I Built an INSANE AI Trading Bot With GPT-6 + TradingView (Full Guide)

Kurze Zusammenfassung

In diesem Video wird gezeigt, wie man mit GPT-6 einen AI Trading Bot erstellt. Das Ganze umfasst die Installation benötigter Tools, die Automatisierung des Handelns sowie die Auswertung der Handelsstrategien. Die Schaffung profitabler Strategien mithilfe von GPT-6 wird eingehend beleuchtet, inklusive Testergebnissen zu Rück- und Vorabprüfungen.

  • Erstellung eines AI Trading Bots mit GPT-6
  • Nutzung von TradingView zur Automatisierung der Handelsstrategien
  • Auswertung der Strategien und Vergleich von Ergebnissen

GPT-6 erstellt einen AI Trading Bot

Im ersten Kapitel wird erklärt, dass GPT-6 in weniger als 8 Stunden einen AI Trading Bot entwickelt hat. Der Ziel ist es herauszufinden, ob dieser Bot profitabler als bestehende Modelle wie Claude oder Grogbot ist. Der Autor kündigt an, dass er eine Schritt-für-Schritt-Anleitung zur Verfügung stellen wird, um den Bot selbst erstellen zu können. Dazu gehört die Verbindung aller benötigten Tools, um live Handelsstrategien zu testen und die Resultate im Trading Arena bekannt zu geben.

Der Plan: Von der Einrichtung zur Automatisierung

In diesem Kapitel wird der Plan vorgestellt, der die notwendigen Schritte umfasst, um den Trading Bot zu erstellen. Der Autor beabsichtigt, zuerst alle erforderlichen Tools zu installieren, wobei er kostenlose Optionen priorisiert. Danach sollen diese Tools miteinander verbunden werden, um Handelsstrategien zu entwickeln und zurückzuprobieren, bevor der Bot live handeln kann. Hinterher wird GPT-6 für 8 Stunden mit der Suche nach optimierten Handelsstrategien beauftragt.

Benötigte Tools: ChatGPT, TradingKit & TradingView

Hier werden die Tools vorgestellt, die für die Erstellung des AI Trading Bots notwendig sind. Der Autor erwähnt ChatGPTs Desktop-Version und TradingKit, ein kostenloses Tool, das die Entwicklung und das Testen von Handelsstrategien mit Pinescript ermöglicht. Außerdem ist ein TradingView-Konto erforderlich, um alle Tools zu verknüpfen. Er diskutiert die Funktionen der jeweiligen Tools und ihre Relevanz für den Entwicklungsprozess des Bots.

Installation von Trader.dev und Testen der Verbindung

In diesem Kapitel demonstriert der Autor, wie man Trader.dev installiert und die Verbindung zu ChatGPT herstellt. Nachdem sich der Nutzer bei Trader.dev angemeldet hat, wird ein Aktivierungscode benötigt, um ChatGPT den Zugriff auf die Backtesting-Funktion zu ermöglichen. Der Autor führt durch die Schritte zur Überprüfung der Verbindung und testet eine einfache Handelsstrategie, um sicherzustellen, dass alles korrekt eingerichtet ist.

Die Eingaben & Acht-Stunden-Forschungs-Runde

Der Autor erklärt, wie man GPT-6 anweist, eine Handelsforschungsstation einzurichten, die Strategien zurücktestet und optimiert. Dafür wird dem Bot eine achtstündige Zeitspanne eingeräumt. Der Autor reflektiert darüber, dass früher manuelles Backtesting zeitraubend war, und hebt die Fortschritte hervor, die durch Automatisierung erzielt wurden. Die spezifischen Eingaben für GPT-6 sind in den Ressourcen des Videos verfügbar.

GPT-6 hört frühzeitig auf: Neustart des Experiments

Leider hat GPT-6 nach einigen Stunden beschlossen, die Arbeit einzustellen, und der Autor musste einen neuen Prompt erstellen, um den Bot fortfahren zu lassen. Trotz dieser Unterbrechung zeigt er sich optimistisch und berichtet von ersten Einblicken in vielversprechende Handelsstrategien, die der Bot während der laufenden Tests entwickelt hat. Er plant, die Tests für weitere fünf Stunden fortzusetzen.

Ergebnisse der Strategie & Missionskontrolle

Nach der Fertigstellung der acht Stunden zeigt der Autor beeindruckende Ergebnisse. GPT-6 hat 184 Handelsstrategien generiert, während der vorherige Bot, Fable 5.1, nur 13 Strategien produziert hat. Die Backtesting-Ergebnisse sind überwiegend positiv, zeigen jedoch eine Präferenz für die 1-Stunden-Zeitrahmen. Der Autor betont, dass die Effizienz des GPT-6-Modells beispiellos ist und hebt die niedrige Token-Nutzung im Vergleich zu anderen Modellen hervor.

Vorwärtstests & Strategien Inkubation

In diesem Kapitel erklärt der Autor, wie erfolgreiche Strategien in einer Vorwärtstestphase verbleiben können, bevor sie auf den Live-Markt gebracht werden. Die Strategien werden für einige Monate in einer simulierten Umgebung getestet, um zu prüfen, ob sie in unbekannten Marktbedingungen weiterhin funktionieren. Der Autor betont, dass es wichtig ist, die Backtesting-Resultate mit jenen aus der Vorwärtstestphase zu vergleichen.

Hinzufügen des Pine Scripts zu TradingView

Hier wird erklärt, wie das erstellte Pine Script in TradingView importiert werden kann. Der Autor zeigt den Prozess des Kopierens und Einfügens des Scripts in den Pine Script-Editor von TradingView. Nach dem Speichern wird gezeigt, wie der Benutzer die Handelsstrategien auf den Charts sichtbar macht und zusätzlich Backtests durchführt, um die Ergebnisse zu validieren. Besondere Aufmerksamkeit wird dem maximalen Drawdown gewidmet und der Notwendigkeit, Regeln zur Risikominderung zu implementieren.

Renditen, Drawdown & Buy-and-Hold-Vergleich

Der Autor beleuchtet die beeindruckenden Renditen, die mit der entwickelten Handelsstrategie erzielt werden können, und vergleicht sie mit der traditionellen Buy-and-Hold-Strategie. Er hebt hervor, dass die handgemachte Strategie eine Rendite von 3.130% erzeugt hat, während Buy-and-Hold fast keinen Gewinn erbracht hat. Die P&L-Daten sowie Anzahl der abgeschlossenen Trades werden ebenfalls analysiert und als sehr positiv beschrieben.

Verbinden Sie Ihren Broker mit Trigger.trade

In diesem Kapitel wird erläutert, wie der autorisierte Handel zwischen TradingView und einem Broker durch das Setzen von Alerts in TradingView und die Verwendung von Trigger.trade als Mittelmann eingerichtet wird. Der Autor zeigt, wie man den API-Schlüssel und andere notwendige Informationen einfügt, um den Trading-Bot über die API des Brokers Handelstransaktionen ausführen zu lassen. Dies spart Zeit und ermöglicht automatisiertes Trading ohne ständige Überwachung des Marktes.

TradingView-Benachrichtigungen & Webhook-Setup

Der Autor zeigt, wie man Alerts in TradingView einrichtet, um den Bot über Trigger.trade mit dem Broker zu verbinden. Er beschreibt die Schritte zur Erstellung der Alerts, einschließlich der Eingabe von Webhook-URLs und der Anpassung von Benachrichtigungsdetails. Dies ermöglicht die automatisierte Kommunikation zwischen TradingView und dem Broker für eine reibungslose Trade-Ausführung.

