Why NBA teams lose on purpose — and what could actually stop them

Every spring, teams at the bottom of the standings quietly start resting their best players, losing close games, and chasing bad draft odds. This isn't an accident. It's game theory.

We'll walk through the math behind tanking: what makes it individually rational, why the 2019 lottery reform didn't fix it, and why a system borrowed from women's hockey might be the most elegant solution nobody's talking about.

Each section has live sliders and simulations. Play with them. The argument is in the numbers.

Losing is worth a lot of money

Before we add rivals, imagine you're a GM of a lottery team with no realistic shot at the playoffs. You have one decision: field your best roster, or quietly lose games to improve your draft odds.

The problem is the relative size of these terms. Revenue sharing means the difference in income between a 20-win team and a 36-win team is surprisingly small. But the expected value of a high draft pick, especially in a superstar year, is enormous. And critically, the value isn't primarily in year-over-year revenue. It's in franchise valuation.

Even at equal weights, tanking comes out ahead for a bottom-10 team. The revenue delta is small; the draft value delta is large. This is the foundation of everything that follows.

When everyone tanks, everyone loses

Now add a rival team. You're both rational. You both see the same incentives. What happens?

This is a textbook prisoner's dilemma. The Nash equilibrium is (Tank, Tank), every time. The 2019 lottery reform attempted to compress odds and push teams toward a "coordination zone." See how well it works:

Note: not every team tanked pre-2019 even when that was the dominant strategy. Some franchises prioritized fan experience, coaching development, or simply disagreed on the math. The model shows the rational incentive structure; reality is messier. That messiness is actually what prevents total league collapse under the current system.

Now watch all 30 teams evolve over 82-game seasons. Each team updates its strategy based on the payoffs it observes. The two panels run simultaneously under pre-2019 and post-2019 rules using the α/β values you set above. Hit Auto-run.

Under pre-2019 rules the league almost always converges to near-universal tanking within a few seasons. Post-2019 can create a mixed stable state, but only when the ratio α/β stays below roughly 3. When draft value weight significantly outweighs revenue weight, which is the realistic case for small-market teams in a superstar year, watch the reform fail. The 2019 reform assumed teams are sensitive to marginal odds differences. In a superstar draft year, they aren't.

A useful thought experiment: what if draft position had nothing to do with record, a pure lottery where every team had equal odds? Tanking would disappear entirely. If you can't improve your odds by losing, there's no incentive to lose. You still pay the revenue cost of a bad season; you just stop getting the compensating draft benefit. That baseline matters because it isolates where the problem actually lives, not in the draft itself, but in the specific link between losing and lottery position. The hockey model attacks that link directly. Compressing the odds, as the 2019 reform did, merely weakens it.

When the prize gets bigger, the problem gets worse

Tanking pressure doesn't stop at the worst records. It cascades up the standings, all the way to teams fighting for the final playoff seed.

The play-in introduced a wrinkle the 2019 reform didn't anticipate. A team on the playoff bubble faces a real calculation: fight for a play-in spot that likely ends in early elimination with no lottery ticket, or miss the playoffs entirely and land in the lottery. In a superstar draft year, the lottery upside of missing the playoffs can outweigh the playoff upside of the play-in, especially for a team with no realistic path to advance. The play-in was designed to keep more teams relevant. It accidentally created a zone where falling short of it is strategically attractive.

But the play-in is a symptom of a deeper problem. Recall the franchise value swings from Stage 1: LeBron's return to Cleveland added $810M in four years. A single generational player doesn't just improve a team. It transforms the entire financial trajectory of a franchise. And unlike large-market teams, small-market franchises can't attract that player through free agency; the draft is their only realistic shot at one. So as the perceived value of any given draft class rises, the incentive to tank doesn't just persist, it intensifies. The basic two-term utility function doesn't capture this. We need a third term:

The γ · S term captures something the basic model misses entirely: the one-time, non-recurring franchise value explosion that comes from landing a generational player. This isn't revenue; it's the difference between a franchise worth $800M and one worth $3B. It's local media rights, jersey sales, national TV appearances, and sponsorship premiums that compound for years. For small-market teams with no free-agency appeal, the draft isn't one pathway to a superstar; it's the only one, which makes γ enormous and essentially permanent if you hit.

