Today I have a great read for both professional and retail traders. By another “Kris”.
Kris Longmore at RobotWealth (many of you will recognize that name because of his terminally online collaborator @theRobotJames) published:
How I Built My Trading Business as a Finance Outsider: A Case Study (63 pages)
This mini-book is so good because of the raw honesty, reflection, tactics, story, and because it’s a rare play-by-play of what it looks like to go from not even knowing what you don’t know to zeroing on on the right ways to think about trading. It’s just outstanding. Takes less than an hour to read. Much less if your in brain is in airplane mode.
The full version is fun to read and worth it but since I snip excerpts you may as well have’em:
Along the way, my friend started talking about “mindset” and “keeping a trading diary.”
The idea was that the key to trading was having the right mindset and controlling your emotions so that your biases and perceptions didn’t interfere with the serious business of interpreting price patterns in real time.
I even read a book devoted to this very topic.
And this was when the alarm bells started going off so loudly that I couldn’t ignore them anymore.
The unsaid premise of the book was that there’s some sort of objective truth in these made-up technical analysis patterns, if only we could see through the noise and take it all in effectively.
That our single biggest battle is to get out of our own way in responding to them.
That the key to trading profits is to deal with our emotional baggage.
Literally the entire book was dedicated to emotional control in pursuit of responding to these signals effectively. It seemed to be saying, “If you can master yourself, you will see the matrix and make loads of money.”
The reality is that mindset does play a big role in trading. But it’s much more boring than you might think.
Primarily, you need the discipline to turn up every day and follow boring processes without having a boss to crack the whip. You also need to have the ability to not believe your own bullshit and to think deeply about the assumptions you are making.
I’ll have much more to say about these aspects of mindset later. But let’s move on.
While reading this book, I remember having a light bulb moment where I was like, “This is insanely, incredibly, and undeniably stupid.”
The answer was so obvious I barely even thought about it – backtesting!
In hindsight, I wish I’d thought about this a little more deeply, instead of buying into the assumptions implicit in those trading forums and books I’d buried myself in.
The lesson this person had learned was “the importance of backtesting my strategies.”
And so my brain connected backtesting with the work of actual hypothesis testing – even though this, too, turned out to be complete bollocks.
Maybe I was just letting hope and ambition interfere with clear thinking again. After all, backtesting is an entirely technical problem, requiring little to no nuanced thinking or decision-making in the face of uncertainty. You just write some code that simulates your trading rules and get your answer.
That doesn’t mean that it’s trivial. Backtesting takes work and technical skill.
But it doesn’t require wrestling with difficult problems – problems that don’t have a clear-cut answer and require weighing up evidence and dealing with uncertainty.
In that sense, it’s easy. And I was being seduced that I could “solve trading” by doing something easy.
Example of backtesting is not research:
I’ll describe the system I set up, even though today, this makes me absolutely cringe.
I had this idea that currency markets would see most of their trading volume in the business hours of the specific currency and thus be most likely to move at those times. For example, that the Australian dollar would move most during Australian business hours.
So I created a system that would calculate the range that, say, AUD/USD moved in outside of Australian business hours, and then traded it during Aussie business hours in line with the longer-term trend.
For example, say AUD/USD was in a longer-term down trend, I would sell the pair when it broke out of its overnight range to the downside.
I had no real economic basis for thinking this might be a real effect. And there are hidden assumptions littered all through that description that it didn’t even occur to me to dig up and critique.
I now know that I was looking at it all wrong – trading is not about something being “likely to move” – trading is about buying stuff cheap and selling it rich.
A better way to look at this trade would be to ask under what conditions is the AUD likely to be cheap? When is it likely to be expensive?
A subtle but important difference.
I won’t go into a ton of detail here, but this paper is a good example of looking at the trading problem in a sensible way.
Compare the description in the abstract, where the author describes a risk premium for holding certain currencies at certain times, with the description of my system above. One has a plausible basis for when currencies might be expensive or cheap, the other… not so much.
Excerpts
As I continued to lose money, my anxiety grew.
I had no idea what was going on. The only feedback I had was my trading P&L, which had gone up quickly, but was now coming down just as fast.
Should I stop trading?
This sort of anxiety is a real problem when your only feedback is your trading returns.
