The NBA Transfer Market: When Even Official Data Sources Get It Wrong

On the night of February 8, 2026, I sat in the twelfth row of Cameron Indoor Stadium with a notebook and a pencil worn down to its tip. Duke against Virginia Tech. I had been assigned to track the rebounds of Zion Williamson — a 129-kilogram forward the entire NCAA was talking about. When the final buzzer sounded, the official organizer's data board lit up on the big screen: Zion had 9 rebounds. I wrote the number 9 in my notebook, filed my piece, and went home. That night, rewinding the tape to write a description of one possession, I counted a different number. I rewound a second time. Then a third. Then a fourth. My number was 11.\n\nI once counted the tape four times, and the error belonged to the source, not to me. The correction I wrote afterward had only 240 readers, but an editor at The Ringer shared it, and that is why I am here, ten years later, still rewinding tape and still counting. That small story is not meant as bragging. It is the foundation for how I read the NBA transfer market today: a market in which thousands of numbers are transmitted every day, most of which nobody verifies, and a significant share of which are wrong.\n\nThere is another night I remember just as well. It was a night during the transfer window, as I prepared a podcast episode about a hot deal. I opened my advanced-data board, and the board was empty. Not empty because the player did not play — empty because the data feed had gone down for a few hours. During that window, at least four social media accounts posted numbers about that player, with different values, and each post claimed to be "per the stats source." When the board came back, I cross-checked: only one of the four numbers was correct.\n\nThat is why I am writing this piece. Not to tell the story of one specific deal, but to offer a filter. The transfer market is where noise overwhelms signal more clearly than in any other phase of the season. And in a market like that, people often confuse data with truth.\n\n## Context: a market that lives on rumor\n\nEvery transfer window, three streams of information pour into one funnel. The first is official information: contracts announced, deals confirmed, team press releases. The second is leaked information: numbers that surface before any announcement, usually from an agent or from internal departments. The third is inferred information: analyses built on the first two, sometimes with no clear source.\n\nThe problem is that, in an era when speed is placed ahead of accuracy, the second and third streams usually spread faster than the first. A leaked number posted at 11 p.m. will draw hundreds of thousands of views before the team confirms it at 9 a.m. the next morning. And when the team confirms it, most of the audience has already memorized the first number — even if it was wrong.\n\nI have held one professional rule for a long time: player agents are the largest hidden cost of this market. Not because they are bad people, but because their job is to optimize their client's interests, and their tool is information. A number released at the right moment can push a contract's value up by millions of dollars. A rumor planted in the right place can pressure a team into acting. Noise, in this case, is not a byproduct — it is the main product.\n\nI am not writing this to accuse any individual. I am writing because I have seen that mechanism operate, and because I believe readers deserve a better filter than "I heard that."\n\n## The core: four layers of verification for one number\n\nLet me start with a concrete example, not to criticize anyone, but to illustrate the method. During the transfer window, a player was said to have an impressive defensive record, and the most-quoted figure was "opponents score only 0.95 points per possession when he is dragged into a pick-and-roll." That number appeared on at least six different platforms within a week. I decided to verify it.\n\nThe first layer of verification is origin. Where did this number come from? It turned out to come from an internal team dataset that had leaked. That dataset covered only 41 possessions. Forty-one possessions is not a large enough sample to conclude anything about defensive ability. In a season, a player is dragged into pick-and-rolls several hundred times. Taking 41 of those and publishing them as a stable metric is a methodological error, regardless of whether the number itself is correct.\n\nThe second layer is definition. What does "dragged into a pick-and-roll" mean? Does it include times when that player is pulled out to the three-point line and forced to change direction? Does it include situations where a teammate helped on defense and the player merely stood waiting? If the definition changes, the number changes. And in most cases, the definition is not published alongside the number.\n\nThe third layer is film. This is the layer I trust most and the one that costs the most time. I rewound every possession supposedly in the sample. I counted. And I found two things. First, four possessions were miscounted — they were plays in which the player did not participate in defense at all. Second, among the remaining possessions, several were plays in which his teammate was the one beaten, yet the points were credited to this player.