← Studio

Collect LineupsDeep dive

Planning Swaps by Tip-Off

collectlineups.com ↗

Problem
Weekly Tip-Off runs Monday to Sunday with unlimited swaps. You fill ten slots with any cards, the same player in several if you have the copies. Games tip off in waves, often every half hour, so the best lineup at 6:00 is not the best at 8:30. Projection sites rank whole games. The contest asks something else: which of your cards earns the most for the minutes left tonight. My sons and I worked it out in our heads. Fill all ten with the first tip’s players, then at each later tip swap in anyone whose full game beat what was left of our weakest card. And the cards we owned changed every day.
Constraints
The site can’t touch Topps, so every swap is one a person makes by hand, in the app, at the moment it pays off. A plan had to be a short list of moves tied to a time, not a projected total. Projections are whole-game totals, with nothing about when in the game the points come. And the plan is only as good as the card counts behind it, so keeping them current had to take seconds.
Process
The first planner followed our rule: at each tip, bring in a card when its whole game beats what’s left of the weakest card, if it adds at least 10 points. Working through an example showed me it was wrong. With ten Jokić cards at 6:00 and a better player tipping at 6:30, it swapped Jokić out at 6:30. Keeping him for his whole game and bringing the other player in when Jokić’s game ended earned more. So the planner moved to points per minute. Each projection is spread over its game, from about ten minutes after the listed tip to about two and a half hours after it. The lineup keeps the cards earning the most per minute, and it re-checks at every tip and every estimated game end. Game ends became their own stops in the strip of times at the top, so a swap shows at the time you would make it in Topps. Later I dropped the 10-point minimum, since swaps are unlimited and a spare 1.7× Flagg was going unsuggested. Dropping it alone made some nights worse, so the planner now builds three plans and keeps the one with the most projected points, fewer swaps on a tie. For the cards, tapping a boost adds a copy with Undo, a lineup can hold a card you haven’t counted yet, and an AI app you already use can read screenshots of your collection into an import.
Outcome
Weekly opens on a strip of the night’s times: each tip, and each game end with something to do. Pick one and it shows the lineup at that moment, the swaps to make, and what each one is projected to add. A star plays his whole game, and later players come in when his game ends. Across 3,000 simulated nights, the plan it keeps was never projected lower than the plan with the 10-point minimum, and was higher on about half. It is an estimate, not a live score: overtime, blowouts, and how a player’s minutes fall aren’t known, and the page says so.
Lessons
The rule we used in our heads was the first thing I built, and it was wrong in a way I only saw once I worked through an example. Most of what changed after that came from using it during games. The game-end stops came from a night when a swap was due at 7:30 and the strip only had 7:00 and 8:00. The minimum went when the planner wouldn’t suggest a spare Flagg card projected to earn about 6 points more.