Quant Mode

Alpha Hunts

Create it once. Run it whenever you want.

An alpha hunt takes one idea the whole way: build, test, fix, stress-test and grade. You create and name the hunt, then choose how much direction to give it. Leave the optional fields alone and we pick the idea, markets and timeframe. Or choose the exact universe, set a minimum prospect grade and add your own recipe before saving the hunt to run again whenever you want.

Every alpha hunt has the same flat 20,000 premium-token charge and the same grade from F to S. The title is the only required field; everything else is optional. See Models, tokens and usage.

Create an alpha hunt

  1. Open Quant Mode in the sidebar of the app.
  2. Click Run an alpha hunt. The first time, the builder opens. After that, click Create a new hunt at the bottom of the list.
  3. Name it. Everything else is optional.
  4. Choose Save and run to start it now, or Save to keep it for later.

Once you have saved a hunt, Run an alpha hunt becomes a list. Click a title to run that hunt again using the same recipe.

What's in the recipe

Title. Required. Something you'll recognise in the list, e.g. "SMC on forex, 15m–4h".

Universe. Choose any combination of Forex, Crypto, Indices, Stocks, Commodities and Futures, or leave them blank and we choose. You can also select Create custom to name your own universe and build an exact list of markets across stocks, FX, crypto and Horizon futures. A custom universe replaces the broad categories and keeps the hunt inside your list.

Base idea. Type the theme you want it to keep working from: "order blocks on the London session", "mean reversion after a liquidity sweep", whatever you actually trade. This is the foundation — the hunt builds around it instead of inventing a new theme each time.

Timeframe. A two-handle slider: 1m · 5m · 15m · 30m · 1h · 4h · 1d. Drag a range and the hunt picks inside it. Both handles on the same tick pins a single timeframe. Leave it untouched and we pick.

Minimum prospect grade. Choose C, B, A or S as the lowest grade you are willing to review. The default is B. A strategy below your minimum is killed instead of appearing under Prospects, so raising the floor means fewer prospects and a longer wait for one that qualifies.

The tighter you make the recipe, the smaller the search. The builder says so when you narrow it. That's a warning, not a lock — a very tight hunt may take a few runs to find a prospect.

Use your own code

Some ideas have a fixed definition. An EMA is an EMA. Order blocks, fair value gaps, "my supply and demand" — those vary from trader to trader. There is no one correct formula. If you type "order blocks" and leave it at that, the hunt codes its order blocks. If you have your own, attach it so the hunt uses your implementation.

  1. In regular chat, set the platform to Horizon (not TradingView, not MetaTrader, not NinjaTrader).
  2. Build the script the way you actually mark the chart — your order blocks, your indicator, your filter.
  3. Back in Quant Mode, open the hunt builder. In Base idea, click + My Scripts or type @ and pick that script.

The hunt is told to use your implementation, not recode it. It builds the strategy around your code.

A few rules, said out loud:

  • Horizon scripts only. Pine, MQL5 and NinjaScript will not appear in the picker. Build the piece in Horizon mode first; you can export the finished strategy later.
  • Up to 3 scripts on one hunt. That's a building block, not a library.
  • If you have no Horizon scripts yet, the picker says so and points you at Create.

Save it and run it again

There is no "run without saving." The first hunt you configure becomes a saved hunt. After that:

  • Click a title in the list → it runs immediately, using the same recipe.
  • Pencil → edit, then Save or Save and run.
  • Trash → two clicks to delete. Delete the last one and the next Run an alpha hunt opens the builder again.

What does not change

You steer the inputs. The research process stays the same: sweep, improve, stress-test, calculate a grade from F to S, and kill failed ideas. Same graveyard, same playbook, same Prospect / Approve / Kill flow.

There is no portfolio picker in the hunt recipe. A hunt can produce a prospect, then you decide whether to approve or kill it. Only after approval can it enter a portfolio, either through Build best portfolio or by adding it manually in My Scripts.

Agents are still "the hunt, on a schedule." Choose a saved hunt when you create one, and the agent runs it while you are away.

Fine print

  • Title is the only required field. A named hunt with no optional selections means we pick the idea, markets and timeframe.
  • Attached scripts are read when the hunt starts, so it always sees the current code, not a snapshot from when you saved the recipe.
  • Narrower recipes (one market, one timeframe, a written idea, an attached script) can take more hunts to find a prospect. The builder warns you.

Next

Quant Mode: what happens after you click →