Over the past 10 years we’ve seen the rise and rise of trading bots and Quantitative Funds and we’ve seen the fall and fall of traditional Asset Managers and Hedge Funds.
In building your crypto trading bot, what did you find was the most difficult human behavior to encapsulate with an algorithm?
such a great question and I’m so glad you asked!
I’d probably say we did everything we could to not replicate human behaviour
If you’re actively trading, human behaviour gets in trouble. Human’s have pack mentality which means you chase after a stock as it goes higher and end up getting into terrible trades. This is also one of the reasons day traders typically never make money. The algorithm wins because it’s not replicating human behaviour and just doing what’s most sensible given the data it has.
Good luck with the move too David!
Excellent insights. Thanks for the read! As a company that is preparing to launch an algo fund, it’s always nice to have some new perspectives.
As of now, profitable trading bots are like God to me – I’ve heard of them and there’s a great probability that they are everywhere, but I’ve never seen or experienced one myself.
But, as you said, I too think they’ll become more common in the near future due to the rapid advancement in A.I. and ML.
Most trading bots that show up on Google search are so not reliable. Anyone who goes with one must have an iron heart to be putting their money straight into another person’s pocket.
if you run a monte carlo simulation on a coin toss (assume heads=profit, tails=loss) you end up with a bunch of paths producing a profit and a bunch not producing a profit. Now, it’s no good looking at the model after it’s run and then picking the one path that has consistently hit heads and saying ok this is a “profitable model” and therefore we’re awesome. This is basically what everyone touting profitable bots does, they either overfit or take a favourable subsample of data to show that theirs is “profitable”. If you take any bot that’s in the public domain today, I guarantee that on a long enough timeline the survival rate goes to zero. The reason for this is simply that the bot developers aren’t incentivised to produce a profit for you. This is the exact reason we’re building Credium - when we launch we’ll have a page on the site where you can see our daily performance so we’re held accountable to what we produce for clients.
as the article points out the most successful trading bots are proprietary;
there is no reason for a developer/creator/owner to open a successful trading bot/program to the public…
yes totally agree Paul - if it was successful opening it up will mean it’ll stop working (alpha decay). The only way to avoid this is to raise capital and scale it yourself
Janny, we are agreeing on opening of the trading will mean diminished returns or stop of working.
How does this fit with your involvement with Credium - which is looking for customers?
good question - so we’re not opening up any of the trades or trade history but we’re opening up a lot of aggregated metrics. It would be pretty difficult to back out the underlying algorithms because we’re running 4 right now and they’re uncorrelated. It’s more of a fund that’s open to retail as well. We don’t need lock up periods because out portfolio turnover is fairly high so it’s easy for us to rebalance whenever we have deposit/redemptions. Appreciate you asking!
in my personal opinion if a program/bot generates revenue it is better for the creator/developer to keep the gains for themselves instead of running a business, pleasing investors and making gains for customers…
Potentially, let’s look at this together.
Scenario 1: ok so let’s say you spent a year developing an algorithm, and that algorithm makes 50% annual returns - this is pretty decent for one guy to achieve compared to what world-leading HF’s produce. Let’s also say you were lucky enough to be in a decent paying field before so you have $100k yourself you can run. After year 1 this is what you have: $50k returns (which you have to pay tax on). Probably tough to live on that given you may be accustomed to a higher cost of living given the previous field you were in. You also need to take into account that Alpha decay hit’s every strategy. The half-life of a strategy is about 1 year, which means if you haven’t come up with a new one after your first year you now have a strategy which only makes 25% annual returns.
Scenario 2: You take the strategy you built and acquire investors. Let’s say you manage to scramble together $10m on a 2 and 20 structure and you hit the same 50% in year 1. You now have $1.2m returns that you can use to go out and hire more quants and you’ve built an impressive track record (you just made $3.8m for your investors) on which you can probably go out and raise $50m in year 2. With the quants you’ve just hired 1 or 2 may come with their own strategies that they’ve already built. So you now have a $50m fund, 5 quants, 2 new strategies and all you had to do was talk to a few investors every now and then.
It’s funny because I’ve heard the “why don’t you keep it yourself” argument so many times. Maybe I’m doing the maths wrong but it never makes sense. Further, if you truly love what you do you’d take the opportunity of scaling and working with a team of ultra-bright people any day of the week.