H1 Total exposure
All positions together stay at most 200% of the portfolio.
Show the numbers 1 value
- Gross exposure, at most
- 200%
Rules
The rules the council works inside
Code enforces every rule; the prompts only explain them. The council cannot change any number here except one: the swing budget, which it votes on each run inside a range code clamps. Any other change is a policy change, recorded in every sealed run by the policy file's hash.
Hard limits
Constants in the code. The policy files may be stricter, never looser, and a test enforces it.
H2 Kill switch
Whatever the policy file says, code stops adding risk and proposes selling everything at a fall of 25% from the best value ever reached, at the latest.
Show the numbers 2 values
- Halt at a fall of
- 25%
- Measured from
- the lifetime peak
H3 A stop-loss on every opening order
Every order that opens a position carries a stop-loss at the broker.
Show the numbers 1 value
- Required
- yes
H4 A person approves every order
No code path places an order without the operator's approval.
Show the numbers 1 value
- Required
- yes
H5 Only the agent portfolio
Only the dedicated agent portfolio is ever touched; never the main account.
Show the numbers 1 value
- Enforced
- yes
H6 Swing ceilings
The swing policy may be stricter than these, never looser.
Show the numbers 18 values
- Open swing trades, at most
- 6
- Shorts, at most
- 2
- One trade's size, at most
- 8%
- New trades in 7 days, at most
- 6
- Loss at the stop, long, at most
- 0.8%
- Loss at the stop, short, at most
- 0.5%
- Stop distance, long, at most
- 12%
- Stop distance, short, at most
- 8%
- Open risk at the stops, at most
- 4%
- Gap allowance, at least
- 1.5×
- Time stop including the extension, at most
- 20 sessions
- Net reward against risk, at least
- 1.2
- An approved entry is stale after
- 60 min
- Model calls per slot, at most
- 12
- Declared cost per leg, at least
- 1.25%
- Loss brake fires no deeper than
- −5%
- Drawdown scaling starts no later than
- −10%
- Size under drawdown scaling, at most
- 4%
H7 Swing budget ceiling
The council sets the swing budget itself, but code never lets it exceed 50% of the portfolio.
Show the numbers 1 value
- Swing budget, at most
- 50%
H8 Go-live switches
Constants in the code decide what may trade at all; changing one is a code change, not a setting.
Show the numbers 3 values
- Swing book trades live
- no
- Stock sleeve live
- no
- Broker news feed allowed
- yes
Core rules
Every rule the risk engine checks on every run of the core council. Shares are of the portfolio's value; a line's "full size" is the weight the rules alone would give it.
R1 Gross exposure caps
All positions added together (long and short) stay under 190% of the portfolio; code refuses anything above 200%.
Show the numbers 3 values
- Proposals stay under
- 190%
- Code refuses above
- 200%
- On a watch, cut back to
- 180%
R2 Net exposure range and short gross cap
Long minus short stays between −50% and 190% of the portfolio; shorts together at most 60%.
Show the numbers 3 values
- Net exposure at least
- −50%
- Net exposure at most
- 190%
- All shorts together at most
- 60%
R3 Soft kill: warn, then halt, as a fraction of the lifetime peak
Measured from the best value ever reached: −20%: no new risk · −25%: stop, and a proposal to sell everything goes to the human.
Show the numbers 5 values
- Warn (no new risk) at a fall of
- 20%
- Halt at a fall of
- 25%
- Confirming reads
- 2
- Seconds between reads
- 60
- Measured from
- the lifetime peak
R4 Every open carries a catastrophe stop-loss
Every new position carries a stop-loss order set well away from the price (at most 35% away), so one bad day cannot sink the book.
Show the numbers 8 values
- Stop distance
- the larger of the class floor and 3 × daily volatility × √5
- Furthest stop
- 35%
- Floor: index
- 8%
- Floor: etf
- 10%
- Floor: commodity
- 10%
- Floor: commodity gold
- 8%
- Floor: crypto
- 20%
- Floor: fx
- 4%
R4d Cool-off before re-entering after a stop hit
After a stop-loss is hit, the line waits 3 days (7 for crypto) before it can be bought again.
