Cadence, decisions and the track record

Left alone, a Squad only wakes when a strategy signals it. That makes it hard to judge: it deliberates at the mercy of how often your indicators fire, so it never accumulates enough calls to show whether it is any good.

Turning on cadence puts the Squad on its own clock. It picks its own markets, wakes on a schedule, and publishes a decision for each one — whether or not any strategy asked. Those decisions are scored against what the market did next, which is what turns a Squad from something you hope is working into something you can measure.

Crucially, a cadence run never trades. It publishes an opinion. Your strategies read that opinion and decide what to do with it.

The five decisions#

Every wake produces exactly one of these per market:

The two risk-off states are about exiting, not reversing. A Squad that wants out of a long but has no appetite for a short says 2, not -1.

Choosing a cadence#

The minimum is 30 minutes, and that is a floor rather than a suggestion. Every wake is a full deliberation — the whole cast argues, in parallel, once per market you have enabled — so a Squad covering five tickers at 30 minutes is doing 240 deliberations a day. Faster than that costs real money without producing better decisions.

A slower cadence is not a worse one. A Squad on a 4-hour clock making considered calls will out-score one on 30 minutes reacting to noise.

The interval is a timeframe, the same vocabulary your strategies use, and a Squad wakes on that timeframe's boundary — a 1-hour Squad at the top of every hour, a 4-hour Squad at 00:00, 04:00 and so on. That is deliberate: it means the Squad deliberates moments after the candle closed, so the readings its specialists get are the closed bar rather than a half-formed one.

Your strategies do not need to match the cadence. The decision line holds its last value until the Squad speaks again, so a 5-minute strategy reads a 4-hour Squad's standing call perfectly well. This is the point of the design: one Squad's opinion serves every strategy on that market, at any timeframe.

What the committee sees#

Add indicator readings and every specialist gets them before arguing. A reading is ticker-agnostic — "RSI, 1 hour, length 14, last 5 values" — and is computed separately for each market you cover, so you configure it once for all of them.

Readings can be on any timeframe, including several at once. A higher-timeframe reading is aligned to its most recently closed bar, so a 4-hour EMA read at 10:15 reflects the 4-hour candle that actually finished.

Use Preview to see the exact text the agents will receive. If a reading shows as unavailable there, that is what the specialists would see too — usually a market with no candle history yet.

How the score works#

Each decision is stamped with the market price at the moment it was made, and graded when the next decision arrives:

• long scores positive if price rose, negative if it fell.
• short scores the reverse.
• risk off long scores positive if price fell — you got out in time — and negative if it rose strongly, because you sold too early.
• risk off short scores the reverse.
• hold inherits the Squad's last directional call as its bias and rides it at half weight: holding through a move that confirms the bias earns, holding through a move against it costs. A hold after a risk off call — or before the Squad has ever gone long or short — is a neutral hold: it scores nothing either way and is left out of the accuracy percentage, because there was no position to be right or wrong about.

Being right during a large move earns more than being right during a small one, up to a cap, so a single lucky call cannot dominate the record.

The deadband (0.5% by default) sets the scale on which a hold is graded. A move inside the deadband still counts for or against the bias, but scaled down by how much of the band it covered — a tiny drift scores near zero, a move at the band edge scores the full half weight. There are no free points for a quiet tape. Raise the band on a volatile market so hold is not graded hard on ordinary noise.

Points accumulate into a standing: positive means the Squad's calls have been paying off, negative means they have not. The Squad is shown its own record before every deliberation, and it reflects on that record first — so a Squad that has been wrong in a particular way has the chance to notice.

Trading a committee's decisions#

Add the Committee Decision indicator to a strategy and pick the Squad. It gives you five lines:

• decision — the standing call, held between wakes.
• confidence — how sure the Squad was, 0 to 1.
• committee_points — its cumulative score. Gate on committee_points > 0 to follow only a Squad that has been paying off.
• last_decision_points — how the most recent graded call scored.
• decision_age_minutes — minutes since the standing call was made.

Always gate on decision_age_minutes. The line holds its last value indefinitely, which is what lets a fast strategy read a slow Squad — but it also means a Squad you switched off keeps publishing its final opinion forever. A limit of roughly two or three cadence intervals is a sensible guard.

A Squad-driven strategy backtests like any other, over the period the Squad has actually been running. Before its first decision the line reads 0, so a backtest starting earlier will simply be flat until the Squad comes alive — that is honest, not a bug. There is no way to manufacture history for a Squad that did not exist yet.