Sina, co-founder of hedge fund 21st Capital and data scientist, publicly challenged a widely cited Bitcoin price prediction model that correlates Bitcoin prices with Global M2 money supply data. The model, promoted by Real Vision CEO Raoul Pal, shifts Global M2 figures forward by 10 to 12 weeks to forecast Bitcoin’s future price moves.
According to Sina, this approach suffers from serious data misinterpretation and overfitting. In a June 24 video posted on X, he described the model as failing basic data analytics principles, saying, “This is a terrible failure of not understanding overfitting.” He explained that manipulating data to align with historical Bitcoin price patterns creates an illusion of predictive power, but lacks a robust foundation.
Sina noted fundamental issues with Global M2 data, which aggregates money supply figures from multiple central banks using exchange rates. This mix of timely and delayed data from various countries generates misleading daily fluctuations, undermining its reliability as a signal.
He demonstrated that minor variations in lead times, such as shifting the data lead by 80 or 108 days, drastically alter the correlation between Bitcoin and Global M2, exposing the model’s instability. “This is not modeling. This is playing,” he remarked.
Sina further illustrated overfitting by comparing it to fitting a curve to a noisy sine wave, where an overfit model captures random noise instead of the underlying pattern, leading to poor predictions on new data.
Challenging the model’s premise, Sina suggested Bitcoin may actually lead liquidity rather than follow it, citing that Bitcoin peaked 145 days before liquidity in the last market cycle. This contradicts the causality implied by the Global M2-based model.
His conclusion warned analysts and investors to be cautious of models that fit historical data without demonstrating true predictive accuracy.
At the time of reporting, Bitcoin was trading near $106,952.