Skip to content
Trading bot
Trading Modes
ChatGPT trading

ChatGPT Trading

With OctoBot, trading according to artificial intelligences such as ChatGPT is possible using both DCA trading mode and Daily trading mode.

In this video, we are showing how one of the ChatGPT trading stragies of OctoBot cloud works.

In OctoBot, ChatGPT is used as a trading indicator and not a way to create orders. This means that multiple trading modes can use ChatGPT.

Using ChatGPT in your trading modes

When using the built-in DCA trading mode and Daily trading mode, you can choose to enable the GPTEvaluator. When doing so, your trading modes will take the ChatGPT prediction into account when trading.

ai trading illustrated by octobot head with chatgpt logo trading bitcoin ethereum litecoin usd logos

You can therefore choose to trade solely on your ChatGPT predictions by enabling the ChatGPT evaluator only or to consolidate them with other evaluators.

The way the ChatGPT evaluator works

The ChatGPT evaluator works in 3 steps:

  1. Gathering of market data: According to your ChatGPT evaluator configuration, market data are processed to be sent to ChatGPT and ask for a prediction
  2. Asking for the prediction: OctoBot asks ChatGPT for a market prediction based on your given data. This prediction consists in asking ChatGPT if the market is more likely to go UP or DOWN in the near future with a % of confidence.
  3. Analysing the ChatGPT prediction: According to the ChatGPT answer, the ChatGPT evaluator emits an evaluation. The side of the evaluation depends on the UP or DOWN answer and its weight is based on the confidence of ChatGPT in its prediction.

Configuring the ChatGPT evaluator

The ChatGPT evaluator can be configured in many ways to customize the way you want ChatGPT to make predictions.

trading with chatgpt in octobot GPTEvaluator configuration

  • Indicator defines the way you wish to preprocess market data before sending them to ChatGPT. You can choose to send the raw candles (no preprocessing), a moving average or other type of processed values.
  • Source is the input to give to the Indicator when one is selected
  • Period is the period setting of the Indicator when one is selected
  • Minimum confidence threshold is a value in % starting from which to send 1 or -1 instead of a value between -1 and 1. This setting is especially useful when using the DCA trading mode in Evaluators based DCA which requires a 1 or -1 evaluation
  • GPT Model allows you to select the OpenAI GPT model to use for your evaluations. Note that models might have a different pricing.
  • Allow Reevaluation is used when using the ChatGPT evaluator alongside other evaluators such as real time evaluators that might require re-evaluations. As re-evaluations can happen at a high frequency, disabling re-evaluations is a safeguard to avoid using too many OpenAI tokens in case this happens
  • OpenAI token limit is the maximum amount of OpenAI token that can be spent within a day. You can use this setting to set a limit of the token use if you are unsure about the total cost of your strategy.

When using the open source OctoBot, the ChatGPT evaluator requires the ChatGPT interface to be configured to be able to operate.

Backtesting with the ChatGPT evaluator

It is not really possible to efficiently ask ChatGPT for the thousands of predictions required when running a backtesting. It would take hours and cost a huge amount of OpenAI tokens.

However, at OctoBot cloud, we decided to pay this cost for you on trading pairs that are used in OctoBot cloud strategies. This means that backtesting with the ChatGPT evaluator is possible on trading pairs used by OctoBot cloud strategies. When selecting such pairs, your OctoBot will automatically fetch ChatGPT historical predictions and run your backtesting accordingly.

Going deeper in ChatGPT trading

Learn more on ChatGPT trading strategies on our dedicated blog articles: