What kind of information do we need to construct a Grid-Spot strategy for cryptocurrency trading?
UPDATE: 27 Sept 23
After much considerations on the current market outlook, we have decided to changed our business model to better adapt to this uncertain environment. We will no longer operate a trading bot.
Instead will be open-sourcing our trading tools (the non-proprietary ones), and transition into a community-driven platform business. If you are interested (early membership will be free as we build the platform together), please take a look here: https://waarn-finance.gitbook.io/waarns-philosophy/
UPDATE: 16 Nov 22
Due to recent events with FTX, we will move our operations to other exchanges once the dust settles. All trading activities and blogs will be paused until then. We apologize for any inconvenience.
Introduction
In our previous [article] we explained our own Grid-Spot trading system. In this article, we dive into the details of what information is needed to successfully implement this trading system. We encourage you to take a look at the article first, to understand why we have chosen the following information to formulate our analysis, but it is not necessary to understand the article. We will also assume you have some familiarity with Grid-Spot strategy, if not, you can check out another [article] for a gentle introduction.
As a quick recap, our particular grid strategy is composed of three simultaneous trading regimes: the short (7 days), medium (4 weeks), and long (3 months) time-frames. The grids are set according to available information on the asset price movement such as macroeconomic factors and graph technicals.
If you are interested in this trading system, you can use our platform https://waarn.finance to automate it for a small fee. Alternatively, you can follow us on Medium, Facebook, or Twitter for a live update of our Grid-Spot configurations. The update is free of charge, and is intended as a way to grow our audience and advertise our platform.
How to formulate a Grid strategy
To be able to configure a profitable Grid-Spot strategy, we must first understand exactly how our analysis informs our strategy. For example, we might want to set up a grid that will trade the SPX for the next week. Say interest rates are going to be announced in two days, and the GDP announcements are coming up in five days. We then have two major events that could make a large price fluctuation in the next seven days. With that, we can try to determine whether the price will be bullish, bearish, or unaffected, given the available information, and accordingly set up the boundaries on the grid. The bars on the grid are usually determined by technical analysis which give us support and resistance. By placing the bars judicially, we can minimize risk. Once new information comes in, or an event occurs, we can continually update the grid according to our hypotheses.
What can we use to inform our Grid configurations?
The key is information. Not just any information though; we want highly relevant information and as much of it as possible, as fast as possible.
Institutional traders use fund flow analysis, tune their order execution algorithms, and try to front-run each other. But these concepts do not have to worry us. In our experience, there are only a few broad categories of freely available information you’d need to successfully construct a grid portfolio.
What are these broad categories of information?
Historical Price and volume
First, we’d want historical price and volume data from our exchange. We will also look (if possible) at historical order book data, aggregated data from all exchanges, and other relevant data like the S&P500, the VIX, US housing prices, and even semiconductor ETF data, and graphics cards prices. These form a cross-correlation reference which gives us a window into the relevant markets that may affect cryptocurrency prices.
Options data
From options data, we specifically want to know implied volatility, open interest, put-call ratio, 24-hr buy-sell volume, among other things. Options data is a good representation of ‘smart money’. Professional traders often use options to hedge against their positions, and this gives us a glimpse into what they are thinking.
Fundamental data
Next is fundamental data. Much of this is qualitative rather than quantitative; however anything qualitative can be quantified. Fundamental data includes microeconomic factors such as protocol revenue, price-to-earning ratio changes, on-chain data like number of active wallets, transactional activities, and short to long term holders ratios. This information is less important as most protocols are run by private companies. This means on-chain accounting information may not be a true representation of protocol value. However, what we are really interested in here is the relative value over-time, which can be used to form a view on future changes in the asset’s underlying value and its volatility.
Macro data
Another important category of information is macro data, which can be separated into 2 categories: macroeconomic data, and regulatory data. The specific data depends on the politico-economic climate at the time of analysis.
Macroeconomic data include GDP, household income, interest rates, inflation rate, PMI, and employment data among others.
Regulatory data include statements by the Fed, introduction of crypto-related laws, and politicians’ stances on cryptocurrencies. Quantifying these is best done by human judgment.
Sentiment data
The last category is sentiment information. Again, this is usually qualitative, but can be quantified. In fact, plenty of AI has been developed to accurately predict positive and negative sentiment based on social media and news data like Twitter, Bloomberg, Financial Times, among others. Google trends can also help us determine interest in certain projects and protocols.
Understanding the role of information on trading
Now is the part where we synthesize the information into market scenarios.
First, we should talk about the principles of how to “predict” something, how they are useful for the Grid strategy, and then the approach to predicting them.
The principles of prediction
We don’t need Beysian probability theory or understand complex mathematical models to make relevant predictions for a profitable Grid. Let’s look at the three key general principles of predictions that we need.
- The closer an event is, the more accurately a prediction can be made regarding its outcome.
- The more information one has, the more confidence one has in their prediction, as long as the information is relevant to the outcome.
- Every prediction always falls prey to the risk of a “Blackswan” event, often with major consequences. A Blackswan event is an event that is extremely unlikely according to the prediction model, but occurs anyway in the real-world.