Show the numbers 2 values
- Wait after a stop hit
- 3 days
- Crypto
- 7 days
R5 Per-line and cluster caps
Each line has its own cap (the largest is 80% of the portfolio); crypto together at most 50%, currencies 50%, the three equity lines 160%.
Show the numbers 12 values
- Line cap: NDX
- 80%
- Line cap: SEMIS
- 40%
- Line cap: SPX
- 80%
- Line cap: GOLD
- 50%
- Line cap: BTC
- 40%
- Line cap: ETH
- 20%
- Line cap: OIL
- 20%
- Line cap: EURUSD
- 30%
- Line cap: GBPUSD
- 30%
- Crypto together
- 50%
- Currencies together
- 50%
- Equity lines together (NDX, SEMIS, SPX)
- 160%
R6 Leverage caps by asset class
Borrowing is capped by asset class: at most 1x on crypto, 2x on indices, 5x on currencies.
Show the numbers 5 values
- Leverage: index
- 2x
- Leverage: etf
- 2x
- Leverage: commodity
- 2x
- Leverage: fx
- 5x
- Leverage: crypto
- 1x
R7 Margin use cap (keeps a cash reserve)
At most 95% of the account may be tied up as margin, which keeps a cash reserve.
Show the numbers 2 values
- Margin in use at most
- 95%
- Cash reserve at least
- 5%
R8 Ex-ante book volatility hard cap
The book's expected yearly swings stay below 30%.
Show the numbers 1 value
- Expected yearly volatility below
- 30%
R9 Volatility breaker
If short-term volatility jumps to 3.0 times its usual level on a line (2.0 times for the whole book), no risk is added.
Show the numbers 3 values
- One line: short-term vs usual volatility
- 3×
- Whole book
- 2×
- Volatility card from
- 2×
R10 Council authority bands around the reference
The council may change at most 3 lines per run, each inside a range that code sets from the line's trend.
Show the numbers 6 values
- Lines changed per run, at most
- 3
- Uptrend: cut with a qualifying card, up to
- 50% of the line's full size
- Uptrend: extra leverage, up to
- 50% (only if it passes the cost gate)
- Mixed trend: range
- 0% to 100%
- Downtrend: range
- −50% to 25% (shorts need a cited risk card)
- Expired cards return the line to the reference
- yes
R11 Deadband: changes too small to be worth trading are skipped
Changes smaller than 25% of a line's full size, or under 2% of the portfolio, are not traded.
Show the numbers 3 values
- Skip changes under
- 25% of the line's full size
- Crypto
- 50%
- Or under
- 2% of the portfolio
R12 Minimum holding periods
A position is kept at least 3 days (7 for crypto) before it is reversed; moving back to the reference is always allowed.
Show the numbers 4 values
- Hold at least
- 3 days
- Crypto
- 7 days
- No flip from long to short within
- 24 h
- Moving back to the reference is always allowed
- yes
R13 Turnover limits
Trading is limited to 100% of the portfolio in 7 days and 300% in 30 days.
Show the numbers 3 values
- Added in one run, at most
- 60%
- Traded in 7 days, at most
- 100%
- Traded in 30 days, at most
- 300%
R14 Cost and carry budgets
One run may spend at most 0.4% of the portfolio on trading costs, and 1% over 30 days on discretionary trades.
Show the numbers 4 values
- Trading costs in one run, at most
- 0.4%
- Discretionary costs in 30 days, at most
- 1%
- Overnight financing a proposal may add, a day
- 0.02%
- Financing that puts the book on watch, a day
- 0.03%
R15 Net-of-cost gate
A trade must be expected to earn clearly more than it costs, after spread, fees and overnight financing.
Show the numbers 3 values
- Break-even Sharpe ratio, reference trades, at most
- 0.3
- Break-even Sharpe ratio, council trades, at most
- 0.2
- Assumed holding period
- 20 days (60 crypto)
R16 No adds around scheduled macro events
No adds from 24 h before to 2 h after a scheduled macro event (Fed decisions, inflation, jobs); selling is always allowed.
