A very interesting test here, the same exact settings were run with Fable 5 against our Gold Standard which is run with Grok 4.3. We've tested just about every model in trading and Grok 4.3 always performs best, but we haven't tested Fable until now. The results: Grok 4.3 - 131.89% profit over a 6 week period, this is the mean of 5 identical backtests, so a very robust result. Fable 5 - 19% profit, same settings same timeframe... exact same test. However, that doesn't tell the full story. Using @Shekel_Agentic MCP tools we drilled down into the Fable test and discovered that it was actually more directionally correct than Grok, also it had a 79% win rate, which is fantastic. Fable however was very cautious with size and put lots of thought managing exits, he was banking profits while Grok was YOLOing the $VVV parabola. In any case, the results are very interesting and worth some more testing with Fable to see if we can craft some adjustments to the prompt to improve his sizing and conviction. If so, the results could be explosive, definitely a thread worth pulling, and all made possible with Shekel's tools.

Shekel was one of the first projects to integrate @AskVenice at the protocol level over a year ago. We saw the value in their business model (staking $VVV, now $DIEM for daily inference power) and built a trading framework on it. Now we're overhauling our business model and $SHEKEL utility based on the same model. It's not completely the same (we don't need 2 tokens) but certainly inspired by it.

Been running several agentic strategies through the @Shekel_Agentic backtesting engine, so far most retail strategies are terrible: momentum (ema200, ema50), mean reversion, RSI focused, etc... I tried lots of them. I finally found a good one, crafted with my @rei_labs agent, based on liquidation hunting with some secret sauce. Been refining and testing it, this is the latest 6 month backtest at fixed 3.5x leverage, trading BTC and SOL, 60 trades. Promising. Now I take the test report back to my REI agent, we look at what worked and what didn't, and refine the strategy. Adjust params, token whitelist, leverage, etc... and run the test again. The shekel backtest engine uses the same data (historical), llm, execute calls, prompting method, virtually everything that is used by your live trading agent, in order to simulate the most accurate results. This test took about 15 hours to run.

The new Meta is: 'Can trading agents be profitable?' @Shekel_Agentic has been trying to tackle this for over a year, and the honest answer is: not quite yet, but we are closer now than ever, and at the rate ai tech is improving it will not be long before you see agents outperforming the market consistently and significantly. This is why all of a sudden new participants are flocking to agentic trading, even @virtuals_io launching the Arena... they can tell the breakthrough is near.