The Ichimoku Kijun Cross strategy (Patreon)

Hello my dear readers,

In this post I am going to discuss the performance of the Kijun cross strategy, which is part of a series of Ichimoku trading strategies. This strategy is well described in the book “How to Make Money Trading the Ichimoku System” by Balkrishna M. Sadekar. In that book the Ichimoku indicator is used as a baseline to create different types of trading strategies.

The Ichimoku Kinko Hyo, also known as the Ichimoku Cloud, is a technical indicator used to gauge momentum along with future areas of support and resistance. It was developed by Goichi Hosoda, a Japanese journalist, and published in the late 1960’s. The indicator is comprised of five lines called the tenkan-sen, kijun-sen, senkou span A, senkou span B and chikou span.

In my video and blog post on “Developing a Crypto trading bot strategy with the Ichimoku kinko hyo indicator Part 1: Introduction” I gave a detailed overview of these indicators and also how to create them in a notebook and Panda’s dataframe so that you can create your own algo strategy, based on this indicator.

I explained The Ichimoku kinko hyo Kijun cross strategy in more detail in an earlier blogpost and video. and there I also described how to create the strong signal version of it.
For my Patreons I created an additional video where I explained how to make the additional neutral, weak and no kumo versions of this strategy. I also provided the notebook I created these in over on that platform.

In this post I will present you the results of the backtests of all these versions to see which one gives the best results. By doing these backtests we will try to get a sense on the best performing variant an the possible future performance of that version.

Also, I will not only test the spot trading version but also the futures version where there are long and short trades. So in total I will test 8 versions of this strategy to see which variant performs best.

The strategy

Let me first explain the strategy and it’s signal strength versions and then proceed with the backtest results.

The complete Ichimoku Kinko Hyo

For this strategy, only the Kumo Cloud and the Kijun Sen should be used.

Also another important note that you should keep in mind is that, in this case I will use the default Ichimoku settings of 9 – 26 – 52 – 26.
I did not test a version that uses the indicator lengths that are sometimes advised for crypto trading. Those are 20 – 60 – 120 – 30.

Entry and exit signals

Anyway, let’s continue with the basic rules of this strategy:

• The Initiate long position is taken when price closes above the Kijun.
• And when price closes below the Kijun a short position is taken.

This signal can occur at different positions from the Kumo cloud. Each position has its own signal strength. For bullish (long) trades:

• If price closes above the Kijun and this crossover is above the Kumo, it is considered a strong bullish signal.
• A weak bullish signal will be considered when price closes above the Kijun but the crossover is below the Kumo.
• If the price closes above Kijun and it is within the Kumo, then it is considered a neutral bullish signal.

For bearich (short) trades these signals are reversed, so:

• If the price closes below the Kijun and it is below the Kumo, it is considered a strong bearish signal.
• Then, If the price closes below Kijun and it is above the Kumo, it is considered a weak bearish signal.
• And finally, if the price closes below Kijun and it is within the Kumo, it is considered a neutral bearish signal.

For the Kijun cross strategy, the stop loss is initially set below the Kijun. In the book there is also a small buffer proposed, but since it is not clear how this buffer is set or which indicator to use here, I will use the close below kijun for the stop loss signal.

The author also proposed to use partial profit takings for this strategy.

The trading bot seems to have this capability. But because it is somewhat tedious work and not easy to do I will not use this capability. I will only use the Kijun exit signal to determine the ideal exit moment in this case.

The code

Now since I explained the code in an earlier post and video about this.  In this post I will only stick to the results of the tests I did on all these 8 variants of that strategy. At the end we will see in the Strategy league which variant scores the highest.

So without further ado, let’s first see what the weak signals deliver and how well the strategy performs.

Weak signal results – Spot and Futures trading

Weak signals spot trading

Spot

The Kijun crossover’s weak signals do not dissapoint and would provide you with some gains overall. Of the pairs I have configured to do the backtest on, 50% respond to this algo and in total would give a 432 percent profit. The negative thing about this is, is its low winrate and large drawdown. The one day timeframe seems to provide the best options here but since the ratios are very low, I would not bet my money on this variant for the future.

