Round up your group, agree on which players and categories you're tracking, and lock in your predictions before the opening tip. That's the whole idea behind a prediction draft: everyone calls their shots before the game starts, then watches to see who called it best.
What is a prediction draft?
A prediction draft is a pre-game ritual where everyone in your group predicts how specific players will perform, then compares results after the final buzzer. Think of it like a snake draft, but instead of drafting players onto a fantasy roster for a season, you're drafting predictions for a single night. Will Tyrese Haliburton go over or under his assist number? Does Jalen Brunson hit a certain point total? You lock those calls in before the game tips, and nobody gets to change their pick once the ball is live.
The format works because it's fast. You don't need a full fantasy league or a season-long commitment. One group chat, one game, twenty minutes of picking, and you're set for the next three hours of basketball. It works for any group size too, from two roommates on a couch to a twelve-person group chat that's been arguing about the same five players for years.
How do you pick the categories?
Start with the stat lines that actually swing the outcome of the game you're watching, not just whatever numbers are easiest to track. For an NBA game, points, rebounds, and assists cover most of what people care about, but three-pointers made and turnovers can be just as fun for players who live and die by pace. If Anthony Edwards is playing a team that switches everything, his usage rate that night might matter more than his season average. Pick categories that fit the matchup, not just whatever showed up in last night's box score.
A good rule: three to five categories per player, and no more than three or four players total. Go much bigger than that and your draft turns into a spreadsheet exercise instead of something you can run before tip-off while everyone's still getting settled on the couch. Keep it simple enough that someone can join five minutes before the game starts and still have fun.
How far ahead of tip-off should you run it?
Aim for 15 to 30 minutes before tip-off, right around the time starting lineups get confirmed. That window matters because injury news and late scratches change everything. If a team announces a starter is out with a knee issue twenty minutes before the game, that changes who's getting extra minutes and extra shots for everyone left on the floor. Wait until lineups are locked and nobody's predictions get blown up by a last-minute scratch.
Some groups like to run it even earlier, right after the injury report drops in the afternoon, so people have time to think it over instead of rushing picks in a group chat five minutes before tip. Either works. The only rule that actually matters is that every prediction locks before the opening tip. No exceptions, no "I meant to say" after the first quarter starts.
How do you set the lines for each prediction?
Use recent, real numbers as your baseline, not a gut feeling pulled out of thin air. If you're predicting Nikola Jokic's assists for a given night, look at what he's actually been averaging over his last ten games, not his career average from three years ago. Player pages on Basketball-Reference break down recent game logs fast, so you can see exactly what a player's been doing lately before you set a number for the group to predict against.
This is also where implied probability comes in handy as a gut check. If a player's odds imply a roughly 55% chance of going over a certain rebound total, the number is close to a coin flip, which usually makes for the most interesting predictions. A number where everyone in the group is confident on the same side isn't much fun to predict. The best categories are the ones where your group actually splits.
How do you score it and keep things fair?
Score every prediction the same way, whether that's straight over/under hits or a points system that rewards being closer to the exact number. The simplest version: one point for every category you call correctly, tally it up after the final buzzer, and whoever has the most points wins bragging rights for the group chat until the next game. Some groups add bonus points for a "lock" pick, one category each player is extra confident about, which raises the stakes without complicating the whole system.
Comparing your calls against the field, not just your own gut, is the whole appeal. That's the same idea behind how GAGE scores predictions against the actual odds instead of just tracking who guessed the highest number. A prediction that beats a tough line means more than one that beats an easy one, and building that into your scoring makes the draft actually reward good reads instead of lucky guesses.
Keep a running tally across a few games if your group is competitive. A single night is fun, but a running leaderboard over a week of games is what turns a one-off draft into something your group actually looks forward to before every big matchup. It also gives you something to talk about during the games themselves, since everyone's watching for their own predictions to hit.
What's the easiest way to actually run one?
Skip the spreadsheet and group chat math and use an app built for exactly this. Download GAGE and you get real lines on real players, scored against implied probability, so you're not just guessing round numbers out of thin air. It turns the whole draft-before-tip-off idea into something you set up in a couple minutes instead of twenty.
If your group already does something like this over text, it's basically a single-game fantasy sports setup without the name. Real odds just make the results feel earned instead of arbitrary, and they give you a clean way to settle the "who actually knows ball" argument that never seems to end in any group chat.
Do you need a big group to run a prediction draft?
No, two people is plenty. A prediction draft scales down just as easily as it scales up, since the format is really just "everyone predicts before tip-off," and that works whether it's two roommates or a twelve-person group chat.
What if someone joins late, after the game already started?
They sit that one out and jump in for the next game. The whole point of locking predictions before tip-off is that nobody gets to react to what's already happened on the court, so a late joiner just waits for the next matchup.
Does this work for any sport, or just basketball?
It works best for sports with frequent, trackable individual stats, which is why basketball is such a natural fit. Points, rebounds, assists, and threes update possession by possession, giving your group plenty to argue about before the game even starts, which is what makes a group watch more fun in the first place.