How Amrullah traded 100 BTC To 120.47 BTC In 76 Days — Set Social Trader Spotlight
Welcome to the Set Social Trader Spotlight Series. In these posts, we’ll give you an insight into our launch traders that’ll be on the Set Social Trading platform once it goes live. Learn more about Social Trading here.
Today, we‘ve got an interview with Amrullah for all of you to enjoy.
Hi Amrullah, nice to have you. Can you give us an intro of yourself?
My background and career has always been in data science. I build and refine deep learning models for unstructured data. An example of the work I do is to build models for images, video or time-series data. Time-series data like a price chart of Bitcoin and Ethereum is at an interesting point in history. Deep learning methods are starting to outperform the more traditional statistical methods and that’s what I’m excited about. Having said that, I am a student of the Bitcoin and Ethereum ecosystem — new things constantly surprise me. Compared to my experience with Bitcoin, I’m fairly new to Ethereum and it’s ecosystem. I’ve only been studying Ethereum for 3 years now.
You told me about your ADL (Amrullah Deep Liquidity) strategy that you posted here. You mentioned it uses AI to calculate when to open a close your positions. Can you tell us more about that?
ADL is built to identify market structures left by other market making bots. I use deep learning methods to build the most accurate model I can. The model works by guessing the inventories of other market making bots and the goal is to detect any imbalance in these inventories. These bots are more likely to skew the order book to restore this balance. So ADL then takes a position in the market to exploit this. But ADL does it on a longer timeframe, not high-frequency trading style. ADL aggregates this data and then positions itself to take a view on the market. Users can find out how ADL has taken past positions here on TradingView.
That’s impressive. Can you talk more about how you were able to create the AI trading model that powers your ADL strategy?
Data, data and more data. Fortunately the crypto markets offer many types of data that’s available to those who are keen to dig deep. As data scientists, we spend 90% of our work cleaning up the data before we start running models to it. When I mean cleaning up the data, I usually mean we build ‘pipelines’. We do that to make the data ‘clean’ enough for our models to start analyzing. You can think of building the pipeline like preparing the oven you use to bake a cake. As long as you prepare the oven at the right settings, any cake that goes in will come out smelling delicious.
We rely on both the on-chain data and the off-chain data that a trading pair can make. For example, if we are looking at ETHUSD, we want to look at all the on-chain data of ETH. We want to see how ETH gets shuffled around from address to address. We also want to look at the stable coin system that represents USD. We may also want to look at how the market makers on Oasis DEX, as an example, is moving their coins or rebalancing.
And then there’s off-chain data such as the order book data we get from exchanges. It is non-trivial to identify the market participants at a certain order book. The goal of many of these market participants who buy and sell is to remain out-of-sight. It’s like an online game of poker. Everyone is trying to bluff each other out. It’s also like queueing for a concert and everyone wants to be the fastest in the queue. And we can know who’s the fastest in just by being attentive to the order book.
We take hundreds of these as examples of potential data points. But we don’t tell the model to look at this and that. The model decides what’s important and what’s not. These are examples of data we take in. Once the ADL is trained on it, ADL then take trades on the market on its own. ADL knows how to position itself in the market and also knows how to size up its position to get the most profit it can. For now, we benchmark this against a known standard called the Profit Factor. World-class algos typically have a Profit Factor of above 3. When we backtest ADL for ETHUSD and BTCUSD, the Profit Factor is consistently above 5.
Is this a strategy that you’d be offering on the Set Social Trader?
Your community will be able to find the ETHUSD ADL strategy on the Set Social Trader to kick this off. We hope your community of users will like it.
How often does your strategy make its trades?
12–15 trades a year for a given trading pair. Average winning trades last around 27 days and average losing trades last around 11 days. Your community can find these numbers on the ETHUSD ADL strategy page on TradingView.
What is something that people don’t know about you that you’d like to share?
That this entire interview is generated by a GPT-2 AI…I’m kidding. I’m an intersections type of person. I truly believe in the mission of liberal arts. We need to understand how humanities, arts and the sciences inform and enrich each other. I spent the last 5 years working with some of the best in the field for deep learning. They build state-of-the-art models for facial and emotion recognition. Lateral thinking is needed when you are trying to solve problems creatively. This also gives me a lot of exposure on newer techniques that emerge out of this field. Building up Amrullah Deep Liquidity is the result of many of the things that I learnt through the years. There are also some of the brightest minds taking a stab in this area of knowledge.
On legacy, one day I hope to join the ranks of J. Welles Wilder Jr. who invented the RSI (Relative Strength Indicator), Gerald Appel who invented the MACD (Moving Average Convergence Divergence) or John Bollinger who invented the Bollinger Bands.
Anything else you’d like to tell us or your potential followers?
I think most of us are in crypto for the long haul. But the truth is a big segment of the crypto community still measure up their USD returns for crypto. So ADL as a strategy keeps the downside risk to a minimum and maintains a high Profit Factor. The Set team has also worked hard to create a great platform. The platform takes care of automatic rebalancing of the underlying assets.
Most of us are also drawn into the non-custodial nature of Bitcoin and Ethereum too. The Set team has set the right vision for what crypto trading should be. Each step is made deliberately to move us in the direction that you should own the keys to your assets. But there are still trade-offs. I caution followers that any smart contract on ETH carries the risk of failure. It should be treated with due diligence and care. Trust, but verify. Past performance does not ensure future performance.
As a software researcher and data scientist, I care a lot about writing good software. Everyone should buffer a reasonable amount of risk. There is so much I can do to mitigate the software risk. My job is done when I’ve finished writing ADL. Amrullah Deep Liquidity is hosted and operated by partners such as Tradingview and other firms now. I do not host the software. I also rely on the Sets platform to do the heavy lifting for the trades.
May the trades always be in your favour!
Conclusion
We hope you enjoyed reading this interview! Amrullah will be one of initial launch traders when Set Social Trading goes live in the next few weeks. If you’d like to be notified when Amrullah’s Set goes live, head here to sign up!
Learn more about Set and join our community
Newsletter | TokenSets | Website | Medium | Twitter | Telegram
