Initial price $50, increases by $1 with each sale
AI Knights of the Round is a fully AI-driven trading expert for gold (XAUUSD). It can freely switch between free and paid models, and it supports backtesting AI performance to optimize and adjust input parameters for better results.
As we all know, using AI (meaning large language models) requires sending requests. This is not only very slow, making long backtests nearly impossible, but also consumes tokens. Free models usually come with call limits, which also prevent long backtests. With paid models, tokens equal money: running a single test may be acceptable, but optimization requires tens of thousands of backtests, which becomes completely infeasible.
AI Knights of the Round implements a highly efficient AI output caching technology. It only requests AI during the first backtest and stores the results in cache files. Later backtests and optimizations can reuse the cached files.
🐶 [Recommended Setup]
[Dual Free Models]
Completely free! I mean, the models are free.
- Models:
- deepseek/deepseek-chat-v3.1:free
- x-ai/grok-4-fast:free
- Timeframe: H4
- Deposit: minimum $100
- Symbol: XAUUSD
- Leverage: minimum 1:30
- Broker: any
- Set file: download from the QR Code
[Why are the models free?]
OpenRouter provides each account with 50 free requests per day. 24 ÷ 4 × 2 = 12 < 50, so you can use them completely free of charge.
🤡 [Preparation]
Please register an OpenRouter account and create an API Key, then enter it in the first input parameter of the cBot.
OpenRouter provides 50 free requests per day per account.
If you deposit $10, the free quota increases to 1000 requests per day, but with AI Knights of the Round, no deposit is needed, 50 requests are enough.
💸 [Paid Model Consumption]
Each request consumes about 3742 input tokens and within 50 output tokens (which can be ignored in billing). The number of requests depends on the timeframe. For example, H2 requires 12 requests per day, and with dual paid models, that’s 24 requests per day.
Recently (end of September 2025), the deepseek/deepseek-r1-0528:free model has been somewhat unstable, sometimes returning a “429, Provider returned error.” Users relying on this model may temporarily need to switch to the corresponding paid version: deepseek/deepseek-r1-0528.
For example, with deepseek/deepseek-r1-0528:
Pricing: $0.40/M input tokens, $1.75/M output tokens.
Each request costs about $0.0015. On the H2 timeframe, daily cost is about $0.018.
Overall, the cost of paid models is very low, even smaller than trading commissions. In normal algorithmic trading operations, you don’t need to worry much about this.
🐥 [Configuration Guide]
[Set Package Structure]
cache/yyyy-MM-akotr-ai-records.cache: AI output cache files split by month. These must be placed in the specific folder mentioned below for corresponding backtests to work.*.cbotset: Set files that allow backtesting based on AI output cache, reproducing the performance shown in screenshots.
[Location to Place AI Output Cache Files]
- Windows: ~
\Documents\cAlgo\Data\cBots\AI Knights of the Round\ - Mac:
~/cAlgo/Data/cBots/AI Knights of the Round/
[If you really can’t find this location (rare case)]
Run a backtest of AI Knights of the Round once, then globally search for the akotr-ai-records.cache file. The folder containing it is the correct location.
🧳 [Set File List]
- Set Package: ic-h1-240922-250922-free2-d-g.zip
- Broker: IC Markets
- Cache file timeframe: H1
- Date range: 2024-09-22 ~ 2025-09-22
- Models:
- deepseek/deepseek-chat-v3.1:free
- x-ai/grok-4-fast:free
