On, on the bench, or off
An algorithm is on (its view is read and, with weight, counts in the score), on the bench (on with weight 0, read but not counted), or off. Nothing on the Signals page can block a trade.
- Who it is for
- Owners & operators
- Reading time
- 3 min read
- Updated
Every row on Signals has one switch, On or Off, and, for a , one weight. That is the whole control surface. What the PM reads and what counts in the score follow from those two things.
The three states a signal can be in
On, with weight
- The PM
- Reads its view on every name
- The score
- Counts, at the share shown on its row
(on, weight 0)
- The PM
- Reads its view
- The score
- Ignored; the row says "On the bench"
Off
- The PM
- Does not see it
- The score
- Ignored
Both switches are one click on the row· Signals › On / Off and weight, take effect at the PM's next session, and are logged under Activity › Config changes with who and when. New arrive on the bench, so a newcomer never dilutes the score you chose until you give it weight.
Two kinds of algorithm have no weight
is a rule machine. The says how much the signals like a name today; Jim's method answers a different question: is there a setup on the name today, and at what prices (the buy-above trigger, the stop, the first target). A fired setup is one of the conditions the PM may act on. Its reading is never averaged into the number, so its card has no weight to set, the blend card never mentions it, and it lives on its own tab, Setups. A setup in a down week is a short setup, which this desk never trades: the table shows its sell-below level for the record and never counts it as armed.
A (see Signal, Screen, Setups) would never move the score either: it could block a buy in a name that fails it and say "Blocked by …" with the reason wherever the name appears, never a sell. No book has one while the buy block is retired (below); the card shape exists for the day a later decision brings screens back.
Nothing here is a veto any more
Until 11 September 2026 the switch had a third position that turned an algorithm into a buy block at the . The first evidence round measured every such block over fourteen years of sessions: none added value and two were harmful, so the setting was retired for every book and the change was written to each book's strategy history. A book that still had one stored shows a one-line note on that card; the next save clears it. The hard stops that remain are the risk rules on Risk, which no algorithm and no agent can argue with.
Choosing
Leave everything you have not looked at on the bench: the PM still sees the view, and the score stays the one you chose. Give weight to a signal once its backtest on its detail page, for example cross-sectional momentum, and its live record on Performance convince you. Switch a signal off only when you do not want the PM to see it at all.