Show the numbers 5 values
- No adds before a macro release
- 24 h
- … and after it
- 2 h
- Before a company's earnings
- 24 h
- … and after
- 30 h
- An estimated earnings date counts as
- ± 5 trading days
R17 Anti-chase: no adds right after a large one-day move
No buying right after a one-day jump larger than 2.5 times the usual daily move.
Show the numbers 1 value
- No adds after a one-day move larger than
- 2.5σ (its usual daily move)
R18 Data freshness limits
Prices must be recent: daily bars at most 30 h old, quotes at most 60 s old.
Show the numbers 4 values
- Daily bars at most
- 30 h old
- 4-hour bars at most
- 5 h old
- Quotes at most
- 60 s old
- Frozen reference share at most
- 30%
R19 Closed markets
A line whose market is closed is held where it is; nothing is bought or sold on it until the market reopens.
R20 Open blockers
While an operator blocker is open (an unresolved incident or check), the lines it covers are held and no new risk is added; swing ideas are dropped too.
MC Executable changes need new material facts
A council change is traded only if some fact actually changed since the last decision.
Show the numbers 1 value
- Required
- yes
IB Initial build
While the book is first built, a line that has never been filled is exempt from R13, R14 and R15 (churn, cost budget and the net-of-cost gate), until every target line has been filled once or five runs have passed; R14's overnight-financing limit still applies.
Show the numbers 3 values
- Rules paused per never-filled line
- R13, R14, R15
- At most
- 5 runs
- R14 overnight financing
- still applies
R21 Legs per proposal
A proposal holds at most 8 orders.
Show the numbers 2 values
- Risk-adding and discretionary orders, at most
- 8
- All orders, at most
- 16
AP Re-checks at approval time and the notification window
A person approves every order, only between 08:00 and 23:00 Lisbon time; prices and costs are checked again at that moment.
Show the numbers 6 values
- Approval window
- 08:00–23:00 Lisbon time
- Price moved since the run, at most
- 1.5σ of 4 hours
- Book drift since the run, at most
- 10%
- Order size may differ by
- 30%
- Cost may rise to
- 1.5× plus 0.05%
- Exposure after the fills, within
- 5%
RC Post-trade reconciliation tolerances
After trading, the real book is compared with the plan; a gap above 3% is flagged.
Show the numbers 2 values
- Gap between the real book and the plan, flagged above
- 3%
- Stop-loss level tolerance
- 0.2%
PR Proposal priority (a lower one never supersedes a higher one)
A sell-everything proposal always outranks a compliance fix, which outranks a routine rebalance.
Show the numbers 1 value
- Order
- flatten > compliance > rebalance
Source: policy/risk.yaml. Its hash is recorded in every sealed run.
Swing rules
The swing book's rules, run twice: on each idea before the Skeptic sees it, and on the manager's entries before a paper leg is built. Exits are never refused. The S11 badge marks the rule that also sets the Skeptic's starting lean.
S0 Not a core holding
A stock the core council already holds as a line cannot also be a swing trade.
Reason codes it can giveS0:vehicle_owned_by_core the core council already holds this stock
S1 Position size
Each swing trade is about 8% of the portfolio, cut so that hitting the stop loses at most 0.8% (0.5% for a short).
Show the numbers 5 values
- Target size
- 8%
- Smallest size after the stop cut
- 4%
- Loss at the stop, long, at most
- 0.8%
- Loss at the stop, short, at most
- 0.5%
- Short with unknown short interest
- 50% of the size
Reason codes it can giveS1:stop_too_wide_for_size the stop is too wide for the position size
S2 Open trades
At most 6 swing trades open at once, 2 of them short, and one per stock.
Show the numbers 3 values
- Open trades, at most
- 6
- Shorts, at most
- 2
- Per stock
- 1
Reason codes it can giveS2:max_open the book already holds the most swing trades allowedS2:max_short the book holds the most shorts allowedS2:already_open a swing trade on this stock is already open
S3 Weekly cap on new trades
At most 6 new swing trades in any seven days: a ceiling, not a quota.