And returns are extremely noisy. Even if you have a real edge, you can easily be underwater for months or longer.
Imagine having a solid edge that would make money in the long run, but turning it off because your short-term P&L is lousy!
On the other hand, if your edge has some basis in reality, then you’re in a much better position to decide whether you think it’s worth continuing to trade or not. You’re not beholden to the fluctuations of noisy, mostly random market moves.
So how do you find edges that are based in reality?
- Think about and question the assumptions you’re making when you talk about a market effect.
- A good question is What needs to be true for this to be a good idea? For example, for breakout to work, the market would need to trend. Markets trend when past movements predict future movements. This gives you something you can actually test.
- When your trading P&L is your only feedback, things become way more difficult than they need to be. Have a good reason for an edge to exist.
You could use machine learning to model an edge. But you don’t find an edge because you’re doing machine learning.
Said differently, you never start with the technique;
you start with the edge
In my case, I’m sad to say that I started with the modelling technique and assumed I would just find an edge.
The second big mistake was becoming enamoured with a vanity project. I was very attracted to the idea of building something complicated and flexing some creative muscles.
But if you’re serious about trading, then you need a relentless focus on things you can do to make money today.
I allowed myself to be distracted by a shiny object with an entirely unknown payoff. I was trying to solve a problem I didn’t have or even understand yet.
This is another thing that you need to flip on its head. You always start with the simplest acceptable thing and only solve the problems you have right now.
For example, don’t set up complicated walk-forward frameworks. Just split your data into subsets and see how different the factor plots look.
When you’re doing it right, trading feels like problems emerging one after another as you learn more. So, you start with the simplest thing, come up against some problems, and then deal with them.
In that way, we move forward.
You should be taking your trading way too seriously to worry about problems you don’t yet know that you’ll have. You’ll have enough real and present problems – you definitely don’t need to go find possible future ones to deal with before you need to.
Convos with professional traders
I met some people who were actually working in the business of trading – some worked in proprietary trading firms, others worked more on the institutional side. I remember there being lots of people who worked in other areas of finance who wanted to become traders.
I found that many of the proprietary trading people were surprisingly open. They didn’t talk in exact terms about things that they were trading, but we had many great discussions. They seemed to like being helpful.
A lot of these people were using Python to do research and data analysis as part of their work, but I very much came from the world of R, which I’d used extensively in my career.
Some of these people were interested in learning about R and what it could do. I was very interested in anything these people would tell me.
So I found myself catching up with these people informally and showing them some of my stuff in R. This was way back when machine learning was just starting to catch on, and people were particularly interested in seeing what I was doing in that space.
I remember proudly showing someone my machine learning framework, and they were like, “That’s really cool, but what’s your edge?”
My blank look prompted further prodding.
“What effect are you trying to model?”
Not really knowing what to say, I mumbled something about feeding returns of correlated assets into the machine to predict SPY.
“Ah OK. So you’re doing lead-lag stuff. How do you know that stuff predicts SPY?”
“I don’t,” I replied, “but I was hoping the machine would figure it out.”
It was extremely gracious of him not to laugh in my face.
This was really the lightbulb moment for me. I’m not sure why I had this assumption baked in that the past price process was predictive – if only I used the right modelling tools. But at this point, thanks to some gentle prodding, I saw how ludicrous it was to just take that for granted.
It was at this point that I became really fixated on edges and what good ones looked like.
We also had some productive conversations around what good trading research really looks like.
One conversation went something like this:
“You’ve got to remember that researching edges isn’t like the sort of data analysis you do in your regular job. The data sets you’re looking at likely have a ton of signal, and they probably don’t change much over time.”
“Financial data isn’t like that at all. It’s highly non-stationary – everything is changing all the time – and the signal-to-noise ratio is super low. Any relationship you do find will be noisy as hell.”
“That’s both a blessing and a curse. It means that your edges are going to have a ton of variance. But it also means that simple tools tend to not only get the job done, but tend to save you from thinking you can be overly precise.”
Say I’m talking to a distinguished options trader, and he says to me “Volatility has been realising significantly under implied, it’s good to be short vega here.”
Can you spot the hidden assumptions?
The first one is obvious: that if volatility has been realising under that implied by the options market, then options might have been too expensive.
