> > >
MKDC Research · Original Studies

The Research
Section.

Original metrics and studies from MKDC Baseball, built the same way the call-up model is built: public data, stated limits, and numbers that stay on the record. The prediction side of the house lives on the PIX board and the methodology page. This is where the new ideas get tested.

Studies

Deadline Analysis · July 30
Two Ways to Be Blocked
James Tibbs III is blocked by other people. Theo Gillen is blocked by a philosophy. Monday’s deadline can fix one of those. Inside the org coefficient that prices both.
July 30, 2026
Board Analysis · July 21
Cleveland Is Promoting Prospects in Waves
Genao +23, Velazquez +21, Doughty +17, all in one week from one system. How a single front office’s promotion wave shows up as correlated movement on the live board.
July 21, 2026
Model Documentation · v0.5
Inside v0.5: How the Model Actually Runs
The 13 features, the four multipliers, the recalibration that cut Brier from 0.269 to 0.148, the live PIX layer, five known problems, and what v1.0 fixes.
July 16, 2026
Prospect Outcomes · 36 Years
What a Top 100 Ranking Is Actually Worth
Making the list turns a 17.6% shot at the majors into a 91% one. The gradient from No. 1 to No. 100, the WAR curve, hitters versus pitchers, and where the stars actually come from.
July 14, 2026
New Metric · BSAx
The Bat Speed the Leaderboard Misses
Bat Speed Above Expected: who swings hardest for their size and age. Carroll, a 35-year-old Yastrzemski, the age-defiers, and Caminero, who needs no adjustment at all.
July 11, 2026
Model Research · Calibration
Inside the June 18 Recalibration
Every checkpoint claim reproduced, the v0.5 probability fit shown step by step, and the honest reason we publish 0.148 instead of the prettier number.
July 1, 2026

How research gets published here

Same rules as the model. Public data only. Limits stated up front. If a metric cannot survive a repeatability test, it does not run. And when a study makes a claim the season can grade, the grade posts either way, the same discipline that runs the calibration log.