The results chart of this variant does not show a steady and healthy growth. My guess is that the bot did nice during the last bull run but you’ll have to face large drawdowns during these bearish conditions.

If you think of this logically, it is actually exactly the behaviour you might expect from this variant. The strategy also (or might I say primarily) receives buy signals from crossovers that happen below the kumo cloud. And the Kumo is a main indication of bullish and bearish markets. And since here only the spot, or long trades, are taken, I might expect that these get stopped out lots of times when the price goes below the Kijun sen once again when diving deeper into the lower price ranges.

Seeing the 683 long entries win only 159 times and exiting with a loss 523 times (which is the 23% winrate by the way) tells me that these large amounts of losses might be partially turned into winners also if I also could have used shorts.

Therefore, let’s take a look at a variant that also is allowed to make short trades and do futures trading.

Futures

Now before I continue, I have to warn you that these shorts results cannot be 100% compared with the spot trades earlier. Not only does Binance not provide the exact pairs for short trading as with spot trading, also the timeperiods of the backtesting data can be different than the spot only version. In most cases these are shorter.
We can however get an idea on the strategies performance when it also is used in futures trading with short opportunities as well.

So let me show the first variant of this strategy.

And as you see below, the difference is not that much. It even performs a little bit worse in total amount of profits. But if you look at the ratios, then you see that this scores much better. Try to keep in mind though that the reverse conditions count for short trades. So if a weak long signal has its close above Kijun below the Kumo cloud, then the weak short signal has its close below Kijun above the cloud.

Weak futures signals

Looking at the enter and exit statistics, we see that there were almost as many long as short trades (353 / 327) and both directions had between 25 – 30% win rates. Which is comparable with the spot trading only variant.

Well, again, the futures timeperiod is not 100% comparable with the spot period. Also, when the bot is in a long or short trade and a similar opportunity exists in the other direction on another pair. Then it is possible that it could not execute the trades becasue of the amount of open positions that the bot has.

Nonetheless, the behaviour of the long and short trades over the last period shows a totally different picture when we take a look at the overall profit plot. It has a nice upward slope and after its maximum drawdown it sort of stayed near its 0 line. Also the bot’s parallelism chart shows that it almost always has its maximum amount of open trades used. So there are no opportunity costs here since the bot is always busy with tradingone way or the other.

If I would have to choose between the spot and futures version to use, you’d probably know which one i’ll take.

Neutral signal results – Spot and Futures trading

Now I will switch over to the neutral trade signals. Here the close above, or close below Kijun should be within the Kumo cloud.

Spot

Seeing these numbers makes me very curious to see the strong signals. A hypothetical spectacular result of 3257% profits makes me also a little bit suspicious because maybe we caught some pumps here that are quite common in crypto. And if I take a look at the best performing assets at that time, then there are certainly some spectacular winners in the portfolio.

My main focus therefore is more on the applicability of the strategy, the ratios and drawdown.

It seems that this strategy does not suffer the faith of the weak spot signals by getting signals below the cloud. In this case the signals within the cloud really do give a higher chance of getting into a trade that has more chance to make some real profits.

The average profits these trades have is more than 9 percent, whereas in the weak signals they were around 5.8 percent. And with more entries in this case, a slightly higher winning rate and lower drawdown, that adds up nicely to the total trading account.

If we take a look at the profits chart, we can discover the crypto pairs that cause this high profit percentace at their vertical lines. The bot is more busy with having trades, probably because there are more signals within clouds than below them.

Unfourtunately this signal version suffers a little bit of the same fate as the weak spot version. The main drawdown starts at the height of the last bull market and we have no point where it ends because the backtest has stopped.

Futures

Switching over to the Futures version of this neutral signals variant, the results get less spectacular but the trustworthyness of the strategy for long and short trades rises. Seeing the rations getting well above one and even almost 3 for the Sortino ratio makes me get more confident.