I’m OK with that assumption, so long as we don’t equate it with “selling options would have made money”, which will likely follow, on average, but with no guarantees.
The other assumption is a little insidious. We tend to make it a lot without even realising we’re doing it.
Our distinguished options trader is assuming that since options were expensive in the previous period, they’re likely to be expensive in the next period too.
This is another example of autocorrelation. You could also call it persistence.
And it’s super important not to just assume it – you must look for it!
So how would you look for it?
I’m so glad you asked.
A lot of beginners, including me, would try to do a backtest or a simulation of a trading strategy and see if it makes money.
That’s nearly always a bad idea.
Backtesting is complicated and subject to a lot of arbitrary decisions, luck and path dependency (path dependency means that the results depend not just on the final prices of the assets in the backtest, but on the sequence of prices over time).
Instead, you want to move quickly and use your data more efficiently.
First, we consider what data we need.
Ideally, we would have at-the-money implied volatility of 30 days-to-expiry SPX options, and SPX returns (for calculating realised volatility).
It would take quite a bit of work to get this data. But we want to move fast. We know that disproving things is easier than proving them, and that we can always loop back if we need to.
So rather than trying to acquire the perfect data up front, we use the closest thing that’s easy to get, so that we can get moving.
The closest thing to our ideal data is VIX index data from CBOE and realised volatility calculated from SPY returns (available from Yahoo Finance and other free sources).
Once we have that data, we calculate monthly realised volatility from SPY returns. We can then compare that with implied volatility for the same month.
In summary, at a very high level, good research for trading requires the following:
- Always question hidden assumptions.
- Make small, testable hypotheses in an attempt to quickly disprove your ideas. Ask things like What would I expect to see in the data if this were true? What would I expect to see if it weren’t?
- Test your ideas by looking in the data as directly and simply as you can.
- Seek understanding, not a result. Cultivate the mindset of a curious scientist, not an engineer with the end goal in mind.
- Favour simple data analysis tools and techniques.
- Precision is unattainable.
- Start with the simplest, easiest data that could disprove your idea. You can always loop back later if things look promising.
- Learn to deal with the anxiety that you’ll never figure things out perfectly. The evidence will often be less clear cut than you’d like.
[a deeply personal story with loads of wisdom compacted into this chapter]
We knew that it wouldn’t make sense to compete with firms like the one I’d worked at. We simply did not have the technology or the resources to do so.
So we spent some time putting together a plan for our trading, one that was organised around the unique constraints of independent traders:
- Low-frequency, at least at the start while we build out our tools
- Forgiving to trade – liquid assets, end-of-day trading
James has always been big on focussing on what you can do to make money today, given the tools and resources that you have at your disposal right now.
That means prioritising simple, obvious trades over complicated projects.
This has time and again proven to be some of the best trading advice I’ve ever received.
Things move fast and won’t look the same in the future as when you started building the things you thought you needed. And without some actual experience in the market, you don’t even know what you need right now.
And the whole time you’ve got your head down building stuff, you’re not interacting with the market. You’re not making trades. You’re not getting feedback.
If you’re building for the future, you’re not solving real problems; you’re solving future problems that you assume you’ll have.
And not only were you not learning market lessons, you weren’t making any money because you weren’t trading!
If you focus on making money today, you find yourself dealing with real problems, not problems you imagine you might have if only you can finish that backtesting framework, optimisation routine, or whatever it is you think you need.
Again,a simple edge that makes sense in terms of the market and its players.
But one thing that you should always ask yourself is Why would I be able to participate in this edge?
For example, if this trade exploits “predictable” rebalance flows, why aren’t the bigger, faster, better-resourced players eating up this entire edge? Why would there be any left for someone like me, trading on my laptop in my underpants on a crappy Australian internet connection?
And if you can’t answer this, or if your answer is Because I’m better/faster/smarter than everyone else, then it’s very likely that you don’t have a trade*.*
It makes sense, right?
If there’s “easy money” on the table, why on earth would the likes of Citadel and SIG leave any of it for you or me?
The answer has huge ramifications for how you think about your trading business.
The reason that you or I can participate in an edge is that there’s something that sucks about it.
For you or I to have a chance at an edge, there must be something about it that makes it unattractive to the big players. There must be a good reason for them to leave it alone or not absorb it fully. Otherwise, being the best players in a competitive game, they would leave nothing for us.
