This is an advanced article regarding different trading styles when using Algorithm Intelligence. We would recommend reading the basic documentation of the Algorithm Intelligence first.
When creating a new AI, it is generally more profitable if you use the AI for a specific purpose. For example, you can create an AI dedicated to scalping or swing trading, but not both simultaneously. If a strategy is made for scalping, it usually works on a much shorter timeframe than one made for swing trading. Lets now dive into how you can create the ideal algorithm for you!
First, decide what type of algorithm you want. For this example, we will go with a scalping algorithm. We will thus need to create several scalping strategies. In this blog, we went over 3 models of scalping strategies. We have created these strategies on Cryptohopper for your convenience. We do recommend reading the following section. However, you can also scroll below to download the strategies (adjusted for the AI) directly to your Hopper. Let's now go over the three strategies:
Momentum Trend Scalper
Combining momentum and trend indicators can work wonders. For this strategy, we have also selected the MESA on the 4h chart to identify the broader trend. We have then chosen Williams %R to identify when the market is oversold and then the crossover of the 1 and 15 EMA to determine when the momentum is back in our favor. We are using a higher timeframe of 15 minutes as both Williams %R, and the crossover of the 1 and 15 EMA tend to give out a lot of signals on the 5 minutes chart. It is important to note that Williams %R and the EMA crossovers are very unlikely to give out signals simultaneously. Therefore we have designed the strategy so that Williams %R signal is kept for 5 candles.
For the exits, we have chosen the following sell settings: Take profit 7% Stop-loss 1% Trailing Stop loss percentage 1%, arming 1.5%
Pure Trend Scalping
This strategy involves using trend-following indicators on multiple timeframes. It first uses the MESA on the daily and 4h charts to make sure that the trend is in our favor. Then on the 5 minutes timeframe, we have the crossover of the 5 EMA and the 25 EMA.
For this strategy, we will use the same exit settings as the previous one: Take profit 7% Stop-loss 1% Trailing Stop loss percentage 1%, arming 1.5%
Volatile Trend Scalper
Let's finish with our most recent analysis on TradingView. This scalping strategy involves using the MESA on the 4h chart to determine the direction of the trend. The crossover of the 15-50 EMA on the 5 minutes chart and an ADX value greater than 20 is then used to determine the entry.
For the exits, we have chosen the following sell settings: Take profit 5% Stop-loss 1% Trailing stop loss percentage 0.5% and arming at 1.5%
We will now move on to creating a suitable trend strategy for our AI. Trend strategies generally need to keep signaling "buy" in uptrends and "sell" in downtrends. The Parabolic SAR on the daily chart is a good example, as it keeps the buy/sell signals according to the trend. However, if you add the Parabolic SAR on multiple chart periods with the option "necessary signal" active for one of the chart periods, you will also get neutral signals.
Therefore let's make a simple Parabolic SAR strategy on the daily timeframe that does not generate any neutral signal.
The three strategies we have previously showcased use the MESA on the daily chart as a filter to only trade in an uptrend. However, as the AI already uses the "trend strategy" feature, this filter becomes obsolete. Therefore, it is preferable to remove the MESA filter from the three strategies.
Here you can download the Parabolic SAR trend strategy we have showcased earlier: Market Trend Parabolic SAR
Let's add them now to the new AI to see how they perform. Please remember to hit save after adding all of your desired strategies.
In this section, we will go through the settings that we have chosen and explain the reasoning in our choices.
Validate signals after X time: Scalping is usually done within 24h as it is essentially short-term day trading. However, due to low volatility during some periods, 1 day may be too short for a position to reach its profit target. Therefore we have chosen two days to be on the safe side.
Minimum percent change buy signal: All of our strategies rely on a Trailing Stop Loss armed at 1.5% to activate. Thus we have selected 1.5%, as this is the minimum threshold for a positive trade.
Minimum percent change sell signal: Here, we have also selected 1.5%. However, in our case, this is not important as we do not have any selling indicators (as we are selling with the sell settings).