Curry was the 7th pick. Giannis was 15th. Neither was seen as a likely franchise cornerstone at the time, but the team picking 8th couldn't draft Curry, and the team picking 16th couldn't draft Giannis. Every lottery spot is a potential miss at a player who could have been yours. That's why even marginal improvements in odds, the difference between 9th and 12th in the lottery, get treated as high-stakes decisions.

Now watch the core dynamic across all three systems. Key experiment: switch to Superstar year and raise γ to 2–3. Under pre-2019 and post-2019 NBA rules that produces near-universal tanking. Under the hockey model the same γ reinforces competing, because superstar value now flows through winning games on the elimination table, not losing them.

Women's hockey figured this out already

Once you're mathematically eliminated from the playoffs, you start accumulating points on a separate league table. The team with the most points gets the #1 pick. Every win after elimination directly improves your draft position.

This completely inverts the incentive for eliminated teams. And worse teams get naturally biased toward better picks, not through lottery odds, but through geometry. A team eliminated in February gets ~25–30 games on the table. A team that just misses the play-in gets maybe 5. That gap is the mechanism. No randomness required.

For teams on the playoff bubble, the hockey model doesn't dramatically change the strategic calculus: whether they fight for the play-in or fall short of it, they accumulate zero table games either way. What changes is the upside. Under the current system, missing the play-in in a superstar draft year still carries a lottery ticket and a real chance at the #1 pick. Under the hockey model, that jackpot doesn't exist. There's no consolation prize for a near-.500 team. The only path to a top pick is being eliminated early and then winning games.

The closer the tiers are in elimination timing, the flatter the effective odds. The more spread out, the stronger the bias toward worse teams:

Notice what happens when you push all three tiers to nearly the same elimination date: the odds flatten toward equal. Pull them apart and the early lottery tier's advantage compounds naturally. The system is self-correcting.

One implementation challenge worth noting: in a 30-team league, schedule difficulty varies considerably across the remaining games of an eliminated team. A team eliminated with 24 games left might face 22 playoff-caliber opponents mostly on the road; another eliminated with 4 games left might face the weakest teams in the league at home. Remaining schedule strength isn't random, and in a system where post-elimination wins directly determine draft position, that asymmetry matters. Any serious implementation would need to either randomize end-of-season scheduling for eliminated teams or publish remaining schedules transparently, so the process can't be quietly tilted in any direction.

The last three drafts, rewritten

If the hockey model had been in place for the 2023, 2024, and 2025 drafts, who would have gotten which pick?

This is a counterfactual. Teams would have behaved differently under different rules, so the actual elimination dates and win totals would shift. But applying the hockey model retroactively to real seasons gives an honest sense of how the pick order changes, and whether the system actually biases top picks toward the worst teams the way it's designed to.

Play-in teams are treated as eliminated on the date of their play-in loss with zero remaining games.

A few things jump out across all three years. Post-elimination winning performance, not regular-season record, is what drives draft position under the hockey model. San Antonio accumulated 9 table wins in 2024 with only 22 regular-season wins and would have picked first. Houston led with 7 table wins in 2023 and would have done the same. Detroit had the worst record in both of those years and lands mid-table both times, because being eliminated early only creates the opportunity to accumulate table wins; you still have to take it.

The zero table wins cases are the starkest. Any team that finishes the season without accumulating a single post-elimination win, whether they narrowly missed the play-in or missed the playoffs entirely, falls to the bottom of the draft table regardless of record. In 2025, Dallas missed the playoffs at 39 wins and won the lottery outright, getting the #1 pick. Under the hockey model, zero table wins guarantees a team will not be picking anywhere near the top of the draft.