The drawdown is starting to get to the lower 30’s and the modest increase in winrate to almost 36 percent gives this version the highest score so far.

The pairs and their win percentages also look pretty normal for crypto. In real Forex trading you probably would never get 308 percent profit within 15 days. But in this market, it’s pretty normal.

The most noticing things about the profit chart of this variant are the steady profit increase, the parrallelism and the vertical line of MATIC/USDT.

I still do not like the drawdowns, there are still too many and also way to deep to my taste. But they are getting predictable because every time the market cools down, the drawdown starts. Let’s hope this improves even more when we look to the other results.

No Kumo signal results – Spot and Futures trading

The next signal types I will discuss, have actually no indication on how strong or weak the signal types are. It only trades on the cross up or cross down of the close price on the Kijun sen. So in this case the Kumo cloud is not involved at all and the bot only trades on higher of lower than the Kijun.

Let’s see what the spot results look like:

Spot

The spot results look pretty decent and on the 1 day timeframe it even has the highest profit percentage until now.
Again don’t stare blindly on profit percentages only but look at the bigger picture. Notice that this variant also has a drawdown of 58 percent. The winrate is also only 24% and the amount of pairs that react well to this variant is also not that high.

Looking at the profit plot we see that this variant has a pretty sharp inclination during the last bull run, but its drop after that is also pretty steep, which results in that large drawdown.

Also notice the large vertical line Shiba Inu has in that last bull run. It went vertical and this probably also explains the large profits of this variant.

Futures

Comparing the spot only version with the futures one, we see that the profits are still admirable. The ratios are much better and the drawdown isn’t that much of that of the spot only.

What I’m also starting to notice here is that all the best timeframes seem to be the 1 day timeframe. Even though that the Ichimoku indicator can be used for all timeframes according to the theory. It is possible of course, and it will bring you hypothetical profits according to these backtests, but with not that high of a profit percentage of other ratios.

What is also interesting here is that the amount of long and short trades during this backtesting period is almost exactly the same. 366 long signals to 365 short signals. Strangely enough the short trades actually result in very low profits and that most of the gains are made by long trades. Even thought that the win percentages are also almost the same.

Nonetheless, watching the profit plot again, I think that there is a nice natural growth curve on the profits and the pairs plot does not seem to have these weird spikes in profits on certain pairs. At some point the Paralellism goes black and stays that way. So here the bot is always busy. Either with long or short trades.

Finally the underwater and drawdown plots shows some cyclical patterns, which I recognize are somewhat comparable with the market bullish and bearish cycles. Also, do not be mistaken by the depth of these drawdowns. At some point it is not lower than 25 percent drawdown. So be aware of the resolution of this chart.

Strong signal results – Spot and Futures trading

Now it is time to look at the strong signal types. According to the book these provide the best signals for trading. Hopefully by doing these backtests we can also confirm this theory.

Spot

After backtesting the spot variant of the strong signals I got the following results. And boy they indeed look pretty spectacular. The 4 hour timeframe even has a high of 7400 percent profits. And I can not remember any strategy until now that has that same amount of profits.

Nonetheless the 1 day timeframe seems to have the best final score here. Even though it scores lower on the Sortino and Sharpe ratio. The main reason here is the drawdown, pairs ratio and poor win rate on the 4 hour timeframe. If you would use that 4 hour variant. You’ll have to have a lot of patience and faith that these drawdown can be again turned into profits again.

Looking a little bit closer to the entry and exit signals of the 1 day timeframe. We can see that the average profit of the trades is almost 12 percent. And that with only 30% win rate.

Looking at the best performing crypto pairs you see on this timeframe, we see that Trading Shiba Inu in this backtesting period had brought you almost 2000 percent profits. And this counts a lot against the total profit percentage. Matic and Solana and other pairs could have also made you very happy with your initial 1000 dollar investment. You’ll have to remember that these can also be considered as outliers or lucky trades and that these will not be happening in the future.

