How to Turn AI Into Your Personal Sports Debate League

Using AI as a role playing mini-game can be really fun.

Sometimes while I fall asleep I’ll play games with AI to have some fun. Have AI pit warriors from all time against each other in hypothetical battle you control. See which cinematic universe would have the best 5 on 5 team squad. Naturally, sports matchups are next in line and I hadn’t done it before.

Two hours later, I had three different AIs arguing about whether Wilt Chamberlain could guard a stretch five, whether Michael Jordan or LeBron James deserved tournament MVP, and whether the Curry-Wilt pick and roll was — and I’m quoting Gemini here — “mathematically the most efficient play ever conceived.”

This article goes through my experience and gives you an idea of the structure to make this kind of conversation and game happen for yourself.

The Setup: Building Your All-Time Squad

Here’s exactly what I typed into ChatGPT:

I want to play a thought game — building the top 5 NBA team of all time. Help me out by going position 1-5, we’ll do one at a time. For each position, list the position, its general role in the game, and then the top 10 players of all time for that position. I’ll select and then we’ll move to the next one.

Nothing fancy. No prompt engineering tricks. Just a clear ask that gives the AI enough structure to make it a real back-and-forth. You pick, it moves on, and by the end you’ve got a roster you actually care about — because you built it yourself.

Position by position, I built my squad:

PG: Stephen Curry
SG: Michael Jordan
SF: LeBron James
PF: Victor Wembanyama
C: Wilt Chamberlain

Yeah, I picked Wembanyama at power forward. He hasn’t won anything yet. I don’t care — this is a fantasy and I wanted to see what AI would do with a 7’4″ unicorn who can shoot threes and block everything.

That’s the whole point. There are no wrong answers. Just interesting ones.

The Twist: Make the AI Pick Against You — Blind

Here’s where it gets interesting. Once I had my team, I asked ChatGPT to build two rival teams without seeing my roster. I just said:

Now build two more all-time NBA teams. Make them competitive but different from each other in style and era. Don’t overlap with any players I already picked.

ChatGPT came back with:

Team 2: Magic Johnson, Kobe Bryant, Julius Erving, Tim Duncan, Kareem Abdul-Jabbar
Team 3: Nikola Jokic, Kawhi Leonard, Larry Bird, Kevin Durant, Hakeem Olajuwon

Team 2 is a classic powerhouse — post play, transition, Showtime-era DNA. Team 3 is a spacing nightmare with five guys who can all shoot and pass. Both rosters made me immediately want to see what would happen in a series.

Which is exactly what the next step is for.

Hand It Off: Let Other AIs Simulate the Games

This is the part most people don’t think to do. Instead of asking the same AI that built the teams to also judge the winner — which felt like asking the chef to review their own restaurant — I copied all three rosters into a simulation prompt and sent it to two completely different AIs: Gemini and Grok.

Here’s the simulation prompt I used:

You are a basketball simulation engine. Simulate a round-robin tournament between three NBA superteams composed of players at their absolute peak. Assume modern NBA rules (2024 era). Teams play best-of-7 series. Consider spacing, defense, playmaking, coaching adjustments, and realistic matchups. After each series: explain key matchups, strategic adjustments, provide the final series result, and average stats for major players.

Then I pasted in the three rosters.

Both AIs went deep. I’m talking matchup analysis, tactical adjustments mid-series, stat lines — the whole thing. It was like reading two completely different ESPN analysts covering the same playoff bracket.

Where Gemini and Grok Disagreed (This Is the Fun Part)

Both simulations agreed on the big picture: my team won the tournament, and Team 2 (the classic post-up squad) finished last. Modern spacing beats old-school interior play — not a controversial take.

But the details? Wildly different.

The Team 1 vs. Team 3 series:
Gemini had it going to a dramatic Game 7, with Team 1 barely surviving on transition offense in the final minutes. Grok called it 4-2 — a more comfortable win where Wilt’s rebounding overwhelmed Hakeem.

Same rosters. Same rules. Different basketball brains.

Tournament MVP:
Gemini picked LeBron James, calling him the “primary processor” who made everyone else’s shots easier. Grok picked Michael Jordan, pointing to his scoring efficiency and clutch elimination-game performances.