Some reasons that an edge might suck include:
- It’s very noisy – maybe it only plays out on average in the long-term, or is very slow to converge. The big players tend to like things that give them a smooth equity curve.
- It’s capital-constrained – maybe you can only get a small amount of capital into it before you move the market so much that the edge disappears. Such a trade might not be worth the time and attention of the bigger players.
- It’s margin-intensive – it chews up a lot of buying power relative to its expected returns. The bigger players tend to favour capital efficiency.
- It has a horrible skew profile – maybe it tends to blow up big time when it goes wrong, meaning you can only trade it at small size.
- It’s operationally awkward – maybe it requires opening a bank account in a foreign jurisdiction or something equally as tedious.
You get the picture, but I’ll say it again:
All of the things that you and I can trade have something that sucks about them.
The things that tend to work in the markets also usually look a lot like “doing something useful that the market values” – just like any business.
For example, a useful thing would be to provide liquidity to traders who need to trade right now.
Another useful thing would be to trade against positioning dislocations such that you push the market towards “fair value” – this is the source of the profits from trading the futures roll, for example.
Yet another would be to trade against behavioural effects that push the market away from fair value – for example, the post-earnings announcement drift effect.
So we have this mental framework of what a good edge looks like: doing useful things that suck.
Around 2021, we started getting interested in crypto and we were really keen to explore this brand new (for us) asset class from the ground up.
By this time, inside RW Pro, we had built out some neat collaborative research tools that we and the wider community use.
We call this collection of tools The Lab.
The Lab is all about facilitating a scalable, reproducible research effort.
It includes a hosted Jupyter notebook service connected to curated, automatically updated data sets. Research notebooks are hosted in GitHub.
What this means in practice is that I can connect to the runtime via my browser, load the latest version of a dataset, create a research notebook, and upload it to GitHub. Then, anyone in the RW Pro community can open that notebook in the same environment as it was created in simply by clicking a button in GitHub. They can run my code, take a copy, or modify it and save the changes for others to see.
Research is organised around themes that we call Research Pods.
This set up is useful for people at all levels of experience:
- Beginners can see what proper research looks like, run the code, copy it, and experiment with it. This shortens the learning curve dramatically.
- More experienced people can submit their own research to the community, receiving valuable feedback and growing the group’s resources.
Essentially, people get a leg up in learning how to do research for trading. Plus they get a bunch of stuff to trade, clean data, clear direction for research efforts, and feedback.
This means that for individuals, two things grow faster than they otherwise would:
- Your research and trading skills.
- The number of edges available to you to trade.
It’s a very cool concept and aims to replicate the environment you’d get in a trading firm, but for part-time solo traders.
Of course, we don’t share everything in The Lab. We’re mostly focused on low-frequency edges on liquid assets, for two main reasons:
- These are more forgiving to trade and don’t rely on execution skill (great for part-time soloists).
- We don’t cut each other’s grass too much.
James and I also trade some more capital-constrained edges, as well as some faster stuff that most people aren’t set up to do. This stuff doesn’t really belong in The Lab.
Anyway, when we got interested in crypto trading, we set up a Crypto Research Pod in The Lab, and the whole ecosystem really matured.
We used it to explore the crypto asset class from the ground up, taking many other newcomers along for the ride…
The beautiful thing is that, since the research is all connected to automatically updating data, anyone can go in there at any time and update the research with the latest data.
We also built a dashboard tool for the strategy that tracks the performance of all the factors that go into it, as well as the weights of the tickers in the tradeable universe.
We also serve this data via an API so that RW Pro people can use it in their trading applications.
As a result, we’ve had community members contribute all sorts of tools that they built on the API that help with trading: everything from Google Sheets that calculate positions and trades for manual click trading, to automated trading scripts for various exchanges and platforms, some of which I use in our own trading stack.
The crypto trading that we do today is truly the result of a team effort – entirely enabled by The Lab.
We’ve explored and traded loads of other edges with the RW Pro community as well. Some of these have come and gone over time – some we’ve retired because they seemed to die, others because we got too busy trading other stuff:
- A portfolio of FX strategies
- VX futures calendar trades
- Straddle over earnings trades
- Post-earnings announcement drift trades
- A meta-labelling strategy for directional trading of US equities
- We compiled a database of equity factors that can be combined into a long/short quant equity type strategy
This entire section on the glorious synergy of diversification is the basis of every successful professional trading org and seeing it applied to individual trading so effectively is uplifting even if it’s expected. It’s a simple idea taken seriously.
