What is service time suppression and how does it affect prospect call-ups?

Service time suppression is the practice of keeping a prospect in the minor leagues for the first 2-4 weeks of the season to delay their service time clock. The MKDC model applies a 0.65x multiplier for this period.

What is the PPI deadline for MLB prospects?

PPI (Prospect Promotion Incentive) is a compensatory draft pick for teams that promote eligible prospects by a specific deadline, typically early April. The 2026 PPI deadline is April 9.

What does ECD mean in baseball?

ECD stands for Estimated Call-up Date. It is a three-value range showing the earliest, most likely, and latest expected date for a prospect's MLB call-up.

How is MKDC Baseball different from prospect rankings?

MKDC Baseball predicts call-up timing using structural factors, not talent evaluation. All predictions are logged publicly before outcomes and scored via Brier score.

Call-Up Probability Engine · Model v0.5 · Calibrated on 28 Resolved Outcomes · 107 Predictions on the Record

How MLB Prospect
Call-Up Probability
Actually Works

Q: What does CUP mean in baseball prospect rankings?

CUP stands for Call-Up Probability. It is a binary 90-day probability that a minor league player will make at least one MLB appearance within the next 90 days.

The full mechanics of the MKDC Baseball call-up probability model: thirteen scored features, four multipliers, a recalibrated probability table, and 107 locked predictions across three cohorts, every one graded when its window closes.

This page is the plain-English version: what the model does and how to read it, no math degree required. The full technical documentation, with citations, known limitations, and the v1.0 build plan, is the research paper Inside v0.5.

MKDC Baseball answers one question: will this minor league player make an MLB appearance within the next 90 days? Not whether he is talented. Not whether he will stick. Whether the phone rings inside the window. Every output is locked at its run date and graded against the outcome, hit or miss.

What This Model Is Not

This model does not evaluate swing mechanics, pitch mix, athleticism, or any subjective scouting factor. It does not predict whether a player will succeed in the majors. It prices one thing: the structural forces that decide when a team makes the call.

Scored Features

Multipliers

Probability Bands

107

Locked Predictions

.148

Corrected Brier

Cohorts on Record

How to Read a CUP Score

CUP is a binary 90-day probability. A 45% means the model gives that player a 45-in-100 chance of at least one MLB appearance before his window closes. It is not a talent grade, and a low CUP is often a statement about a calendar or a roster, not a player. The fast-moving companion number, PIX, tracks live promotion pressure between batch runs and is never used for scoring or calibration. Original studies and new metrics, including BSAx, live in the research section.

The Four-Stage Pipeline

CUS

Score 13 fixed features into a raw total (max 153), normalized to 0-100.

→

Multipliers

Four sequential adjustments: org, position, service time, PPI.

→

CUP

Map through the v0.5 band table to a 90-day call-up probability.

→

ECD

Assign a projected call-up window via one of three pathways.

13 Feature Inputs

Level, age vs level, 30-day performance, 40-man status, org opportunity, options, position scarcity, trend, spring signal, ranking tier, Rule 5 exposure, and PPI status. All from public data at the run date.

Stage 1–2

CUS Score 0–100

All 13 feature scores sum to a raw total out of 153, normalized to 0–100. Weights live inside the defined scoring ranges. Locked at run date.

Stage 3

CUP Probability

Four multipliers adjust the CUS first: org behavior, positional base rate, service time, and PPI. The adjusted score then maps to a probability through the recalibrated v0.5 table.

Stage 3

ECD Engine

CUP converts to an expected call-up date through one of three pathways, each with a confidence tier.

Stage 4

The 13 Model Features That Drive Call-Up Probability

Every feature has a fixed scoring range, published here in full. No freehand points. The maximum raw total is 153.

