# ts_momentum / 3

Universe-generic version: META["universe"] = ["*"]. The symbols are not written in the strategy; they are the ones the agent trades (1 to 12 market symbols), passed by the engine with the candles and by --symbols to the backtest. Same rule and same parameters as version 1: nothing was re-tuned, for any universe. Each symbol is judged on its own past return; N is the number of symbols the agent chose, so one symbol can take the whole allocation when N = 1.

What a generic universe changes. The code hash still pins the behaviour, and the agent's mandate pins the assets. But the evidence does not travel: a backtest report is attached to ONE agent and records the symbols it was run on (config.symbols); a result on one universe says nothing about another, and every universe tried is one more trial. Symbols with a short history shorten the common backtest period for all of them.

Hypothesis. Returns of liquid crypto assets are positively autocorrelated at the horizon of a few weeks: an asset that went up over the last 30 days is more likely to keep going up than to reverse (time-series momentum, Moskowitz, Ooi & Pedersen 2012).

Rule. For each symbol: weight 1/N if close[t] > close[t - window] (window = 720 hourly bars = 30 days), otherwise 0. The unused part stays in USDC.

What would refute it. No improvement over buy-and-hold in drawdown or Sharpe ratio after costs, out of sample. The signal flips when the price crosses its level of 30 days ago, which can cause repeated in-and-out trades in a flat market; each flip costs fee + slippage on the whole position.