sizing discipline binds per market: adds capped at the current stake size

Proportional add-mirroring was unbounded after max_position_usd was lifted -
fortuneking doubling into his own bet took one game to $90 on a $42-stake
book (2.15 stakes). Adds now only top a position up to stake_usd(); the
backtest was already one-market-one-stake, so bot and model agree again.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

README.md

@@ -144,10 +144,14 @@ backtest and bot share:
 - **Lag + slippage**: the bot fills at the live CLOB ask at detection (~60s poll),
   logging per-fill `detect_lag_s` and `slippage_pct`; the backtest applies a
   +0.5%/~90s haircut. Measured so far: ~48s avg / +0.8% avg slip.
-- **Dynamic sizing**: each bet stakes **4% of current equity** (compounds both
-  ways), halved while equity is below 80% of its high-water mark. No per-trade
-  cap. A per-event correlation cap exists (`risk.max_per_event`) but is **off** —
-  every conviction trade is followed.
+- **Dynamic sizing**: each bet stakes **4% of current working equity** (compounds
+  both ways), halved while equity is below 80% of its high-water mark. The rule
+  binds **per market** — adds that mirror a sharp scaling in can only top a
+  position up to the current stake size, never past it.
+- **Profit ratchet** (`stake_cap_usd: 250`): stakes pin at $250; once the book
+  outgrows that level, surplus cash sweeps to a **banked reserve** that never
+  bets — locked-in profit, and fills stay inside realistic book depth. A
+  per-event correlation cap exists (`risk.max_per_event`) but is **off**.
 - **Entry cap 0.95**: entries above 95¢ are skipped (`follow.max_entry`) — the
   June sweep showed >95¢ favorites *lower* final equity even while winning
   (slip + fee eat the 1–3% payouts; the capital compounds better elsewhere).

archive/copytrade.py

@@ -427,11 +427,21 @@ class CopyTrader:
             return
 
         if is_add:
-            # proportional add: grow my position by the same fraction they did
+            # proportional add: grow my position by the same fraction they did —
+            # but SIZING DISCIPLINE binds per market: total position cost never
+            # exceeds the current stake rule. Unbounded mirroring once took one
+            # game to 2.15 stakes ($90 on a $42-stake book) when fortuneking
+            # doubled into his own bet; the backtest is one-market-one-stake,
+            # so the bot must be too.
             their_prev = self.state["their_pos"].get(wallet, {}).get(token, 0)
             frac = their_size / their_prev if their_prev > 0 else 0
             want_shares = mine["shares"] * frac
-            want_usd = want_shares * price
+            room = self.stake_usd() - mine["cost"]
+            if room < self.risk["min_order_usd"]:
+                self.log(f"ADD  {label} — skip (position ${mine['cost']:.0f} already "
+                         f"at the stake size)")
+                return
+            want_usd = min(want_shares * price, room)
             kind = "ADD "
         else:
             want_usd = self.stake_usd()      # fraction of current equity (compounds)