\n\nThe fourth layer is cross-checking against an independent source. I pulled data from a second provider using a slightly different definition. The number I got was 1.12 points per possession, not 0.95. The gap between the two numbers is not a small margin of error. It is enough to shift an assessment of a player from "good defender" to "average defender."\n\nThe conclusion from this example is simple: a number without an accompanying source, definition, and sample is not data, but an opinion wearing a number's jersey.\n\nA rebound the organization recorded incorrectly still counts — if you are willing to rewind the tape. That principle applies not only to one possession in one Zion game. It applies to every number transmitted during the transfer window.\n\n## Follow the money, follow the story\n\nDuring the transfer window, the most effective way to read the market is not to read rumors, but to read contract structure. A contract has four elements that outsiders usually overlook but that determine real value: base value, term, performance bonuses, and team protections.\n\nWhen a team announces a contract, the most-cited number is the total value. But total value is not the most important number. The most important number is the structure of the clauses. A "200 million dollars over four years" deal may in reality guarantee only 120 million, with the rest contingent on the player hitting performance thresholds or making an All-League team. Most coverage will cite only the 200 million.\n\nI have spent years tracking contracts and learning to read the fine print. When I see a team willing to include an injury-protection clause, that is a signal. It tells me how the team assesses the player's physical risk. When I see a games-played bonus, that is another signal. It tells me both sides are negotiating on a basis of mutual doubt about availability.\n\nThose signals matter more than any rumor, because they are the product of a real negotiation, with real money, real legal force, and real consequences if either side breaches. Rumors have no consequences. Contracts do.\n\nThat is why I always tell my podcast listeners: when you see a rumor about a deal, ask yourself who benefits if that rumor spreads. If the answer is the player's agent, that rumor needs double verification. If the answer is a team trying to sell a different player, that rumor also needs double verification.\n\n## Systemic error and the lesson of transition defense\n\nIn February 2026, I produced an investigative podcast series on a nine-game losing streak by the New York Liberty women's basketball team. It was a stretch in which the team was heavily criticized, and most of the criticism was aimed at the head coach. I decided not to start from emotion, but from data.\n\nI used data from Second Spectrum and found a clear pattern. Rookie center Han Xu was exploited an average of 14 times per game in pick-and-roll situations, and opponents scored an average of 1.17 points per successful exploitation. That number was not Han Xu's problem alone. It was a problem of the transition defense system. The team was asking a young center with limited lateral movement to defend in areas where she had no advantage.\n\nWhen I published that podcast series, head coach Sandy Brondello declined an interview request. I was not surprised. But three weeks later, the team changed its scheme: Han Xu was kept closer to the rim, and the guards had to work harder on the perimeter. That series drew 80,000 listens, five times a normal episode.\n\nWhat I learned from that experience is not "I was right." What I learned is: an analysis is only valuable when it points to a mechanism, not merely a phenomenon. If I had only said "Han Xu defends poorly," I would have said nothing at all. But when I showed that the system was systematically placing her in a disadvantageous position, that became actionable information.\n\nI also learned something else: never forget to credit the analytics assistants. They are the ones who provide the underlying data. In this industry, sources are assets, and the only way to keep a source is to never steal another person's credit.\n\n## The counterintuitive angle: official stat sheets are not perfect either\n\nThis is the section I want to dedicate to those who believe that using "official data" is enough.\n\nI once had an experience that completely changed my view on that. In 2026, during the World Cup in Russia, I was an intern at a local radio station in New York. I was assigned to analyze the defensive tactics of the Croatia national team. I rewatched all seven of their matches and tallied that Ivan Perišić ran an average of 12.3 kilometers per match.\n\nBut when I dug deeper, I found something more interesting: only 31 percent of that distance was directed toward the opponent's goal. The rest was lateral running, backward running, positioning. I wrote a 19-page internal memo emphasizing this imbalance. The editor did not use it, calling it too dry. After Croatia reached the final, he admitted my read was correct.