That’s a genuine basketball debate. And two AIs had it without either one being wrong.

The moment that stuck with me:
Gemini identified the Curry-Wilt pick and roll as the single most unstoppable play in the tournament. Its logic: if you switch, a guard is stuck on Wilt (immediate lob). If you stay home, Curry has an open three. If you trap Curry, Wilt rolls to the rim and kicks to a wide-open Jordan or LeBron. Gemini called it “mathematically the most efficient play ever conceived.”

Grok never mentioned it specifically. Instead, Grok focused on Team 1’s defensive versatility — Wembanyama and LeBron switching onto guards while Wilt protects the rim. Same team, completely different reason for why they win.

Why This Actually Works as Entertainment

You might be thinking: it’s just an AI making up fake stats. And you’re right — none of this is real simulation with actual physics or play-by-play engines. The AI is reasoning through matchups based on what it knows about these players, then constructing a plausible narrative.

But that’s exactly what sports debates are. Nobody can prove that prime Wilt would beat prime Hakeem. The fun is in the argument — the reasoning, the “what ifs,” the moments where someone makes a point you hadn’t considered.

AI is genuinely good at this because:

• It knows these players’ careers in detail
• It can reason through matchups and spacing
• It generates specific, arguable takes (not wishy-washy “both teams are great” answers)
• It produces different results every time, even with the same prompt

Running the same rosters through multiple AIs turns a solo thought experiment into a panel of disagreeing analysts. And disagreement is where sports conversations get good.

Other Games Worth Trying

Once you get the hang of the basic team builder, there are a bunch of variations that add constraints — and constraints are what make games fun.

The $15 Draft Challenge

Players get assigned to price tiers. You have exactly $15 to build a starting five.

You build a team. Then ask the AI to build one against you with the same budget. Run the simulation. The salary cap forces creative roster construction — you can’t just take the five best players.

Ringless Legends Draft

Only players who never won a championship. Charles Barkley, Karl Malone, Allen Iverson, Patrick Ewing, Steve Nash, Tracy McGrady — suddenly you’re building a team from basketball’s greatest “what if” careers. The AI’s analysis of why these players never won adds a layer of real basketball history to the game.

One Player Per Decade

Pick one player from each era: 1970s, 1980s, 1990s, 2000s, 2010s-present. This forces weird, fun combinations. A team with Kareem, Magic, Hakeem, Kobe, and Giannis plays nothing like a team with Dr. J, Bird, Jordan, Duncan, and Luka. The era-mixing is where the AI really shines, because it has to figure out how players from completely different styles of basketball would fit together.

The Prompts (Copy and Use These)

All-Time Team Builder:

I want to play a thought game — building the top 5 NBA team of all time. Help me out by going position 1-5, we’ll do one at a time. For each position, list the position, its general role in the game, and then the top 10 players of all time for that position. I’ll select and then we’ll move to the next one.

Blind Rival Builder:

Now build two more all-time NBA teams that could compete with mine. Make them different in style and era. Don’t use any players I already picked.

Simulation Prompt (send to a different AI):

You are a basketball simulation engine. Simulate a round-robin tournament between these three NBA superteams at their absolute peak. Assume modern NBA rules. Play best-of-7 series. Consider spacing, defense, playmaking, and coaching adjustments. After each series: explain key matchups, strategic adjustments, final series result, and average stats for major players. After all series: rank teams by record, identify the champion, explain why they won, name the tournament MVP, and identify the most unstoppable lineup combination.

The Real Game Is the Argument

AI gets marketed as a productivity tool. Summarize this document. Write this email. Automate this workflow.

But some of the best hours I’ve spent with AI had nothing to do with work. They were spent arguing about whether LeBron or Jordan deserved MVP of a tournament that doesn’t exist, between teams that never played, judged by artificial intelligences that couldn’t even agree with each other.

That’s not productivity. That’s a Tuesday night well spent.

If you’re an NBA fan with access to ChatGPT, Claude, Gemini, or Grok — you already have everything you need. Build a team. Let the AI build a rival. Hand it off to another model and see what happens.

The best sports debates aren’t the ones with right answers. They’re the ones you can’t stop thinking about.