Feature	What it measures	Scoring
Level Proximity	Level of assignment	AAA=18 · AA=10 · A+=4
Age vs Level	Age vs league average	+3yrs young=18 · +1=15 · even=12 · old=9
Performance (Pitcher)	30-day K-BB%	28+=19 · 22+=17 · 16+=14 · 10+=11
Performance (Hitter)	K%, BB%, ISO composite	0 to 20
40-Man Status	On the 40-man roster	Yes=16 · No=4
Org Opportunity	IL count, churn, win%	0 to 20, three sub-scores
Options Remaining	Minor league options left	0=16 · 1=12 · 2=8 · 3=4
Position Scarcity	MLB depth at position	Sparse=18 · Thin=13 · Avg=8 · Deep=4
Performance Trend	30-day vs 60-day direction	Improving=+5 · Stable=0 · Declining=-3
Spring Signal	Spring training tier	Elite=5 · Strong=3 · Avg=0 · Poor=-2
Prospect Ranking	Org priority proxy	Top10=3 · Top30=1 · Unranked=0
Rule 5 Exposure	Rule 5 draft risk	Yes=6 · No=0
PPI Status	Org incentive to promote	High=8 · Medium=4 · None=0

The Four Multipliers

Org Coefficient

0.85 to 1.02

Thirty organizations, one behavior prior each. Colorado tops the scale at 1.02. Tampa Bay anchors it at 0.85. The full reasoning lives in the org guide.

Behavior prior

Positional Base Rate

0.85 to 1.15

Some positions get called up more. Relievers ride the high end at 1.15. Catchers carry the 0.85 anchor because teams promote them last and slowest.

Position prior

Service Time

0.65x early season

First-call-up players in the season’s opening weeks get suppressed, because that is what teams do. The window passes, the multiplier releases.

Calendar-driven

PPI Multiplier

1.12x / 1.06x / 1.0x

The CBA pays teams a draft pick for promoting stars early. When the incentive is live, the model prices it. After the April deadline, it zeroes out.

CBA incentive

PPI: The Draft-Pick Incentive, Priced

Since 2022, an org that carries a qualifying top prospect on its Opening Day roster can earn a bonus draft pick if he wins Rookie of the Year or places top three in MVP or Cy Young voting. The model scores the org’s incentive to try, never the player’s award odds. Eligibility requires two of three major Top 100 lists, no prior debut, no pre-debut extension, and promotion by roughly the 13th game. For 2026 that deadline was April 9, which is why every mid-season evaluation carries PPI None.

Player	Org	Year	Award
Julio Rodriguez	SEA	2022	AL ROY
Gunnar Henderson	BAL	2023	AL ROY
Corbin Carroll	ARI	2023	NL ROY
Bobby Witt Jr.	KC	2024	AL MVP top 3
Drake Baldwin	ATL	2025	NL ROY
Hunter Brown	HOU	2025	AL CY top 3

Six picks awarded from roughly forty eligible players in four seasons, about a 15 percent conversion. The incentive is real. The payoff is rare. The model prices the first thing and refuses to conflate it with the second.

CUS to CUP: The Recalibrated Mapping

The June 18 checkpoint showed v0.4 ranked players correctly but printed probabilities that ran low. The best-rated group reached the majors 100 percent of the time against a 41 percent average prediction. v0.5 rebuilt exactly one thing, the mapping table, from the 28 resolved outcomes. The refit shrinks small samples toward their predictions and fits in log-odds space. On the resolved cohort it cut the Brier score from 0.269 to 0.148. The full workup is in the technical report.

CUS band	Base CUP
under 30	5%
30 to 39	12%
40 to 47	22%
48 to 54	33%
55 to 61	45%
62 to 69	62%
70 to 77	78%
78 and up	88%

under 305%

30-3912%

40-4722%

48-5433%

55-6145%

62-6962%

70-7778%

78+88%

v0.4 · original

0.269

Ranked players correctly, but probabilities ran low.

v0.5 · corrected

0.148

Same 28 outcomes, remapped. Lower Brier is better.

ECD Pathways

High conf

Service time driven

Tight window

The calendar decides. Suppression windows and control math produce narrow, predictable dates.

Medium conf

Opportunity driven

Medium window

An injury or roster move opens the door. Validated at the June 18 checkpoint through the Konnor Griffin resolution.

Low conf

Performance driven

Wide window

The default pathway. The player forces the issue and the date floats with the performance.