\n\nCroatia was not the team that ran the most — it was the team that ran in the right direction most. That 31 percent figure for distance toward the opponent's goal is the number I wanted to talk about. But let me be clearer: that 31 percent was also a number I calculated myself, and I know it carries error. My definition of "toward the opponent's goal" is not the same as any data provider's definition. That is its weakness.\n\nI wrote 19 pages only to extract one sentence worth saying. But the real lesson lay elsewhere: even when I verify by hand, my number can still be challenged on its definition. No number is entirely neutral. Every number carries the assumptions of the person who created it.\n\nThat is why, when I read an official stat sheet, I do not read it as a fact, but as a claim. A claim can be right, can be wrong, can be right in one context and wrong in another. My job is not to believe, but to check.\n\n## The 2.8 percent number and the limits of small samples\n\nIn 2026, when leagues shut down because of the pandemic, I defended my master's thesis on the effect of empty arenas on free-throw efficiency. I collected data from 612 NBA games from March to October. The result I found: the free-throw percentage of young players under 25 fell an average of 2.8 percent when there was no crowd pressure. Meanwhile, the EuroLeague showed no significant change.\n\nMy thesis was rejected by the committee for too small a sample. They were right. I did not argue. But I used it as the foundation for the first episode of my own podcast, and I always state the limits of my data in every episode.\n\nA thesis being challenged is fine; the data does not argue back. What I want to say here is: even a study with serious methodology can be challenged on sample size. So how credible is a number leaked from an internal dataset with 41 possessions, spread across six platforms?\n\nWhen the crowd disappears, young free-throw shooting disappears with it — unless you are in the EuroLeague. That sentence sounds decisive, but it is only true within the limits of a 612-game sample. I do not know what would happen with a larger sample. And I say so clearly.\n\n## Load management and the art of reading minutes\n\nDuring the transfer window, there is one type of data I always look at first: minutes played and rest patterns.\n\nA player averaging 34 minutes a game during the season is usually not a problem. But how those minutes are distributed is a problem. If a player plays 12 straight minutes in the first half and 12 straight minutes in the fourth quarter, his body bears a different load than if he plays 34 minutes spread evenly. Modern teams track this with motion sensors, but most fans only see the total.\n\nDuring one transfer window, I analyzed a deal in which a team acquired a player with an injury history. The most-cited figure was "he played 68 of 82 games last season." That sounds fine. But when I looked at the minute distribution, I found that over the final 14 games, he averaged only 18 minutes per game, down from 31 minutes before that. That was a signal about physical condition that the "68 games" figure concealed.\n\nThis is the biggest blind spot of the transfer market: people read games, not minutes; people read scoring average, not scoring distribution. And in a market where one percentage point of efficiency can be worth millions of dollars, those blind spots are extremely valuable.\n\n## Agent noise and how to spot it\n\nI want to say plainly something I have learned over many years: most rumors in the transfer window are not intended to inform, but to apply pressure.\n\nThere is a pattern I recognize. When an agent wants to renegotiate a contract for his client, he often plants a rumor about another team's interest. That rumor does not need to be true. It only needs to spread widely enough to create a threat. The player's current team, under fan pressure, may act.\n\nSpotting these rumors is not hard, if you pay attention to detail. Real rumors usually name the specific parties involved, contain a preliminary contract structure, and carry a timeline. Planted rumors tend to be vague about structure, lack a timeline, and appear exactly when the player is negotiating.\n\nI am not saying every rumor is false. I am saying every rumor needs to be placed in the context of the speaker's motive. When you read a rumor, ask yourself: who wants me to believe this, and why now?\n\n## Four questions to filter a rumor\n\nAfter years of work, I have distilled four questions I ask myself before putting any information into my podcast.\n\nThe first question: who is the origin of this information? If the answer is "a source close to the situation," that information is not yet good enough for me to use. If the answer is a reputable reporter with a track record of accuracy, I start to consider it.\n\nThe second question: does the accompanying number have an origin? If there is a number but nobody says where it came from, I treat that number as nonexistent.