A lot of things going on at the moment. I did some infrastructural upgrades on the backtesting engine, database, and admin dashboard. This is to provide me better status, data, visuals, activity, and performance feedback. The entire CoinBakers server infrastructure spans over 23 servers, not counting any client trading servers, each with their own dedicated tasks.
The reason the infrastructure, both hardware and software, is important is because it speeds up the actual development of strategies and provides better feedback for deeper analyses. The strategies are only so good as the environment it is developed in.
I'm sending out this mailing earlier to discuss the V5 EUR position trigger. I didn't expect for V5 to trigger. However it did at 3AM UTC. Because of my concerns outlined in my last blog post I paused V4 from trading as V4 has increased risk in this situation. My general concern related to the whale pulling out if the market doesn't move up, still stands. As we do not have a proper solution yet to properly de-risk this situation. At least not something I'm confident about.
I therefore closed our EUR positions with ~0.5% loss once I woke up this morning and saw our EUR position.
The markets might move up and we might miss out on profits, but that would be a pure guess and with increased risk as the buy trigger was not naturally formed due to the pump. Nothing wrong with that, but as long as we do not have measures in place when the rug is pulled out under us, we should tread this line carefully.
Missing out on profits is no problem, there will be many opportunities in the future.
V5 runs separately on each currency pair so it's possible EUR will trigger while USD might not.
I will enable V4/V5 once the whale pulls back, finish the supportive strategy, or when the market situation kind of times out in time. I'll keep monitoring it for now. Hopefully any of these by next week.
I'm going fully focus on the supportive strategy for now and integrate the short term solution into that. The supportive strategies' job is to advice V4/V5 on current market situations based on different time-frames. V4/V5 will then use this advice to decide if they go into a position or not. This is better use of my time as the supportive strategy was already planned.
At the moment I'm live running a new strategy that is able to trade crashes. It's definitely in its early stages, but it looks promising. Check backtest performance below. One interesting fact is that it throws out any concept I used in V4/V5 to make this work. Which makes it some sort of diversification when running side by side with a V4/V5.
The reason why C1 (C stands for crash, as it trades crashes), is important now is because it's able to make money in December and January. While V4/V5 struggle with this.
In this backtest screenshot you can see "N", which is the net result based on fees and slippage. Binance would have better result as the fees are lower there. This btw is CoinBakers custom backtest engine.
Most importantly about this strategy is that it's working well in recent periods. Once it makes its next move I'll share how it performed and possibly have an early release. Stay tuned.
The development of C1 brings us to the next topic. By introducing V5, we can also somewhat reduce our risk by running two strategies side by side. If we take this one step further and diversify using 3 strategies, for example, V4/V5/C1, we reduce our risk even more.
A master strategy that can manage all other strategies by dynamically assigning the optimal portion of balance, and prioritizing and deciding which signals to take and assign to new positions. It wouldn't take in raw data by itself, but rely on the advice of strategies. Say for example, V4 signals long, a crash happens triggering C1, it would decide to pull out of V4 and assign 50% to C1. If then during this period V5 signals a buy it would add another 20% or the rest of the portfolio to its position.
Based on market conditions it can decide to override certain signals, add or remove to the current position.
The best thing about a master strategy is that we can feed different external data like sentiment, tweets, hashrate, market conditions etc. and all the strategies by themselves are single purpose. For example, looking for pumps, trading crashes, etc. This makes updating and keeping the strategy performant over time easier and more effective.
Basically one strategy to rule them all.
This is something that I strive to develop in Q2 2019.
Another strategy I want to explore is to use predictive analysis. I have a huge dataset of over 4 years of minute candle data. What I plan to do is to create a fingerprint over different sets of candle patterns on different periods. We can subdivide them into strict and loose matches to form a confidence score for let's say the next 5 candles. Let's see how this will unfold and where it will take us. Probably looking into Q2/Q3 for this one.
Besides all the stuff that will happen in the future, I'm also excited to announce that we are now officially supporting Binance. You can choose Binance during signup or for existing customers, you can click the "Update API keys" in your dashboard at the bottom of the page.
Simply follow the instructions and once saved, your trading server will instantly use Binance. Reasons to use Binance are lower fees, 0.075% compared to Coinbases' 0.3%. And 0.1% if you do not hold BNB token from Binance. If you check the backtests, you can see this can make a huge difference longer term.
We are supporting BTC/USDT and BTC/USDC. USDC is native to CoinBase as it has been developed by CoinBase and Circle. At the moment I do not have measures in place for a collapses of stable coins, but expect this in the future. If it's implemented I will mention this in a blog post. I personally do not trust USDT. If USDT collapses, or any other stable coin, we should have measures in place to quickly get out.
By having multiple strategies now and support for multiple exchanges, its only natural to support multiple trading servers. For example one running V4 and another running V5. Or C1 once released. It will costs $29/mo for every extra trading server on top of your first server and will appear in your dashboard with its own data.
This will be supported soon once the supporting strategy is finished and everything is resumed.
In a later update when things settle down a bit I will improve the app dashboard to give some more feedback on where we are at for triggering a new position. Think like 5/8 indicators are OK for next position.
That's it. More next time about the supportive strategy launch 🚀