A note on the open cohorts: Sub-Batch B and C windows are derived by a fixed documented rule from the model’s own features, expressed by month with a confidence tier. The rule and every input are in the ECD run record. ECD is directional and is never graded by Brier.

The Cohorts: 107 Predictions on the Record

Three batches, each locked at its run date and graded when its 90-day window closes. Together they cover every healthy player on the current Top 100 plus tracked carryovers.

Cohort	Run date	Size	Status	Notes
Sub-Batch A	2026-03-20	28 scored	Resolved June 18	14 of 28 reached MLB · all 28 ranked in order · corrected Brier 0.148
Sub-Batch B	2026-06-11	14 scored	Closes Sept 9	First cohort on the recalibrated v0.5 mapping
Sub-Batch C	2026-07-10	65 scored	Closes Oct 8	15 featured + 50-player full sweep of the healthy Top 100

By October 8, the model will have 107 resolved binary outcomes. The v1.0 training gate was 50. The Sub-Batch C full sweep deliberately scored fifty long-shot players so the low-probability bands, thin at the June checkpoint, get real outcome data before training begins.

The Full CUP Board: Every Open Prediction

Every unresolved locked prediction across the open cohorts, refreshed as players resolve. Live movement between run dates lives on the PIX board and never feeds back into these numbers.

The full CUP leaderboard has moved. All 79 open predictions, with projected call-up windows (ECD), live PIX, and cohort, now live on the home page board, alongside the live PIX board. Resolved predictions are graded on the calibration log.

Calibration Commitment

Every prediction gets graded, hit or miss

Outputs are locked at run date. Checkpoint results post with the original numbers, a Brier score, and a named reason for every miss. June 18: resolved, 14 of 28, perfect ordering, corrected Brier 0.148. September 9: Sub-Batch B resolves. October 8: Sub-Batch C resolves and the v1.0 training set locks. The record lives on the calibration log.

What CUP Measures, and What It Does Not

Inside the model

Timing of a first MLB appearance inside 90 days

Structural forces: levels, rosters, calendars, incentives

Org behavior priors from documented history

Public data only, locked at run date

Outside the model

MLB success, stardom, or fantasy value

Scouting judgments and tool grades

Front-office intent it cannot observe

Award probability, even inside PPI

Version History

The complete v0.5 documentation, including known limitations and the v1.0 build plan, is published as a research article.

v0.1–v0.3

Core features, service time suppression, org coefficients, positional rates, the ECD engine.

v0.4

PPI feature and multiplier. Normalization to 153. Sub-Batch A locked March 20.

v0.5 — Current

Recalibrated CUS-to-CUP mapping from 28 resolved outcomes. Injury humility band. Powers Sub-Batches B and C.

v1.0 — 2027

Trained model on 107 labeled outcomes. Empirical org coefficients and positional rates. Training set locks October 8, 2026, when the 50-resolved-outcome gate clears.

Frequently Asked Questions

What does CUP mean in baseball prospect rankings?

CUP stands for Call-Up Probability: a binary 90-day probability that a minor league player makes at least one MLB appearance within the window. A CUP of 45% means a 45-in-100 chance. It does not predict success, only the appearance.

How many prospects does the MKDC model track?

107 scored predictions across three 2026 cohorts: 28 resolved in Sub-Batch A, 14 resolving September 9 in Sub-Batch B, and 65 resolving October 8 in Sub-Batch C, which covers every healthy player on the current Top 100. The live PIX board tracks the same players between batch runs.

What is the difference between CUP and PIX?

CUP is the scored probability, locked at a run date and graded against outcomes. PIX is a live directional index that moves on sourced news events between runs. PIX is not a probability and is never used for calibration.

How accurate is the MKDC call-up model?

At the June 18 checkpoint, every prediction band resolved in the correct order and all six players rated 30 percent or higher reached the majors. The corrected Brier score was 0.148, a recalculation on the resolved data; the original 0.269 stays on the log. The cleaner test is Sub-Batch B, graded September 9. Every checkpoint publishes the same way, misses included.

How MLB ProspectCall-Up ProbabilityActually Works