Also, you also had to have to endure this chart, where from the top of the last bull market your acquired fortune slowly bleeds back into the market again. You’ll have to see your bot doing nothing most of the time as you can see at the parallelism chart. Now for spot trading this actually is not bad, because there are not that many long trades to make in a bearish market. But again, these can be considered opportunity costs if you know that you also can go short during these periods. So let’s find out how the performance would be when the bot also makes short trades during these bear markets.

Futures

So now we maybe we are comming to the pinnacle of this trading strategy. Having strong signals and making use of both the long and short side of the market. I expect that this last variant will perform the best of all, and hopefully I am right about this.

Let’s watch the results of the backtests I did with this variant.

The first thing that is interesting to see is the timeframe. Until now the 1 day timeframe was always having the the highest scores. But this time the 4 hours seems now have the best cards.

Then the amount of profits that this timeframe has. It is quite comparable with a lot of all the other variants of this Kijun crossover strategy. All these profits hover between 1800 and 3800 percent and this one is somewhat in the center. The win percentage and drawdown are also similar from what we have seen before. But what this variant makes scoring like it is, are the ratios. The CAGR, Calmar, Sortino, Sharpe and Pairs ratio score the best of them all. Let’s take a more detailed look and watch the enter and exit reasons.

By looking at these reasons we can see a little bit more in detail what happens with all the trades. Again what strikes me here is that the long signals make all the profits happen and that going short actually makes the losses. There are more short than long trades and their winrate is almost the same as the long trades. I suspect that in this case the algorithm also catched some big pumps of a couple of pairs and therefore has more profits to the long side instead of the short side.

Actually I hoped that this variant also would catch some big crypto dumps as well, because they happened. And I did a Little investigation on what the possible causes are.

So below you see a small snippet of the Bitcoin plot and you can see that there is indeed a sharp drop in may 2022, which is caught with a 16 percent wining short trade. However, Before and after that one trade the price stayed relatively stable and zigzagged along the Kijun sen below the Kumo cloud, which resulted in a couple of losing trades. Now, I saw this behavior a couple of times on this chart and my theory is now that, while crypto pairs can have very steep drops, they are relatively stable after that for some time and hover around already existing support lines. The possible reason behind the ginormous profits can be that prices do not really have resistance lines after they break the highest high and then have price discovery. Which can result in parabolic moves. And with these altcoins the parabolic moves can be sometimes ridiculously high. But again, that is just a theory and that is why I place more importance on the ratios than on the profit percentages.

Getting back to the profit plot, here you can see again closely that some coins have vertical lines to the upside and then have a slight drop off to lower profits. You could say that with the coins that I use for this backtest and that were in the top 50 of january 2022 you better have bought them when they just appeared on the Binance platform to trade, and then hope they have some parabolic moves during bull markets.

Strategy League

So with all these backtests and findings, which variant of the Kijun crossover trading strategy has the best performance and where will they end up in the strategy league?

It might not be a surprise anymore, but the strong signalled Kijun cross on futures and the 4 hour timeframe has the best performance of all these variants.

Again, comparing this variant with the other Kijun crossover ones does not have specatcualr differences on the profit percentage, winrate and drawdown. But it’s score is mostly dependend on the good ratio’s this strategy has.

Comparing this best variant against all the other trading strategies I tested earlier shows that it certainly not has the best winrate, but the drawdown is somewhat comparable with the top 5. Again the CAGR, Calmar, Sortino and Sharpe ratio makes this strategy score pretty high up.

Please take into consideration the fact that pumps are probably the main cause for the profit percentage, but also that these other strategies are tested in almost similar conditions.

Nonetheless, If you combine the best parts of this strategy with the best parts of the other ones on this list, I think you could come up with an algorithm that has certainly potential to make you money in the future.

Files and more

The files of this strategy are located in a Patreon post that was uploaded earlier. You can find the files here: https://www.patreon.com/posts/82552983

There is also another additional Patreon exclusive video on this subject in this post: https://www.patreon.com/posts/82552777

Thank you for visiting!!