charts one by one and looking for your setup and recording stats. Most of popular ideas just don't work so there is no point in investing into becoming a backtesting guru to figure that out. Often, after doing it manually for 20 trades you will see that idea is not worth it. But, unlike with automated backtest, you will pick up little observations on a way and that can help to refine idea and eventually to come up with something that might work. In the simplest form - maybe you will notice that doing exact opposite to what everyone else is saying you need to do works better. It's not that easy, but that phase of manually looking at charts has its learning advantages. Don't come to Automation as a rescue for unprofitable trading. Sometimes it might seem that only if emotions would be left out your executions you're golden. Unfortunately, it is often no the case. And on that note - taking more courses is not a solution either. Most people probably go thru 2-3 and it makes sense, as we expect that there is sort of formal education in trading. Takes 2-3 to realize that there isn't. Taking more is likely just postponing that inevitable realization and clinging to the uplifting feeling of hope each new one of them gives. Added that last paragraph after reading a story of a poor soul who gave up trading after 8 years of switching from method to method and taking more and more courses. That is not to say there is no hope. Hope is in stats. The only question to aks yourself really - if I do that exact trade 1000 times what the outcome will be? And then gradually remove anything that stands on your way to answer that question. Val
Good post. One large problem with attempting automation is wasted time. I'll explain.... 1 Creating an automation trading system is highly complex. 2 While you attempt to develop this you wont be trading, ie, you'll be distracted with automation creation. 3 If you aren't trading, you are losing opportunities. 4 99% of the time you will be changing your mind and going down another rabit hole. 5 Automation then becomes another job, What do you want to do, trade or code? 6 Going back to point #4, nearly in all instances, trading is a journey of discovery, it morphs, what you loved about your system 12 months ago now no longer applies, you have moved onto 'better ideas'. Ya ok, redesign your automation system and back onto the loop of wasting time. 7 imo, automation requires a team approach of several people to implement, there are way too many balls in the air for a one man band.
Exactly. Automation should be viewed as a tool, not as the objective. As a tool it can either make your life easier, or make certain things possible which aren't possible manually (e.g. the screener/filter you describe).