Maximum percent change neutral trend: We have left this at the default of 5%. This is not as important for us, as we have chosen a trend strategy that doesn't provide any neutral signals.
Minimum percent change uptrend: We have chosen 2% here, as the trend strategy should experience a more significant move than the strategy itself.
Minimum percent change downtrend: We have chosen 2% here, as the trend strategy should experience a more significant move than the strategy itself.
Maximum score for neutral signals: We have chosen 1 here as we do not have any neutral signals, and we do not want to give any importance to them.
Since we have only analyzed how the strategies behave on BTC/USDT, we have chosen the pair BTC/USDT to have the most accuracy. Since these are scalping strategies, we should still have enough signals to get an accurate picture of how well they can perform.
As can be seen in the editor section, the scalping strategies have performed well. They had over 75% of the trades reach the 1.5% mark.
As predicted earlier, we have had a significant number of signals (almost 200 in total). If you look in detail at each of the strategies (you need to click on them), you will find that the winning % is more significant in an uptrend market than in a downtrend (hence why they had the MESA as a filter initially).
This is also evident when you look at the results by "Trend" in the "Results" section. The uptrend signals are around 20% higher than the downtrend ones.
Now that we have trained the strategies with our AI, let's configure our hopper. To do this you can go to config/baseconfig. Let's now go through the most important settings that we have selected:
Order type: Limit orders can be considered better in our case as we have a scalping bot, and in scalping, even a small difference of 0.1% profitability per trade is significant.
Max open buy time: Five minutes can be the right choice for the "open time buy". We do not want to get in a trade an hour after the opportunity was found.
Max open positions: We have set the max positions to 1 as we are only trading against BTC, and we haven't set any DCA. If you are trading more coins, then you may want to select more. "Max percentage open positions per coin" and "only open 1 position per coin" were also set accordingly.
Coins and amounts:
Base currency: USDT should be selected here, as we trained our strategies against USDT.
Percentage Buy amount: Since we only open 1 position at a time on BTC, 50% could work well here. If you are new to Cryptohopper and trading in general, you may want to select a smaller % first to learn, before risking large amounts of your capital. You can also choose more and trade with all of your funds at once. However, if you decide to do this, make sure to select 98% and not 100% as you need to leave some free funds for the fees and such.
Minimum USDT amount per order: Here, just make sure you trade with more than the minimum required by your exchange.
Take profit: 7% was the "take profit" we were using with most of these strategies. Therefore it should fit well with the scalping bot.
Order type: "Limit orders" will also generally be the right choice because, as we stated earlier, small differences can be significant when scalping.
Max open time sell: Five minutes will be a can work here too. If the hopper is unable to sell your position in 5 minutes at a specific price, it is better to cancel the order and try to sell it again at another rate. It is generally more advantageous to have a shorter selling "time sell" so that the order can be canceled if it is not filled and placed again.
Stop loss: 1% can work here, as that is what we use for most of the strategies on their own. You could test out wider settings here (such as 2%) and see how that works out.
Trailing stop loss:
Trailing stop loss percentage: 1% should be fine as it gives your coin room to grow, and it is also sure to lock in profits once it is armed. (if you have at least 1.5% as the arming level, and are using an exchange with low fees such as Binance or Kucoin).
Arm trailing stop loss: 1.5% should be the obvious choice here as we have tested the AI with 1.5%, and our Algorithm is based upon our strategies reaching 1.5% every time.
This is the most crucial part when loading the AI into your hopper!
Strategy: Select the AI that you have just created. For us, this is "Scalping Bot Example".
Signals only: This should remain disabled.
Number of targets to buy: This one is set to 1 as we only have 1 open position at a time and 1 coin.
Use corrected score for buys/sells: "Yes" is usually the better choice here, as the "corrected score" take your real-time data into account and allow your AI to improve.
Minimum score for buys/sells: A decent choice here is around 80% as you want to be selective with the signals your AI acts on, for the best profitability.
Congratulations, you are now ready to start trading with your new AI!