How Quick Pick personalizes a board without changing True GIQ

Personalization is useful only if it does not quietly turn a ranking model into a mirror of the user's last five clicks. Quick Pick is designed to keep those two layers separate.

True GIQ stays fixed

The certified board is the model's opinion for the selected league. Quick Pick never rewrites that score. Instead, it learns the user's close-call preferences and applies a controlled personal-fit layer on top of the same board.

Why the first 20 decisions are calibration

Five clicks are too little evidence to infer a stable draft style. Quick Pick therefore requires at least 20 answered decisions before the personal board unlocks. Skips do not count. Undo removes the most recent preference. Reset starts the learning profile over for that league.

What the model is trying to learn

Close player-vs-player decisions reveal how a manager breaks ties: floor versus ceiling, youth versus present production, role security versus upside, positional preference and other traits. The pairings are intentionally close enough that either choice can be defensible.

Movement guardrails keep the result realistic

Personalization is capped more tightly near the top of the board and more loosely deeper in the rankings. The purpose is to reorder close values, not to turn a player ranked 50th by True GIQ into a top-five asset because the user repeatedly clicked one archetype.

Why stars are not training data

Saving a player to My Board means “remember this player.” Choosing him over a similarly valued player in Quick Pick means “this preference tells GIQ something about how I draft.” Keeping those signals separate prevents bookmarking behavior from contaminating the learning model.

What personalization cannot override

Quick Pick is deliberately prevented from becoming a second ranking engine. The strongest movement limits apply near the top of True GIQ, where large jumps would imply a meaningful difference in player value rather than a personal tiebreaker. Deeper on the board, the system can move players farther because the underlying values are closer and draft style can reasonably matter more.

This also protects the usefulness of the research layer. A user can disagree with GIQ and build a board that reflects those preferences while still seeing the original certified rank. The model's opinion remains auditable, the user's opinion remains personal, and the two can be compared instead of being blended into one unexplained score.