With that, let’s look at how we should be constructing three grid setups for each time-frame.
Short time-frame grid (1–7 days)
From Principle 1, we assume that the closer to the time of the price movement we are trying to predict, the more accurate our prediction. This is good for our short time-frame grid. However, we also have far less relevant information to predict a 1-hour price move than a 1-week price move. And according to Principle 2, less information means lower accuracy. Short time frame grid setup also suffers heavily from Principle 3, as they are more sensitive to sudden news and events compared to medium and long time frame grids, which have wider range and can tolerate larger price moves.
Long time-frame grid (1–3 months)
From Principle 2 alone, we assume that our long timeframe grid setup will be more accurate, as we will have more relevant information on hand. This includes options open-interests data, short to long-term token holder ratio, and macroeconomic viewpoints, among others — information which are not as useful for shorter time frame grid setup as shorter time frames are usually influenced by more immediate factors like news and sentiments. And while Principle 1 is not in favor of this timeframe, the amount and quality of information available for longer time frame prediction outweighs the cost of uncertainty. We will also go into more details on this in future articles where we analyze the market weekly to set up our grids.
Still, while we may be more confident in the long time-frame grid setup, because it has a wider range, we need to wait longer for this to be profitable.
Medium time-frame grid (1–4 weeks)
Finally, we find the medium time frame grid setup to strike the ideal balance between Principle 1, 2, and 3 of predictions, while still being relatively quick to turn a profit. We are at a short enough time period that our prediction is likely to be more accurate (Principle 1). We also have enough relevant information to be confident in our predictions (Principle 2). And while a medium timeframe prediction suffers from Blackswan events, the range is wide enough to tolerate most of these events (Principle 3).
So why set up three grids if medium time-frame is the best?
Ultimately, it is about finding the balance between risk and reward. Short timeframe grid setup can generate more profit in a shorter period of time, but it is riskier; Long timeframe grid setup requires more time, but it is safer. Allocating your money towards the three timeframe grid setups is how we adjust the strategy to our risk tolerance.
However, It is important to note that we always allocate to all three grids, as any of the setups can deplete our capital if we only allocate all our money to just one setup at any given time.
Now, let us look at some of the information we can obtain to inform our grid setups. We will look at how we can use it to form a bull/bear case, and find the volatility and support/resistance level of assets at a certain time — metrics which are important to deciding our three grid ranges.
Gathering compelling information
Long timeframe (1–3 months):
The first thing we analyze is the long time frame, as that will form the overall market view, which will help inform what we should be looking out for when constructing the other 2 grids. Here are the information we need:
General economic indicators — while this usually isn’t significant, it can form a framework to view the economy as a whole. We know that cryptocurrency price and US stock price correlation is generally higher during market corrections, so if we expect the economy to be in trouble, we can reasonably expect the price volatility to be high. In fact, we generally prefer to look at economic indicators for bad signs, since outside of bearish periods, stock prices and cryptocurrency generally have low correlations.
Options open interests — for longer timeframes, this information is very useful, as we can see traders’ expectations of the asset’s price support and resistance level. The general rule of thumb is that we skip a couple of strike prices nearest to the current price with the highest open interests, and find a wider range that also has relatively high open interests, and use those as our initial estimates for the upper and lower bounds of our grid system. The intricacies of exactly how we use this information will be available in future articles.
Price and volume data — there is a lot you can do with this. But the first step is to build a “bull-case” and a “bear-case”. Pick your methods, whether it be Fibonacci extension/retracement, Elliot Wave, or good old moving average. The important thing is that we want to confirm whether the signal for bear case corresponds to the lower bound that we get from Options information, and vice versa for the bull case and upper-bound.
Volatility measurement — The next thing is to construct a volatility model. We generally prefer GARCH(1,1) due to its simplicity, but you may use your own preferred volatility model. For a long time frame, if you want to use implied volatility (obtained using Black-Scholes), you should be careful as most IVs are calculated from options that expire in 30 days. Again, here we are trying to see whether the distance between our upper and lower bound respects the volatility measurement or not.
There are a few more factors you may want to consider for a long time frame grid configuration. These include whether the protocol is viable long term, the number of wallet transactions, and the ratio of long-term holders to short-term holders.
Medium timeframe (1–4 weeks):
Ideally, we are aiming to be in-profit within 2 weeks. However, if volatility is low, we may extend it to 4 weeks to allow the price to rack up enough buy-sell transactions. Some information to look for are as follows.
Economics calendar: this includes announcements of GDP, inflation rate, interest rates, among others. We are also looking for periods of potentially high volatility in the broader financial market like Triple witching hour (when options, futures, and index options expire on the same day), where volatility in the stock market may propagate into the crypto market. This information should be used in context. So for example, when there is talk about recessions, interest rate and unemployment rate announcements are periods where you should expect more volatility.
Options open interests and other options-related data: we can predict a lot with options data at this time frame since most options expire on a running 30-days period, matching our medium time frame. We can simply use open-interests data to determine our upper and lower bound, although it should be noted that there are whales playing games here where they will manipulate the price so that some unaware traders have their options expire out of the money. Therefore, when setting stop-losses and take-profits, it should not be too close to the range indicated by open interests. Other data include 24 hour volume data, and the put-call ratio.