Show the numbers 1 value
- New trades in a rolling 7 days, at most
- 6
Reason codes it can giveS3:weekly_cap the weekly cap on new swing trades is reached
S4 Open risk
All open and proposed swing trades together may lose at most 4% of the portfolio at their stops, allowing for gaps.
Show the numbers 2 values
- Open risk at the stops, at most
- 4%
- Gap allowance
- 1.5× the stop distance
Reason codes it can giveS4:open_risk the open risk across swing trades is at its limit
S5 Stops
Every trade has a stop between 2% and 12% away (8% for a short), outside one day's normal range; stops only ever move toward safety.
Show the numbers 4 values
- Closest stop
- 2%
- Furthest stop, long
- 12%
- Furthest stop, short
- 8%
- At least
- 1 × the average daily range
Reason codes it can giveS5:stop_missing no stop levelS5:stop_out_of_range the stop is outside the allowed rangeS5:stop_inside_atr the stop sits inside one day's normal rangeS5:atr_unknown the stock's normal daily range is unknown, so the stop cannot be checked
S6 Targets and reward against risk
The target must pay at least 1.2 times what the stop risks, after costs, and be a move the stock can plausibly make in the time allowed. The code gate checks the Scout's levels before the Skeptic sees them.
Show the numbers 5 values
- Net reward against risk, at least
- 1.2
- Target, at least
- 3 × the round-trip cost
- Target, at most
- 30%
- Target, at most
- 1.5 × daily volatility × √sessions
- Daily volatility, at most
- 4%
Reason codes it can giveS6:target_missing no target levelS6:vol_unknown the stock's volatility is unknownS6:vol_too_high the stock is too volatileS6:target_too_small the target is too smallS6:target_beyond_vol the target is further than the stock usually moves in the time allowedS6:cost_unavailable the trading cost could not be pricedS6:net_rr_below_min the reward is too small for the risk once trading costs are counted
S7 Time stop
A trade closes after 3 to 15 sessions; the manager may extend it once, with a new cited fact, to at most 20.
Show the numbers 3 values
- Time stop
- 3–15 sessions
- One extension, at most
- 5 sessions
- In total, at most
- 20 sessions
Reason codes it can giveS7:time_stop_out_of_range the time stop is out of range
S8 Liquidity
Only stocks that trade enough every day and are not penny stocks; shorts need much more trading volume.
Show the numbers 1 value
- Thresholds
- a minimum average daily traded value and share price, higher for shorts; the amounts are in policy/swing.yaml (this site shows no money amounts)
Reason codes it can giveS8:illiquid too little trading volumeS8:price_too_low the share price is too low
S9 Earnings
No entry within 3 sessions of earnings, an exit proposal 1 session before them, and a 30-hour wait after a report.
Show the numbers 4 values
- No entry within
- 3 sessions
- Exit proposal before
- 1 session
- Wait after a report
- 30 h
- An estimated date counts as
- ± 5 days
Reason codes it can giveS9:earnings_window earnings fall inside the trade's windowS9:earnings_window_estimated estimated earnings fall inside the trade's windowS9:post_earnings_wait too soon after earnings
S10 Similar trades and market exposure
At most 2 trades on the same kind of bet, and the swing book's net market exposure stays small.
Show the numbers 3 values
- Trades per bucket, at most
- 2
- Counts as the same bet above a correlation of
- 0.6
- Swing book's net beta, at most
- 0.3
Reason codes it can giveS10:bucket_full the book is already full of similar tradesS10:swing_net_beta the swing book's market exposure is at its limit
S11 No chasing
An idea is dropped if the stock already moved more than 4σ since the news; from 2σ the Skeptic leans towards wait.
Show the numbers 2 values
- Dropped above
- 4σ since the news
- The Skeptic's wait prior from
- 2σ
Reason codes it can giveS11:chased the move since the news was already too large
S12 Fees
Fees are measured and published, never a brake: a budget of 1% of the portfolio over 30 days.
