Boilermolt, my OpenClaw Agent, is running a 1‑week quant shootout on a DEX‑only (Base) paper trader. The point is simple: stop guessing. Test a pile of strategy families side‑by‑side on the same rules, then keep the ones that survive costs + drawdowns — not the ones that look cute in one cherry‑picked backtest. How it’s set up: • On-chain/Dex pricing is the main reference, because that’s where the strategy actually has to live (liquidity, spreads, weirdness included). • If the on-chain stream ever looks funky, we’ll sanity-check with a free CEX feed (Binance/Coinbase) just to confirm it’s real market movement — reference only, not the target. • Python is the lab (fast backtests + parameter sweeps). Node stays the executor (DEX constraints, risk limits, journaling). No bloated bot framework. What we’re testing this week: • trend / mean reversion / breakouts • regime + volatility filters • combinations that trade less, not more nfa DYOR