What I hope you take away from these examples is the realisation that systematic trading can work for the individual part-time trader, especially if you harness your greatest edge: the ability to trade multiple edges simultaneously.
Solo trading looks different to the world of professional trading in many respects – you won’t have the tools, resources, or the time to compete in the most lucrative areas.
But if you pick the right games by focusing on doing useful things that suck, and leverage your greatest edge by trading multiple strategies, and do the work well and diligently and consistently (and the reality is that it’s more work than most people imagine), then you can absolutely make good money as a solo trader.
I’ve been lucky enough to build a lifestyle that’s aligned with my goals and values. And while your goals and values will differ from mine, I can share some lessons on thinking about the sort of life you want to engineer for yourself, and what it’s really like in the trenches.
I think it’s important to ask yourself what you want from your trading.
- Do you want to pursue a career in trading?
- Do you want to trade for someone else?
- Do you want to trade your own account for a living?
- Do you want to create a part-time business that grows your capital or provides an income?
- How much of a life outside of trading do you want?
There are no right or wrong answers to these questions, but they will have an impact on the sorts of things you’ll trade, the time you’ll need to put into it, and your performance expectations.
Some people want to do nothing but trade. And I totally respect that.
Personally, my objectives have changed over time.
I’ve gone long periods where I’ve traded every single day as if it were my last day in the markets – spending long hours doing intense work. When you’re a solo trader, you’re working on your trading business and doing the actual trading in parallel. It’s not for the faint of heart.
I’ve also taken things easier from time to time. I’ve taken time out of active trading and just run some risk premia harvesting positions so that I could pursue other projects or go on holidays.
There’s no rule that says just because you trade, you have to do it full steam ahead all the time.
It’s a balancing act and a trade off.
Sometimes an edge will come along that you know won’t be around for long, and so you whack it as hard as you can for as long as you can.
But you also recognise that there will always be things to trade, and so you don’t have to go after absolutely everything right now.
If you have the motivation and energy to do so, then go for it. But don’t think that you have to. Do it on your own terms.
it’s not all roses. It always feels like there’s work to do, but the rhythm is conducive to the life I want to live. And I think that ultimately that’s what it’s all about – doing things you enjoy that provide meaning, and bringing in enough financial support to enable that lifestyle.
Actively trading for a living is a big job.
As a solo trader, you’ll need to wear a few different hats. You’ll do the trading, do the research, manage your own data, look after business administration, and a bunch of other things.
In that respect, it’s harder than professional trading where you have a team to do the things you’re not good at or don’t want to do.
You also need to be prepared for some boring work. Trading is less glamorous than most people think and more of a grind. A lot of it is running processes and not messing up.
Be realistic about your return expectations, and don’t expect to quit your day job in the near future. Having some sort of regular income is a great offset to the reality that trading income fluctuates and is highly uncertain.
There are other upsides to teaching and running a membership business as well.
When you explain something to someone, you’re forced to really understand it on a deep level. When someone asks you to explain something in a different way, you come at it from another angle, again deepening your own understanding.
In addition, through our membership community, I’ve been privileged to form relationships with people I’d have never otherwise met. Trading solo can be a lonely existence, so having access to a community of likeminded people has made a huge difference for me. You can’t put a price on that.
Some social media gurus sell confident answers to trading problems. They make it look like you can make a quick, easy buck trading.
That’s almost never true.
Someone long ago told me that trading is the hardest way to make easy money.
This rings truer than ever.
The reality of trading is that it is messy and complex. There are no rules. No right answers. Only trade-offs and varying degrees of uncertainty.
It takes a special kind of maniac to trade solo.
It takes the kind of person who finds this reality much more exciting than the simple, confident answers some would have you believe.
where I(S; O) represents the mutual information between the signal and execution outcomes, and H(S) is the entropy of the signal. This tells us how much actual information we’re getting about execution outcomes versus how noisy the signal is. Translation? We can spot real market changes worth adapting to and ignore the head fakes that would otherwise trip us up.