Echte Gespräche: Diversifizierung & Risiko

Der Autor diskutiert die Bedeutung von Diversifizierung im Trading, basierend auf seinen Erfahrungen in den letzten fünf Jahren. Er rät dazu, nicht das gesamte Kapital in eine einzige Strategie zu investieren, um das Risiko zu streuen. Er betont die Wichtigkeit, nur Geld zu investieren, das man sich leisten kann zu verlieren, um Stress und unnötige Risiken zu vermeiden und empfiehlt, mehrere Bots zu nutzen, um potenzielle Verluste abzufedern.

GPT-6 vs. Fable & Grok: Die Arena-Rangliste

Hier wird die aktuelle Leistung von GPT-6 im Wettbewerb mit anderen AI-Modellen wie Fable und Grog vorgestellt. Der Autor zeigt die Rangliste der Ergebnisse, wobei GPT-6 derzeit den zweiten Platz belegt. Die Ergebnisse umfassen Aspekte wie Geschwindigkeit, nachgewiesene Effizienz und Konsistenz. Er erwähnt, dass GPT-6 kostengünstiger als die Konkurrenz ist, weist aber darauf hin, dass es sich um einen laufenden Test handelt.

Erkunden Sie die Arena und treten Sie der Community bei

Im letzten Kapitel lädt der Autor die Zuschauer ein, die aitradingarena.com zu besuchen, um mehr über die Handelsstrategien zu erfahren, die von KI-Agenten entwickelt werden. Er erwähnt auch seine eigene Schulungsplattform, die über 400 getestete Handelsstrategien anbietet und ermutigt die Zuschauer, an seinen wöchentlichen Livestreams teilzunehmen, wo er weitere Einblicke in das Automatisieren profitabler Strategien gibt.

Fasse jedes YouTube-Video in Sekunden zusammen

Hol dir Zusammenfassungen wie diese für jedes Video, in deiner Sprache, mit der kostenlosen Summ-App.

Laden im App Store
Jetzt bei Google Play

Transkript

Vollständiges Transkript von „I Built an INSANE AI Trading Bot With GPT-6 + TradingView (Full Guide)“. Klicke auf einen Zeitstempel, um das Video an dieser Stelle zu öffnen. Du brauchst ein anderes Video? Nutze den YouTube-Transkript-Generator.

0:00 This insane AI trading bot was built by GPT6 in less than 8 hours. And by the end of this video, I'm going to show you step by step how you can build your own using GPT6 and Trading View. But building another bot is great. The big question is, can it be more profitable than Claude or Grockbot? To find out, I'm going to enter GPT6 into my AI trading arena. By the end of this video, you will get all of the tools, the prompts, the step-by-step guides to allow GPT6 to research, build back test strategies, and also actually live trade. And then I will reveal the results of the trading arena. But for that to happen, we're going to have to start right now. Okay. So, here is

0:45 today's road map. We're going to be first of all installing everything we need. I'm going to be using tools that are free where possible and where not. You most probably already have a subscription to those tools. Once everything is installed, we're going to connect them all together so that they can actually back test and create strategies before we end trade anything live. Then we're going to give GPT6 8 hours to go away and find the most profitable trading strategy possible. And then we will put GPT6 in the leaderboard to find out whether or not it is profitable. And then in the next step, I'm going to show you how you can take the trading view strategies and connect them over to your broker. And then finally, I'll give you my thoughts as an algo trader and AO trader over the past 5 years. I will tell you what I

1:33 think of the strategies and how I would go forward. So GPT6 Astra was recently launched and the things the benchmarks are absolutely insane with ARGI 3 being maxed out at 98.6% and most of the benchmarks absolutely demolished by GPT6. More importantly, the things that people are building with this model are absolutely insane. And from my experience, GPT6 is a lot cheaper than Claude Fable.1. So, but on this channel, we don't go through all these benchmarks and things like that. The only thing that is important to us is can this thing be more profitable than Claude or Grobot. So, let's go over

