10th June 2021.
Welcome to the Bonfida Ask Me Anything (AMA) with Overhead Gains. For the month of June to promote the Bonfida x Serum Trading Competition we have lined up a series of guests to help provide insight to automated trading and how to get involved.
The transcript below follows the format as BOLD TEXT is interviewer and NORMAL TEXT is the interviewee.
— — — — — — — — — — — — — — — — — -
Welcome to our AMA with @OverheadGains!
For the next hour we will be discussing the technical aspects of building some of the best bots available in this space.
Hello everyone, happy to be here!
Glad you are here.
So, tell us a bit about yourself!
I worked in finance 15 years ago, and got into engineering 10 years ago, so I guess my life kinda prepared me for algo-trading.
In the past 10 years I worked at many different companies, being involved with various projects and technologies. I developed front and backend, got into game development, shipped games for consoles, mobiles and PC, worked with autonomous cars, and the list goes on. My main interest always revolved around AI, and I always found it fascinating what can be achieved by translating seemingly complex tasks that only humans can perform to a language that even computers can understand.
I’ve been building various bots for a long time now. I worked on online poker bots, automated sports betting, did arbitrage on both sports betting as well as crypto, experimented with automated DCA/Grid strategies and smart-trading, explored various portfolio manager strategies, and the list goes on. I love to automate stuff, and all the experience learnt during these adventures can be used going forward.
That is an impressive resume you’ve got there OverheadGains.
AI is moving in leaps and bounds in this past decade.
Glad to see that your experience and knowledge is all transferable skills and can be deployed into algo/bot development.
How long have you been involved in cryptocurrencies?
I got into crypto in 2017 January. Been actively trading forex and stocks before that and the platform I was active on added crypto so I thought I’d check it out, and I instantly fell in love with the premise of blockchain. I got a friend who mined BTC back in 2013, but I didn’t take it seriously at that time.
That is impressive.
I guess all your trading is automated at the moment?
Correct. I don’t trade futures manually, and my spot buying is more of an investment, rather than a short-term trade.
This transitions onto our next question perfectly.
In terms of strategies, how do you feel about automated trading vs human trading?
I prefer automated trading a lot. I believe most strategies can be translated to code with the right mindset. It can be difficult at times since our brain can recognize patterns in a way that’s not straightforward to code, but on the other hand, we are also mostly driven by emotions, and taking this out of the equation is a big motivator for me.
I agree with the notion that emotions interfere with our trading capability.
Even though when I trade I strictly stick to no more than 2–3 indicators. Try to keep it simple but there is the issue with human trading and your own bias.
Bias is an absolute monster. There are often times when I don’t agree with my bots’ decisions — but in the long run, they always prove that they know better than me.
Couldn’t agree more! When it comes to trading do you feel that there is one strategy that can be applied to all market conditions?
I think it highly depends on how you define a strategy. For example, if you are a breakout trader who uses triangles to identify potential setups, you can do this all the time, but you will be presented with more opportunities and better profits during a bull market compared to a bear market.
On the other hand, if you trade based on a simple crossover strategy, for example, you will be able to identify profitable trades regardless of where we are in the market.
Makes sense. So there are different bots that are better suited for different conditions.
Like recently, we had some significant chop in the markets. So a scalping bot would’ve thrived more compared to a trend bot.
How many bots do you have published at the moment?
I got 2 bots on the market that are publicly available, one for BTC and one for ETH.
Would you like to introduce your bots to the audience?
Currently the bots are available on Stacked Invest as well as on CompendiumFi. You can access historical performance and subscribe if you are interested here:
Earthshaker (BTC) — 93% profit in 3 months:
Dreamcatcher (ETH) — 480% profit in 3 months:
Those are impressive stats. And that is running 1x leverage!
Talk about beating the market
So lets say we want to build a bot. What would you suggest for someone looking to create one?
How would you go about planning one?
The most important advice I can give (and what I’m always telling people who are looking at algo-trading) is to start simple. I can’t stress this enough. Don’t go and build your own trading simulator in python or c++, download all the historical data, make sure executions are correct, take fees into account then start building visualization and basic indicators… Been there, done that. Just don’t. By the time you get to a point where you can start considering a strategy you wasted so much time you might even forgot why you even started.
The best thing to look at is TradingView and their scripting language, Pine script. Even with a few lines of code you can set up a basic strategy and then backtest it on various assets and time-frames with excellent visualizations and deep performance analytics. It’s just as easy as it sounds. Even if you aren’t code savvy, you can learn the basics in just a few hours and then look at automating your strategy. TradingView gives you all the tools you can possibly need to verify your strategy and see how it works. There will be times where you will hit limitations — faced some too. But at that point you already saved yourself several hundreds of hours by not going the hard way and developed everything from scratch, and you now have a viable strategy that you know will work, and then at this point you can start considering going native and developing your standalone bot using your favorite language if you still need to. But don’t start with this one!
So to sum it up.
Take into consideration all the factors
Build up slowly and get familiar.
In terms of building a bot how would you go about choosing leading or lagging indicators.
What do you find effective for varying market conditions?
I think each indicator has a good purpose and you just need to figure out if they are a good fit for your strategy or not. If you want to trade trends, you will never be able to catch exact tops and bottoms, so lagging indicators are more reliable. For quick scalps, you want maximum noise, so you need to take leading indicators into account. I think here, balancing is key.
