CLOSE = you have banked >= 80% of the premium - lock the win and free the shares (the last sliver is not worth the pin/tail risk), OR a still-OTM call whose survival has fallen below 67% (cut it before it breaches). ROLL = already ITM and losing: the LEAPS defends it, so roll up/out (see roll_calls) rather than realize the loss. CHECK REBUILD = still OTM but deteriorating AND underwater: closing would realize the loss, and roll-vs-cut on a fortress is fortress_rebuild's call - the row mirrors rebuild's own routing so you see its answer here. ACT TODAY = expires today ITM: buy back or roll before the close. Everything else HOLDS - a safe call with real premium still decaying to you. The decision is about THIS call's own premium, not whether a fresh CC exists (you find those in Insight 2). Surv colours mark the two levels that actually do something: red below 67% (holding has gone -EV, so the verdict flips), amber up to 80% (safe to hold, but you could not write this strike today), green at/above the 80% write floor.

ONE view (per Abhi, 2026-07-10): sell every available contract at each fortress's income pick - the projection's New-CCs term uses exactly this number. Cell shows monthly / weekly income and contracts. CC-SS is the scalar safe floor; every strike passed fortress_rebuild's per-expiry exit-model safety check.

These fortresses carry Skip; in the CSV Comments column, so they are held out of New CCs, E[rollover] and E[fight] - the projection forecasts income you will actually write, and you have deferred these. They are still fully live everywhere else: realized income, the Base pipeline and every buyback verdict in Insight 1 are unaffected, because those are about calls you ALREADY sold. The goal ($30,000 floor / $50,000 ideal) is deliberately NOT lowered to match - this is an income-for-safety trade and the number above is its price. Clear the tag in the CSV and the income returns on the next run.

Read from fortress_fight --all run Sep 21, 21:50 · 2h ago. THE panel for every uncovered deep-drawdown fortress, at the near-term safe-yield strike you actually sell weekly (survival >= 90%). ✓ safe = strike at/above CC-SS (assignment wouldn't lock a loss); ⚠ below = below CC-SS (assignment locks a loss - the risk you accept for the near-term tenor). This E[net] IS the projection's E[fight] term (prorated): you sell it, it lands in actuals. This cycle = the gross cash you collect on THIS sale (bid × 100 × contracts). Gross/mo is that ×(30/DTE), so a 2-3 DTE rung is a ×10-×15 extrapolation that assumes you re-sell it all month. E[net]/mo = gross × your 68% historical keep rate, the same rate the Goal Tracker net-adjusts with. (It is NOT fight's own vol-edge EV: on 3-DTE penny strikes that number is a difference of two near-equal pennies ×10 proration, so a single IV tick swung it $64k→$22k on an unchanged bid.) Skipped: names whose live book shows an open call (already in Base/E[rollover]). The 45-DTE at/above-CC-SS alternative is FYI in Insight 2c below. Trade plan / escape doors live on the FIGHT dashboard.

One row per open LOT. Old cap = the start of the lot's ROLL CHAIN: buybacks link to the leg that replaced them (re-opened within 3d; an expiry gap breaks the chain), walked back from the most recent buyback - so a progressive escape (435 → 475 → 540) scores its ENTIRE move even when early steps predate the 30d window, which only decides whether the chain still counts as a live escape. DOWN-roll steps net through: a lot whose episode net-LOWERED its cap shows muted as “net down”, outside every total. Lots pair to chains by contract count (the sheet has no account column), so same-size lots share a chain start; unmatched sizes fall back to the ticker's highest recent buyback. A row tagged [roll] in the log links to its predecessor AUTHORITATIVELY (tag every rolled-to leg; fresh writes never), and a tagged chain STOPS at the first untagged leg: the cap that stood before the first tagged roll is where counting starts. Tagged escapes STAY VISIBLE while the leg is open and bypass the up-and-out tenor bar; the 30d recency window and the 21d tenor bar gate only the untagged heuristics; for untagged history an open leg links to a closed one only when the close was a PENNY/zero buyback (a roll: the cost lives in the successor's net premium or negok row) or a real buyback continued DOWNWARD (post-breach de-risk); a real buyback followed by a HIGHER fresh strike is an exit plus a new campaign and never links (BMNR 17.5 exited → fresh 20.5, one day apart), and an escape must go up AND OUT: tenor at entry >= 21d, so a next-weekly re-strike (MU 950→1000, IREN 40.5→45) is the income machine adapting, never an escape. Headroom = (new - old) cap × contracts: LEAPS upside the roll unlocked. Banked by roll = spot minus the OLD cap when spot trades through it: gains the old cap would have handed to the call buyer at expiry. Marked to spot, UNREALIZED - it firms up only as the LEAPS are monetized or spot holds. Puts and down-rolls never count. Never in income totals, projections or --score.

The flip side of the buyback toll: income given up NOW comes back as cap P&L on this schedule. Each up-roll lands in the month its rolled-to call expires; IF spot finishes that expiry through the new cap and the trade is closed (assigned / LEAPS monetized), the month realizes the lot's full headroom over the old cap. Conditional on the rally holding to expiry - a lot that finishes below the new cap realizes nothing here (and keeps the upside unrealized instead). Display-only, like the ledger above.