2:18 to the road map and start installing all of this. First of all, we're going to be installing Chad GPT's desktop version. I'm going to be using a Mac. So, I'm going to click here, download it. As you can see, I already have it here on my screen ready to go. But the problem with Chat GPT is it doesn't have the tools installed to be able to build a trading bot and back test strategies to give us something profitable. So, the next thing that we're going to need to use is something called tradingkit.com. It is a free tool and allows you to practice trading strategies using Pinescript, which is the scripts that are used for trading view. It has an absolutely huge community where everybody's building together trading strategies and they have done almost 1 million back testists

3:04 is very simple to install and I'll show you how to do that in a couple of seconds. And finally, we will need a trading view account to link everything together so that we can actually get these trading bots actually trading and verify our back testing results. But let's go over to chat GPT desktop. As I said earlier, I already have it installed. We're on GPT6, Astra, High Thinking, and we're over in the work tab up here. Now, you can choose between GPT and Codeex, but I've recently been working a lot on the work tab up here, as I feel as though the sessions actually last longer and are just a lot simpler to set together. Chat GPT work is kind of like having the best of both worlds. It gives you the full model of chat GPT as if it is running in your web

3:51 browser, but it also gives you access to your computer and allows you to run things like schedules, which is absolutely perfect for back testing. So, once we have that running, we're going to go over to traderdev.com and we're going to sign in. Once again, this is absolutely free. Once we're logged in, we're going to click on my account up the top here and then we're just going to copy this code here. Go back over to ChatGpt, paste it in there, and say, "Hey there, buddy. Could you please install this MCP server globally? I want to be able to use it in chat GPT work." And then you'd hit enter. I've already got it installed, so I'm not going to install it again. It will take you through the steps for installing this. One step that you will need to know is it will ask you for your

4:37 activation key. Just copy and paste this activation key over to Chat GPT. paste it, hit enter, and it will install everything for you. Once that's done, we're just going to check the connection between tradingkit.com. Hey there, buddy. Could you bank test a very simple trading strategy on BTCUSDT for me? And this is going to go away and back test a simple strategy and make sure that we are fully connected. And boom, there you have it. We can confirm the connection the back test of a simple moving average strategy on 20 and 50 where the AC is below the cross. Here are the back tested results. There is a net return of 4.74% a win rate of 38% a profit factor of 1.5

5:24 and a maximum draw down of 5.81. So it absolutely did work. Even added commission and things like that. So we now know that GPT6 is connected over to tradingkit.com. Okay, that's fantastic. So, we now have our AI connected over to a back testing engine, which allows Chat GPT to back test any trading strategies. The next step is very simple. We're going to give it a prompt asking it to set up a loop. A loop is like a schedule where it will do exactly the same thing over and over again every hour. In our prompt, we're going to ask GPT 6 to set up an AI trading research desk where it's going to back test and find strategies, refine them, optimize them, and find the best results. And we're

6:11 just going to literally give it 8 hours. I'm not going to sit in front of screens. It's going to do all of the hard work for us. Now, if you would have told me that 3 4 years ago where I was back testing manually trading strategies every single day, I'd be absolutely mind blown. This can be a very complex process which used to cost thousands of dollars per strategy and now is easily accessible by anybody. Okay, so I'm going to paste in the prompt. If you want the prompt, you'll be able to find it in the resources down in the description. In this prompt, it's going to onboard chat GPT, explaining exactly what its job is and how it should set up its first trading bot. I'm going to hit enter. It's going to go away and on board. It'll cap take a couple of seconds. And this is a series of two

6:58 prompts that we're going to give it. After it's on boarded, we're going to ask it to set up that 8hour loop. Okay. Once this is done, it's going to build you an AI trading arena, very much like the one that you can see here, where you can see all of the agents fighting against each other and the equity curves. So, we're going to tell CHJBT, which is a little bit more strict, and asks us to accept. And then we're going to ask it to review and execute on the second prompt. And boom. Right. Okay. Now that's going to go away and build our trading bots one by one back testing and trying to find the best settings so that we can actually trade them live on a live broker. It's important to understand exactly what each and every one of these bots are actually doing. That is why we're going to be back testing before we ever put these onto

