How the NBA Draft Lottery Actually Works
The lottery system has been in place since 1985, and despite decades of familiarity, there are still plenty of people who misunderstand what the odds actually mean. I spent years building projection models around the lottery, and the first thing you need to understand is that it is not a random drawing in the way most people imagine. The process is heavily weighted toward worst records, yes, but the distribution is not linear. It is deliberately shaped to balance competitive integrity with the incentive structure of losing well. Before diving into the mechanics, let me say this plainly: betting markets and fan discussion consistently overvalue the team with the worst record. The Boston Celtics at 10.5% for the top pick do not stand a 50-50 shot against Detroit's 14% just because the gap looks small on paper. The difference between the worst team and the second-worst team in the current setup is more meaningful than the headline numbers suggest. The lottery weights are designed so that the gap from first to second is larger than the gap from second to third, and that matters when you are pricing risk. The process starts with 14 ping pong balls numbered one through fourteen. Four balls are drawn at random. The combination determines the pick. Since order does not matter, there are 1,001 possible combinations. The team with the worst record gets 250 of those combinations. Detroit at 14% roughly corresponds to that proportional allocation. The Celtics at 10.5% get proportionally fewer. The remaining teams outside the lottery receive their picks in reverse order of record. This is the basic architecture.
I run these numbers constantly, and the thing I see most often is people misreading the implied probability as a simple percentage of possession. It is not. Each team gets a fixed number of combinations, but the drawing mechanism is combinatorial. That distinction matters when you are building a simulation. If you treat each team's chance as a flat probability rather than a combinatorial weight, your model drifts quickly out of alignment with reality. The real work happens after the lottery. The top three picks are determined by the lottery draw itself. Picks four through fourteen are assigned by record among the non-lottery teams. Then the regular draft follows. The NBA has experimented with changes over the years, including modifications to the weight distribution in 2019, but the core mechanism has remained stable. The league wanted to reduce tanking incentives while preserving a meaningful advantage for the worst teams. The result is a system that feels fair but operates on mathematical principles that reward careful study. When I help front offices or serious analysts work through the lottery, the first thing we do is build a Monte Carlo simulator. You run it 100,000 times. The output should closely match the published odds. If it does not, you have a bug in your combination generator or your sampling method. A common error is treating the draw as sequential without replacement but then allowing duplicate combinations across iterations. That skews the tail probabilities noticeably. I caught this myself early in my career. It cost me an afternoon and a credibility hit with a colleague who reviewed my code.
The workaround is straightforward. Generate all 1,001 combinations upfront. Assign each combination to its weighted pool. Sample with replacement from that pool, not from the balls themselves. This ensures the combinatorial weights are exact and eliminates any edge-case drift. It runs in under two seconds on a modern laptop. You should not need a cluster for this. One counter-intuitive point that beginners miss is that having the worst record does not guarantee the best player. The lottery is a random event. The 2019 draft is the textbook example. The Lakers had the worst record and won the top pick. The Suns held the second-worst odds and landed the fourth pick. The Kings won the third pick. If you were pricing this ahead of time, you needed to account for the variance, not just the expected value. A model that treats worst-record teams as automatic top-two locks is fundamentally broken. Another nuance involves the second lottery draw. After the first night, the remaining combinations are redistributed among the teams still in the lottery. This means the odds shift slightly between the first and second drawing nights. The league publishes these adjustments, but they are small. The real shift comes from how teams position themselves for the second round if they miss the top three on the first try. A team with a mid-lottery chance that loses the first draw still faces a recalculated probability landscape. Most public discussion ignores this entirely.
Get the Full Details
Here is the blunt truth about limitations. The lottery system cannot eliminate tanking. It reduces extreme incentive, but a team with the worst record still gets a dramatically better shot than a team with a 40-win record. That gap is large enough to create strategic behavior. Some teams will still make roster choices that tilt toward long-term asset accumulation. The league has tried to mitigate this with playoff revenue sharing and competitive balance mechanisms, but the fundamental tension remains. No weighting scheme solves it completely. For practical analysis, I recommend combining lottery simulation with a draft value model. The lottery gives you the distribution of likely picks. A draft value curve, like the one used by Jon Rothstein or the one adapted from ESPN's draft analysis framework, tells you what each pick is worth. Multiply the two and you get expected value for trades, roster construction decisions, and free agency planning. This approach cuts analysis time from hours to minutes once your calculator is built. Before I had a working script, I spent about forty minutes per lottery simulation by hand. That is not sustainable. If you want to build your own simulator, start with a simple Python or R script. Use the official NBA lottery combination table as your source of truth. There are public datasets available that list every team's allocation. Do not guess at the weights. The NBA publishes them, and they change slightly between years. The 2023 modification is a good example where the top three probabilities shifted by fractions of a percent. Small changes, but enough to matter if you are building a pricing model.
The broader takeaway is that the lottery is a tool, not a crystal ball. It provides a probabilistic framework for understanding where your team might land. It does not tell you who you will draft. That requires scouting, evaluation, and a willingness to accept that randomness will play a role regardless of how well you prepare. I have seen too many analysts treat the lottery as destiny. It is not. It is a structured randomizer with edges, and those edges are real but limited.
