Grade a roster
grade_rosterGrades which jobs a roster can do on the table, as the roster builder's grades do: score 0 to 100 and grade A to D overall and per category (board control, mobility, damage against reference targets, durability), placed among tournament lists of the same battle size and centred on comparedWith's armies, so a list's distance from its army's typical list reads on every army's spread. A typical tournament list scores about 70. It measures coverage, not strength: in tournament games the players' skill decided far more than any list measure, the overall grade went with winning only slightly, and more units with an even spread of points went with winning most. So never call a list competitive from its grade, and never present it as a win rate. missing names the reference targets nothing in the list can take half of alone. units gives each unit the jobs it is credited with; a leader or support character counts inside the unit it joins, so its buffs never show as its own job, and cutting it for another unit always raises board presence. A unit with one joined to it lists it under attached, with each target whose share of wounds the unit takes changes without it: with and without, where 1 is one whole target. Weigh that before cutting it. tournament, when the army has enough top-four lists, gives stapleShare (the share of its points in datasheets at least 3 in 10 of those lists field, which went with winning slightly), the staples it leaves out and its unproven datasheets that no winning list fields. calibration.lists is 'drawn' when the battle size was compared with lists drawn from 2,000-point tournament lists rather than real lists of its size, a rougher comparison, and there nothing shows that more units or an even spread go with winning. It measures units at full strength with passive rules only, from the exact simulator, so it rarely tells detachments apart; compare detachments with simulate_combat. Each row's notCalculated names units or rules it could not measure. Takes a few seconds; check the list with check_roster first, since a list that cannot be fielded still grades.
Tool call template
Replace each placeholder with a value matching the input schema. Only required parameters are included; add optional parameters as needed.
{
"id": 1,
"jsonrpc": "2.0",
"method": "tools/call",
"params": {
"arguments": {
"roster": "<roster>"
},
"name": "grade_roster"
}
}Input Schema
| Name | Required | Description | Default |
|---|---|---|---|
| roster | Yes | A complete roster in the shape save_roster takes, such as the roster get_my_roster returns or a draft built from list_units and get_unit_options ids. | |
| llm_model | No | The model identifier you are running as, for example "claude-opus-4-8" or "gpt-5.2". Pass "unknown" if you are not certain. Recorded for the server developer's analytics. | |
| conversation_id | No | The conversation_id this server returned earlier in this conversation, unchanged, including after later user messages. Omit it if you have none. Recorded for the server developer's analytics. |