7:44 live markets. So, we're going to set the clock for 8 hours and I will be back in a couple of seconds as we're going to fast forward through all of that and leave it to actually back test those strategies. Okay, it has been a couple of hours and we've come across a bit of a problem. Unfortunately, chat GPT decided that after a couple of hours that it no longer needed to go any further. So, I did have to create a new prompt that asked it to continue. honestly had a bit of a peek at some of the strategies that it's building and I have never seen anything quite like this. We have multiple strategies already and a lot of them look very very profitable. So, I'm going to leave it for another 5 hours and 16 minutes. Let's hope this time they all go all the

8:30 way till the end. 3 2 1 Boom. Okay, guys. So, Chat GPT has had over 8 hours to actually find strategies. I can tell you I had a bit of a sneak peek and my socks were completely blown off. We're going to go over to chat GPT and unfortunately because it's in a loop, you can't see very much what's happening. But let's go over to our mission control where we'll be able to see the insane results that Chat GPT has made for us. So in 8 hours, Chip has made 184 strategies in compared to the Fable 5.1 sprint which only created 13 strategies. That is absolutely insane for such a short amount of time. And when you look at the back testing

9:16 results, they are pretty pretty awesome. We have different kinds of pair, different strategies, different equity curves. But one thing that I did actually notice is it they are all on the 1 hour time frame. Is this because Chibbit found it easier or it just found its edge only on that time frame? So if we go through we can see some of these insane equity curves. Some of them are extremely flat. A few are not profitable at all. And as we go along they get more and more profitable. So for me this has been the best model that I've ever tested building trading bots. 185 strategies in less than eight hours means that we could build thousands of strategies every single week. Now, the other thing that was very very important

10:03 was that chat GPT used a very small number of tokens per strategy in comparison to Fable 5, Claude, Opus 5, and Grog. Now, when I say per strategy, that is per sprint. I've had going for multiple days in a row, but I constantly get hit by the usage limit. But JBT, as you can see, I have over 96% left, and that was with back testing way more strategies than any other model. Okay, so I said I was going to show you the full process to get these all the way to an automated trading bot. Once you have a strategy, there are a couple of steps to make sure that these strategies actually work. One of them is to put them into an incubation period. Incubation means that you just leave the

10:49 strategy in a simulated environment like a paper trading account. Now, what is absolutely fantastic about the mission control that you actually built with the prompts that I gave you, these strategies go into forward testing. So green is back testing and every single day all of the profitable strategies will be back tested and will continue that equity curve in pink. So next step which is long unfortunately is leaving those strategies for a couple of months to see whether they actually react to conditions that they have never seen in the same way as they did in the back test. Once that's done, we're going to take those strategies, put them onto Trading View, and then we're going to automate those strategies. Okay? Once

11:34 you've put them into a forward testing period, we're going to have to use our imagination here. They will build an equity curve in pink and you're going to have to compare the back testing results to the forward testing results. We're going to click on that and you're going to be able to download the Pine script by clicking the download. You should have this on your mission control as well. that will download a file which we'll take and import into trading view. To import this into trading view, we just literally take the file that's downloaded. We're going to open the pine script in a text editor and copy and paste it here onto Trading View. To find that, we just click on this little pine tree icon just here, which will open this editor. We're going to save it, and we're going to click add to chart. This will add the trading strategy with all

12:21 of the stop- losses, take-profits all on your chart. You'll be able to see them, every single one of them. And what is amazing about Trading View is that we can actually back test those strategies. Now, although we did already back test these strategies, it just gives us another level of confirmation that the back test is exactly the same as the forward testing. Now, there's one thing that actually jumps out to me here is the max draw down is quite high. We should actually set some rules for our AI trading bots to make sure that we don't have strategies that have a such a high max draw down. The P&L looks great. The number of trades looks absolutely amazing at 695 closed trades. And the profit factor looks pretty good at 1.35. This strategy absolutely destroyed the