Show the numbers 4 values
- Mode
- reported only
- Fee budget over 30 days
- 1%
- Fees against gross gains, at most
- 33.3%
- Measured after
- 10 closed trades
Reason codes it can giveS12:fee_budget over the fee budget
S13 Shorts
Shorts are 1x with a hard stop, and never on a recent listing, a crowded or squeeze-prone name, a takeover target or a stock that just flushed.
Show the numbers 7 values
- Listed at least
- 180 days
- Short interest, at most
- 20% of the float
- No short after a fall of
- 3σ
- No short after a 20-day rise above
- 20%
- Squeeze risk: price within this far of its high
- 3%
- … with volume above
- 3× normal
- Takeover news looked back
- 90 days
Reason codes it can giveS13:short_new_listing a recent listing cannot be shortedS13:short_crowded too crowded a shortS13:short_takeover_target a takeover target cannot be shortedS13:short_into_flush a short into a flush is not allowedS13:short_squeeze_risk squeeze risk on a short
S14 Cool-off
After a stop is hit the stock waits 5 sessions; after any other exit 2.
Show the numbers 2 values
- After a stop
- 5 sessions
- After another exit
- 2 sessions
Reason codes it can giveS14:cooloff a cool-off after a recent loss on this stock
S15 Loss brake
A net loss of 5% of the portfolio over 30 days, after all costs, stops new swing trades; only the operator lifts it.
Show the numbers 3 values
- Loss that stops new trades
- −5%
- Window
- 30 days
- After all costs
- yes
Reason codes it can giveS15:brake_on the swing brake is on (30-day net loss limit): no new entries until the operator lifts itS15:brake_engaged new entries are paused after the Skeptic canary check failed, until the operator lifts itS15:brake_unknown the swing brake could not be checked, so no new entry
S16 Entry guard at approval
An approved entry is a market order within 60 minutes of the run; it is dropped if the price already ran too far or hit the stop.
Show the numbers 3 values
- Approve within
- 60 min
- Price run, at most
- 50% of the stop distance or 1.5%
- Re-proposals, at most
- 2
Reason codes it can giveS16:swing_entry_ran the price ran past the entryS16:swing_entry_stopped the price hit the stop before entryS16:expired the approval expired
S17 Drawdown scaling
From a fall of 10% below the peak, trades shrink to 4% with at most 3 open; the core kill switch's warn or halt blocks new entries.
Show the numbers 3 values
- Starts at a fall of
- 10%
- Size then
- 4%
- Open trades then, at most
- 3
Reason codes it can giveS17:kill_state the risk engine's kill switch blocks new entriesS17:drawdown_unknown the drawdown could not be checked
S18 The swing budget is full
An entry is refused if it would push the swing trades above the swing budget the council set (below).
Show the numbers 1 value
- Budget
- see the swing budget card
Reason codes it can giveS18:swing_budget_full the swing budget the council set this slot is full: no room for another entry
Budget The swing budget (set by the council)
The swing manager's attempts each vote a swing budget from 0 to 50% of the portfolio in steps of 5; code takes the median and clamps it. Unused swing money is invested in the core.
Show the numbers 6 values
- Range
- 0–50%
- Steps
- 5%
- Decided by
- the median of the swing manager's valid votes
- Before the first valid vote
- 40%; later the last budget is kept
- Code ceiling
- 50%
- Core re-sized when swing exposure moved by
- 4%
SB16 Which setups may trade
Only some setups may trade; the rest are tracked on paper only, to measure them.
Show the numbers 2 values
- May trade
- news continuation, post-earnings drift, second-order effect, day-2 confirmation: a Skeptic 'wait' re-checked a day later
- Paper only
- gap fade, breakout, mean reversion, event run up
Reason codes it can giveSB16:setup_paper_only a paper-only setup: tracked on paper, never traded
Source: policy/swing.yaml. Nothing here trades with money while the swing book's go-live switch is off.