\n\nThe third question: what is the motive of the person sharing it? This is the hardest question, because motives are usually hidden. But there is one clue: if the information is released at a moment that benefits one specific party, that party's motive needs to be examined.\n\nThe fourth question: can I verify it independently? If I cannot verify it with a second source, I do not use it. This principle sometimes makes me slower than other colleagues, but it keeps my podcast from having to issue corrections.\n\n## When data does not exist: the lesson of an empty board\n\nI want to return to the story at the start of this piece. That night, my data board was empty, and during that window, four accounts posted four different numbers.\n\nWhat is notable is not that someone posted a wrong number. What is notable is that none of those four accounts said they had no data. When data does not exist, some people choose silence, but others choose to fill the gap with guesswork. And in the transfer market, guesswork spreads faster than truth.\n\nThis is a lesson about honesty in analysis. When I do not have data, I say I do not have data. That does not make me less credible. On the contrary, it makes me more credible, because listeners know that when I offer a number, that number has been verified.\n\nPeople see a mistake and laugh; I see a mistake and look for the source. That is the fundamental difference between a commentator and a verifier.\n\n## The rise of advanced metrics and the new trap\n\nOver the past decade, basketball advanced metrics have developed very quickly. We have True Shooting percentage, Estimated Plus-Minus, Usage Rate, and dozens of other indicators. This is a major step forward from an era when there were only points, rebounds, and assists.\n\nBut a new trap has come with that development: people have started to believe advanced metrics as if they were absolute truth. A high EPM does not automatically mean a player is good in every system. A high True Shooting percentage does not automatically mean a player will succeed on a different team.\n\nAdvanced metrics are built on models, and every model has assumptions. If the assumptions are wrong, the metric is wrong. During the transfer window, teams often buy players based on advanced metrics without checking whether the metric fits their system. That is a new, more sophisticated kind of error, but it is still an error.\n\nI am not against advanced metrics. I use them every day. But I always remember that a metric is a tool, not a verdict. And tools need to be checked before use.\n\n## The economics of the transfer window\n\nTo understand the transfer market, you need to understand one basic thing: this is a market with limited supply and unlimited demand. The number of players who reach a certain level in a season is very small. The number of teams that want them is every other team.\n\nWhen limited supply meets high demand, prices rise. But in basketball, price is not only money. Price is also opportunity, roster position, and developmental rights. One team may pay less money but promise a larger role. Another team may pay more money but promise only a small role.\n\nThis is why rumors about "Team A offering more than Team B" are usually incomplete. Money is only one part of the equation. The rest is role, championship opportunity, geography, family, and dozens of other factors that no stat sheet can measure.\n\nI learned this when I interviewed a former player about his free-agency decision. He told me he had turned down a higher-paying contract because that team had no clear plan for him. The number the public saw was the number he declined. But the number that truly mattered to him was the minutes he would get to play. And that number appeared in no transfer report.\n\n## The counterintuitive angle: the market is inefficient, and that is the opportunity\n\nThis is what I want to emphasize in this section: the NBA transfer market is inefficient. If it were efficient, a player's value would always reflect his true ability. In reality, it does not. Some players are paid above their true value, and some are paid below it.\n\nThis inefficiency comes from several sources. First is imperfect information. Second is psychology — teams are often swayed by herd effects. Third is timing — a player coming off a strong season right before his contract expires will be valued higher than a player of equal ability whose strong season came two years earlier.\n\nFor an analyst like me, this inefficiency is an opportunity. If I can identify a player who is undervalued and explain why with data, I have created value. But to do that, I must trust my data more than I trust the market consensus.\n\nAnd that is where verification becomes most important. If I am going to go against consensus, I need to be certain my data is right. A small error in verification can lead to a large wrong conclusion.\n\n## My verification method, step by step\n\nI want to share the verification method I have built over ten years. Not because it is perfect, but because I believe transparency of method matters more than transparency of conclusion.\n\nThe first step is establishing origin. Every number must have a source. If there is no source, the number is discarded.