chore(polymarket-trader): widen max_edge 0.15 -> 0.30 (volume trial)

[jmoreira-valory]

What

Widen the Polymarket trader's edge cap max_edge from 0.15 -> 0.30 in polymarket_trader/service.yaml (STRATEGIES_KWARGS). Omen (trader_pearl) is untouched. No FSM change; only the polymarket_trader service CID re-locks.

Why

At max_edge=0.15 the live edge gate admits ~0 bets/day in the current market regime -- the forecaster has drifted more overconfident over time (median execution edge ~0.18 -> ~0.36 since January; the share of bets landing in [0.01, 0.15] fell ~31% -> ~10%). The trader is effectively silent. Widening to 0.30 roughly doubles admitted volume so the service actually trades.

Tradeoff (please read before merging)

This is a volume-over-ROI trial, not a profit change. The max_edge ROI sweep shows widening is not ROI-positive in the current regime -- the extra bets it admits are net-negative:

max_edge	last-30d capital-weighted ROI
0.15	~-4.8% (least-bad)
0.30	~-12.9%

So 0.30 buys activity at a documented ROI cost. It is a defensive, monitored trial to keep the trader active while the durable fix -- upstream forecaster calibration (a high p_side must actually be right) -- is developed. A downstream edge cap cannot fix an overconfident forecaster.

Rollback

• Operated deployments: STRATEGIES_KWARGS env flip (immediate).
• Pearl: a follow-up release (reaches all users; no per-agent switch).

Monitoring

Track capital-weighted ROI / bets-day / win-rate vs the ~-5.4% pooled baseline weekly; revert if the current-regime ROI degradation outweighs the volume benefit.

🤖 Generated with Claude Code