Will the staff act on it, and what do we learn when they don’t?
How much the plans ask
Every development goal, by what it asks of the player and staff
- Low 11 · 19%A game-plan change: same pitches or swing, used differently
- Medium 18 · 31%Practice and targets: command, approach, velocity he has already shown
- High 23 · 40%Mechanics, a swing change, a reshaped or new pitch
- Medical first 6 · 10%A health check before anything else
| Goal type | Ask | Goals | Players |
|---|---|---|---|
| Diagnose | Low | 6 | 6 |
| Health note | Low | 4 | 4 |
| Health check | Low | 1 | 1 |
| Command | Medium | 8 | 8 |
| Approach | Medium | 5 | 5 |
| Velocity | Medium | 4 | 4 |
| Position | Medium | 1 | 1 |
| Contact | High | 9 | 9 |
| Impact | High | 5 | 5 |
| Pitch shape | High | 3 | 3 |
| Stuff | High | 3 | 3 |
| Bat speed | High | 2 | 2 |
| New pitch | High | 1 | 1 |
| Health | Medical first | 6 | 6 |
2 players on the Winter Plan’s 2027 roster have only high-ask goals (Jared Triolo, Spencer Horwitz): start that conversation in person, not with a card. Low-ask goals are rare because the models look for pitches and swings to change; that’s where the staff’s judgment matters most.
Pitchers already make the easy change
2025 dead-weight flags graded on 2026: did the pitcher cut the pitch, and what happened?
Observational and small: pitchers who change may differ in other ways, so this is a reason to lead with low-ask goals, not proof that they work. The 2025 flags were picked with Stuff from a model trained on 2024 only, so nothing the selection used had seen 2026. It’s on the Track Record with the other tested claims.
Staff responses
Accept, Modify or Not now on any coach card or Decision Brief. Saved in this browser; in production these write to the club’s decision log.
No responses yet. Open the coach cards or a Decision Brief and answer a recommendation: it shows up here, grouped by ask level, with the reasons.
When the staff says no, the rules change
Feedback from reading Keystone the way a front office would, and what it changed
- “Pitchers already throw with as much movement as they can.” Shape goals now target only a pitch the player has thrown before, by more than a pitch’s grade drifts by chance; velocity goals use his own best season net of aging.
- “That club would never trade him.” Every trade target carries an availability read (just signed, young core, a contender’s regular), and the Winter Plan only acquires players a club would plausibly move.
- “He missed the whole season.” A healthy 2026 is now part of the acquisition rule, and trade targets who missed most of it carry a medicals flag.
- “Scouts can’t confirm that; you can.” Generic scouting prompts became computed checks with a run value; scouts keep what numbers can’t answer.
Each one is in the decision log with the alternatives considered.
What we’d measure in production
- Adoption rate by ask level. Low-ask goals should clear 70%; if high-ask goals sit far below, the plans are asking too much.
- Time to decision. Days from a recommendation to Accept, Modify or Not now; anything older than two weeks is a stalled conversation.
- Outcome by response. On May 31, goals met for adopted vs. not-adopted recommendations, so the models are graded on what was actually tried.
- Reasons, coded. “Not now” reasons tagged (health, timing, disagree with the data, player buy-in) and reviewed monthly; repeated reasons become rule changes.