I see. So different indicators for different approaches.
Should you pair two types of indicators together to provide more confidence in your bots trading plan?
Combining multiple indicators can be an interesting approach, but they also carry an inherent risk with them. What I mean is if you just blindly add a bunch of indicators in your strategy and then tune their parameters until you get a strategy that performs well, you will eventually risk what I’m calling overfitting on historical data. Yes, you created a strategy that worked very well in the past, but you created this strategy in hindsight, and it’s entirely possible that the market will never reproduce the patterns you fine-tuned your bot for. There are lots of strategies out there that are marketed only by using backtested data and suffer from this phenomenon. The main reason I prefer Stacked and CompendiumFi as a bot platform is because they show real, historic trades that happened in real markets, and are immune to backtest-manipulations.
That is often misleading only marketing a bot on his backtesting.
Glad to see there is credibility in reporting honest bot performances.
So with too many indicators it can lead to “noise” that affects bots performance?
Or as I like to say. The bot could be sucked into a “fake out”
It depends on the indicators used, but I think most indicators will help you to reduce the noise compared to looking only at the charts, so eventually, multiple indicators will reduce more noise, and will lead to potentially “perfect” trades. Just need to be on the lookout not to chase perfection too much.
Preventing fakeouts is a real challenge. Obviously, you want your bots to be resistant to them, but on the other hand, you also don’t want it to miss potential good trades. In most cases, fakeouts are not identifiable in advance, you can only tell in retrospect that it was a fakeout. In my strategies I spent a lot of time on identifying potentially unwanted market conditions where the bots just shouldn’t issue new signals since there was a high probability of fakeouts, but obviously, they can’t be eliminated completely. For example, the current long positions opened this morning that reacted to the market might be a fakeout, but it might be a continuation of yesterday’s movement. If the former, we hope the bots will quickly realize this and go back in a short, while if it’s the latter, we will be happy that we caught it. Only time will tell.
That makes sense.
So reducing the noise will provide more confidence on the signals being produced
Its difficult, especially in the current environment.
So, your bots Dreamcatcher and Earthshaker are always in a trade?
Yes, the bots are always in a position (so they either long or short the market). They won’t win all trades, and have a ~50% win-rate. However the wins will always outweigh the losses in the long run, and this is what makes them the most profitable strategies on 1x leverage right now. The only bots that have higher returns in the past 3 months are running on 2x or higher.
That is great! On 1x and you are able to beat the competition.
What time frames would you suggest for a scalp and trend bots?
I saw trend bots working on 15 min candles and scalp bots on the 4 hours. Generally speaking, if you want to catch small moves with high leverage you want to look at low time frames, usually 3, 5, or 15 min candles, while if you want to trade trends, it’s better to eliminate the noise and focus on 1–4 hours or even dailys. I wouldn’t automate strategies that trade on the weekly or monthly candles, as they are just too slow for me, but on those levels it’s pretty easy to create a very slow strategy with a very high win-rate.
That is interesting.
I always thought that trend bots would favor more of the higher time frames.
However, I’m guessing that it is all subjective due to the indicators you are using
My bots use 4 hour candles as I found it to be the sweet-spot. They usually capture trends successfully, while also able to scalp smaller moves.
Never seen a bot that has been able to do both successfully.
What were some of the best trades you’ve bots have done and some of the worse?
The best trades from a single position for both algos were above 100%. They don’t happen often, but if we are in an “up only” market (and this is where most other bots will shine), the bots can hold a single position for over a month if the market allows it, and during these time periods, big gains can be achieved. If you are looking at Stacked marketplace and see BTC bots that peaked in performance in December 2020/January 2021, this is exactly what you are seeing. :)
As for worst trades, we just had the biggest loss on Dreamcatcher almost 2 weeks ago, which resulted in a 23% loss. Now to go back to an earlier point, we got caught by a fakeout, followed by a slowly bleeding market without any significant price action that the bots would have acted on, in a zone where they also wanted to avoid overtrading. This trade was unfortunate, but they happened in the past and will happen in the future. What matters more is that the bots will eventually get back in sync with the market and continue winning. That’s what makes them profitable.
That being said, in a trending market most bots will do well.
I guess the real test for any bot is to successfully navigate a bear market and consistently beat the market in its returns.
So @OverheadGains, where could our audience find out more about your bots?
The best place to start is my website, where detailed trade history, statistics and lot of helpful information can be found:
I also recommend following on Twitter, where some of the bot signals are posted, as well as all info concerning bot availability and promotions. Very rarely I also share stuff I find interesting: https://twitter.com/OverheadGains
And your bots are in the Bonfida Bots trading competition as well.
How are they performing?
That is correct, they can be found in the current competition. The pools only support spot trades right now, so while above performances include wins on both ups and downs, here we can only profit from the market moving up. So if you are afraid of futures trading, check out the Bonfida pools and you can still enjoy some action of the bots there. ;)
Thanks for being our AMA guest tonight.
It has been a great insight to the technical aspect of building algos/bots.
If our audience has any questions please feel free to comment here and we will get back to you.
Thank you for having me, and for building Bonfida. I believe it has a really promising future!