What this
is

The game

A legendary team-season lands on your draft board. You pick one player and take one of his ratings — your choice, not the best one automatically. Do that 13 times and you have filled all 13 attributes. Then the build plays an 82-game season and you find out what it was actually worth.

The format was originated by Build-A-Bucket from @Build_A_Player. This is our own take on it, with our own engine.

Where the ratings come from

We made them up. They are our own editorial estimates, informed by how these players are remembered and what they actually did, but they are not official and they are not from any licensed product. Nobody publishes a canonical "1971-72 Jerry West had 99 clutch", so treat every number here as an opinion with a method behind it.

The method: each player is authored once at his peak season across 13 attributes, then resolved per season through an age curve. Athletic traits (ATH, DNK, BLK, FIN) fall away hard on both sides of the peak; feel traits (PAS, CLU) hold or improve. That is why the same player reads differently in 1988 and in 1996. There are 99 team-seasons in the pool.

How the simulation works

It is not a random number with a nice animation. Each attribute contributes to net rating through a weight that depends on the position you were dealt, and the season report shows you that contribution attribute by attribute — what carried you, what cost you, and which single pick did the most damage.

Overall is a weighted sum, not an average. A 96 three-point rating is worth about +2.89 overall to a shooting guard and +0.89 to a centre. The full matrix is on the attributes pages.

Everything is deterministic from a seed. The same seed and the same picks always produce the same season, which is what makes a shared link honest — the card you send is a replay, not a screenshot.

Balance is measured, not guessed

The engine is calibrated by running 40,000 drafts and reading the distribution. That is where the rarity thresholds come from, and it is how we know the median build is 88 OVR and wins 41 games. Those figures are on the benchmarks page, read live from the calibration file rather than typed into the copy.

What is deliberately not here

• No player photos and no uniforms. Team marks identify the team on the draft board; player names are plain text.
• No signup to play. Every mode, including the daily board, works signed out. Signing in with Discord is optional and exists for one reason: to put a run on the leaderboard under a name.
• No currency, no codes, no redeem system — despite what a lot of sites claim.
• No Roblox version. It is a browser game.

Not affiliated

Not affiliated with, endorsed by, or connected to the NBA, any team, NBA 2K, or Take-Two Interactive. Player names are used descriptively to identify real performances. Team names, marks and colours are the property of their respective teams and are used here for identification only — no endorsement is implied or claimed. If you own rights to something here and want it removed, get in touch and it goes.

What is still missing

Being straight about it: there is no stored leaderboard yet, so results are not ranked against other players — the benchmarks page compares you against simulated runs instead. Doing it properly needs stored results and server-side verification, otherwise the top of the table is just whoever edits their own numbers first.