How to build a screen

Three kinds of row exist, and they are separate worlds. A batter row knows the starter he faces. A pitcher row knows the order he faces. A team row knows both starters, both bullpens, both orders and the park.

Picking a market picks the rows for you. Hits and home runs are batters, strikeouts are starting pitchers, run line and game total are teams. Switching market can therefore switch what you are screening, and because the available stats are different, it starts you on a clean screen rather than leaving blocks behind that could not be evaluated.

Every block is a sentence: Hits in last 5 ≥ 4. Click one in the catalog to add it to the end of your screen, or drag it to drop it exactly where you want. Everything on a block is editable after you add it, including the stat itself, so the catalog is a starting point rather than a fixed choice.

Blocks in a group interlock, like puzzle pieces. The group header says how they combine: ALL of them must pass, ANY one of them, or NOT the one inside. Nest a group inside a group (up to 2 levels) to say things like “top of the order, and either hot in the last five or facing a soft bullpen”.

The stack marked Screen is the one that runs. Anything else on the canvas is a draft: drop a block on empty space to start one, and build there without touching your live screen. Drafts are never evaluated.

When a draft is ready, → screen folds it in as a group. Give it a name and save and it joins My blocks at the top of the catalog, ready to drop into any screen. Saved blocks live in this browser, so they do not follow you to another device.

Above the results, the funnel shows how many rows were on the slate, how many survived your filters, and how many were finally selected. When a screen returns nobody, the funnel tells you at which stage they went.

Click any row to open Explain: the actual value of every block for that row, against the threshold you set. If a row only passed because a stat was missing and your policy kept it, Explain says so.

Backtest replays the exact screen over past slates and reports the hit rate. It runs the same code the live builder runs, so what you test is what you get.

Two numbers always appear together: the rate, and how many picks settled. They mean nothing apart. A perfect record on four picks is not a better screen than 58% on four hundred, which is why the leaderboard ranks nothing below 25 settled picks.

You can keep 10 playbooks privately. Publishing freezes a copy and lists it under your username, where anyone can fork it into their own library and edit it freely. A published playbook cannot be edited in place: its record has to belong to the screen that earned it.

Unpublishing removes it from the catalog and drops the standing it built. Publishing needs a username, which you set on your account page.