Price and volume data — Similarly to the long timeframe analysis, we will build a “bull-case” and a “bear-case”. Usually, 1-hour chart patterns work quite well at this timeframe. The point isn’t trying to predict whether the price will go up or down, but how far up and down the price could go. Look for price/volume divergences, rounded number support and resistance levels, trend lines, and triangle chart patterns, among other traditional TA techniques.
Volatility data — Implied volatility can be calculated from options data using Black-Scholes. If you are quantitatively inclined, you can construct your own model and derive the implied volatility. More mainstream indicators include Donchian channels, the Average True Range, and even Bollinger Bands or Ichimoku Clouds. You may also use DVol, a VIX-equivalent created by Deribit (current market leader in crypto options trading).
Sentiment data — At a medium timeframe, sentiment analysis starts to become important. We can go fully manual for this. Find a few Crypto Twitter influencers and check their tweets. Ideally, we’d want to cross reference their Twitter posts and price data, but it’s not necessary. Instead, look for contradictory viewpoints. If they all say the same thing, it is likely that the market is trending in one direction, but if there starts to be a fracture in sentiment, then it can be a good representation of uncertainty and high volatility.
We also want to look for signs of conviction. Tweets and articles that ‘hedge’ their opinions is a sign of low volatility, whereas tweets with high conviction tend to give clear price targets supported by charts. For us, we use a trained sentiment-analysis AI for this analysis. While there is a correlation between price and sentiment, it is usually a lagging indicator. But we find they are sometimes a leading indicator of increased volatility. The best way to use this data is as a piece of supporting evidence, not as an indicator in itself.
Fundamental data — There is a lot we can tell with on-chain data; one example is to look for large transactions, strange non-repetitive movements in the Treasury wallet (if you are trading platform tokens), or other breaks from the usual patterns of transactions. This could signal internal troubles, ongoing hacks, or new product launches, organizational changes, etc., which will significantly impact price movements. We will go into more detail in our future articles on other relevant data.
Short timeframe (1–7 days):
We are almost there. The last grid construction in our strategy requires less information than our medium and long term grids. It is a high-risk high-reward configuration, and it will be the profit-driver out of the 3 grids. Here are the information you need to know:
The economic calendar: similarly to the medium term grid, we want to know the specific dates of economic announcements and basically avoid trading during periods where important economic metrics will be announced. The key is to not avoid all important economic events, but only those that are most influential in a particular economic climate.
Options data: at a short timeframe such as this, we can ignore most of the options data, since it is generally a representation of long term traders mentality. However, that is not to say it is useless. We can still obtain some insights into the potential grid range and the take-profit and stop-loss points by looking at the nearest strike price with the highest 24hr volume, and the relative put-call ratio at those prices.
Price and volume data — This is probably the second most important data. At this time frame, we are competing against day-traders, high-frequency trading bots, and other retail traders who essentially trade on news and sentiment. Technical analysis is pretty good at capturing this short-term market psychology. Our ideal predicted scenario should be when there is no clear bull or bear, and when there is relatively low volatility. This may sound paradoxical since grid trading profits off volatility, but what you want is to trade the range and high volatility can indicate a coin is beginning to trend. We also want to err on the side of caution due to Grid-Spot’s long-bias nature.
Volatility data — This is by far the most important data set, and it is derived from the price and volume data or options data most of the time. We want to make sure that the ranges we set for our short time-frame grid respect our volatility model. So if the model says that in the next 3 days the volatility will likely be 2%, and our analysis shows that the price will likely be ranging, we would want to set our range between 2–4%.
Sentiment — For this time-frame, we want quiet — not loud voices. We prefer to trade during relatively low volatility, and sentiment analysis is a great tool to understand how ‘sure’ the public is of the market. In a short timeframe, we want lower volume — figuratively and literally.
This is it for a short time-frame grid. It’s important to not get caught up in too much information, because we are one whale’s MO (market order) away from destroying our setup. Shorter time-frames necessarily mean lower volume in the order book, and therefore any large fund flow in either direction can easily shift the support and resistance levels.
To put it another way: you cannot win if somebody decides to flip the table on you.
Conclusion
Now that we know which easily obtainable information we can use, we are ready to construct our Grid-Spot portfolio. For more details on the logic behind the construction, you can check out further information [here].
If you are unsure of this methodology, we will be publishing live grid set-up weekly, where you can follow the progress of our trades. So if you are interested, feel free to subscribe to our Medium channels to get these updates. We will also occasionally publish articles based on our other strategies and general opinion pieces relating to the cryptocurrency market as a whole.
At https://waarn.finance we are developing an automatic grid trading tool that allows you to copy-trade this Grid-Spot trading system without constant monitoring for a small fee. This is under construction, and we will update this article once this feature is ready to ship. Meanwhile, you may enjoy free usage of our manual Grid bot! We also have a couple more strategies that you might be interested in. We hope you take a look at them and try them out while our AUM capacity is still not full.
WAARN Finance Quant Team
