As I mentioned in the sponsored project thread I had a discussion with Glassnode to redistribute their metrics. We only talked about the possibility of redistributing daily data for about 330 metrics (their T1+T2 metrics).
You can do rudimentary backtests in their website for $50/mo. On P123 we would make it available for ranking and AI factors, with macro/market data, so a lot more powerful, and would be $100/mo.
Below is an example of the backtesting capability they currently offer.
First you define the factors . These are binary , either on or off.
• Machine learning effectively predicts the cross-section of cryptocurrency returns.
• Simple models prevail; key predictors are price, past alpha, illiquidity, and momentum.
• Contrary to stocks, abnormal returns come from long trades and persist over time.
• ML strategies prove profitable despite high turnover and transaction costs.
• Alphas originate mainly from small, illiquid, and volatile cryptocurrencies.
They used only three on-chain/fundamental features.
I performed (rather basic) backtest using on-chain data from Coin Metrics including data such as rank of network value to transactions (NVT) ratio and active addresses for 5-6 top crypto, However, it turned out that these two factors are not predictive.