13:07 buy and hold. Buy and hold would have almost made you nothing. And this strategy made 3,130%. Okay, so you've validated that strategy by forward testing it. You've validated it on Trading View to make sure that the back test is exactly the same as what we're expecting. Now, we're going to need to use the Trading View alert system to connect them over to your broker. Now, this one here was on MNTUSDT, so I will need to find a broker that allows me to trade crypto. And we're going to need a middleman service. A minimum service sits between Trading View and your broker and is a bit like a translator that translates the code from Trading View over to your broker, setting the stop- loss, take profit, and everything in your broker as soon as you

13:52 get a trade. I'm going to be using Trigger Trade. It is absolutely free of charge. You'll have to set up your broker in here. I'm going to do it very quickly as an example. And once that's done, you will have some code that you have to take over to Trading View so that it will make the link between your Trading View account. Now, obviously, you're going to have to set up your broker. It will take a couple of minutes. You just have to find your API key and secrets by clicking on the avatar up the top where you'll be able to find API key and secrets. These are a bit like a secret word and a secret password that allows your trading bot to actually take those trades. Now, one thing that I like about Trigger Trade was it was built by my community and none of those API keys and secrets are actually stored on any of our

14:37 servers. So, first thing we need to go do is go back over to Trading View. Click up the three buttons just here. Click on add alert. We're going to go down and click fill orders only. We need to click on this message just here and go back to trigger trade. Press copy. Paste it in just here. Give it a name. Press apply. Finally, we need to change this web hook URL just here. Copy, paste. Mine was already good. Click apply and Bob's your uncle. Your trades will be sent over to your broker. As soon as Trading View opens a brand new trade, the entry, the stop- loss, and the take-profit will all be sent over to your broker. So, you don't have to sit in front of screens all day every day trading. you have a fully back tested

15:23 and forward tested strategy which allows you to put a little bit of trust into your actual back testing results. Okay guys, before we go just a couple of things, some real talk here. In this video, we learned a little bit about forward testing. Forward testing is extremely extremely important. But once you take it out of forward testing, I would not go and risk the whole of your account on one single strategy. I've back tested thousands of trading strategies in the past 5 years and I currently trade seven bots. This allows for me to diversify my portfolio, iron out that equity curve a little bit, hoping when one strategy is taking a couple of losses that maybe one or two other ones have your back and are actually in a winning streak. And also

16:09 always with trading, don't trade with money that you can't afford to lose because that is just going to stress you and you're just going to go from one bot to another bot to another bot and that is worse than manual trading. Okay guys, one last thing before we go. Let's go over and have a look at the leaderboard. This leaderboard is actually judged by AI and not myself. Is judging all of the strategies that have actually been built by each and every single agent. Now, this is an ongoing experiment, which means that the leaderboard is updated every single day, and these agents are all running building strategies in the background. So, as you can see, GPT 6 is in second place, not far behind Claude Fable 5.1 for the overall score. Build

16:56 speed, actually, Grock was above GPT6. Proven Edge, GPT is in third place. velocity GPT is in third place, consistency in second place, and then we have risk where GPT is actually in sixth place. Turned out to be a lot cheaper than the other models, but as I said, this is an ongoing test. These other models have had weeks and weeks of back testing results as CH GPT has only had one single sprint of 8 hours. On the right hand side here, we also have the leaderboard of forward tested results, which means that if the strategies continue to be more profitable, the AI models will gain even more points and more credibility. This website is called aitrading arena.com. It's free. You can

17:44 come and have a look. And you'll also find a link down in the description to my school where I will give you access to over 400 back tested and forward tested trading strategies. And you're more than welcome to join me in the live streams that I do weekly where I teach everything I learn about building automating those profitable strategies. Okay guys, without further ado, I'm going to wish you a wonderful day. Take care, keep safe, trade like a boss, and I'll see you in the next one. Goodbye. Come on, David. And once again, he's gone to code. A new strategy. He'll be back in just five minutes. RSI and MACD. If he doesn't run into any bugs, there will be a new strategy.

Share

Weitere Video-Zusammenfassungen