\n\nThe second step is establishing definition. Every metric must have an accompanying definition. If the definition is unclear, I redefine it myself and note that clearly.\n\nThe third step is checking sample size. Every conclusion must rest on a sufficiently large sample. I usually set a minimum threshold depending on the type of metric, but the general principle is that the smaller the sample, the more cautious the conclusion must be.\n\nThe fourth step is cross-checking against film. This is the most time-consuming step but also the most important. Film does not lie. If the data says one thing and the film says another, I trust the film first, then investigate why the data was wrong.\n\nThe fifth step is cross-checking against an independent source. Every important conclusion must be verified with at least two sources. If the two sources disagree, I do not conclude until I understand the cause.\n\nThe sixth step is noting limits. Every analysis has limits. I always state my limits clearly, even in a short podcast episode. This does not weaken me; it strengthens me, because it tells listeners I understand my own weaknesses.\n\n## Case study: reading a deal through multiple layers\n\nLet me apply this method to a typical transfer-window situation. Suppose there is a rumor that a team is negotiating to acquire a scoring player. The rumor comes with a number: that player averaged 22.4 points per game last season.\n\nLayer one: origin. Where did the 22.4 points come from? If it comes from the official stat sheet, I accept it as a starting point, but not an endpoint.\n\nLayer two: context. 22.4 points in how many minutes? If the player played 36 minutes a game, his per-minute efficiency is lower than a player scoring 18 points in 28 minutes. Scoring average does not indicate efficiency.\n\nLayer three: efficiency. The player scored 22.4 points on what shooting percentage? If he took 25 shots to score 22.4 points, that is poor efficiency. If he took 15 shots, that is good efficiency. The scoring number does not reveal this.\n\nLayer four: team context. Did the player score 22.4 points on a weak team where he was the first option, or on a strong team where he was the third option? This heavily affects whether he can sustain his efficiency on a new team.\n\nLayer five: transferability. Does the player fit the new team's system? If the new team plays fast and he plays slow, the 22.4 points may not transfer.\n\nAfter these five layers, the 22.4 points is no longer a simple fact. It becomes a data point that must be placed in context. And that is the real work of an analyst.\n\n## When the market overreacts\n\nOne of the phenomena I observe most during the transfer window is the market's overreaction to a small sample.\n\nA player with a five-game hot streak gets praised. A player with a five-game cold streak gets criticized. But five games is not a large enough sample to assess anything. In an 82-game season, five games is about 6 percent of the data.\n\nThe problem is that the transfer window usually comes right after the season ends, and the season ends with the playoffs — where the sample is even smaller. A player with a good playoff run can be valued far above his true ability. A player with a poor playoff run can be valued far below it.\n\nA small sample is not yet wrong; a hasty conclusion is wrong. I always remind myself of this when reading transfer reports in June and July.\n\n## My own limits\n\nI want to devote a section to my own limits, because I believe an analyst who does not acknowledge his limits is a dangerous analyst.\n\nThe first limit is time. I cannot rewind the full tape of every game. I must be selective, and that selection can miss information.\n\nThe second limit is resources. I do not have access to every team's internal datasets. I must rely on what is public or what my sources provide.\n\nThe third limit is personal perspective. However hard I try to be objective, I still carry my own biases, and those biases can influence how I select data.\n\nThe fourth limit is the language of data. Every metric is a way of expressing reality, not reality itself. When I use a metric, I am using a way of seeing, not absolute truth.\n\nI list these limits not to belittle myself, but to set a correct expectation. My podcast listeners should know that I am not the holder of truth. I am just someone who counts a little more carefully.\n\n## The role of analytics assistants\n\nOver the years, I have realized that the most important people in basketball analytics are often the least mentioned: the teams' analytics assistants.\n\nThey are the ones who collect data, build models, and prepare reports before every game. Their work demands patience and high precision, but they rarely appear in public.\n\nI always credit them in my analyses. Not because I want to seem kind, but because it is the right thing to do. If I used their data without credit, I would have stolen their work. And in this industry, once you steal credit, you lose sources.\n\nI learned this after I produced the series on the New York Liberty. I used data from an analytics assistant, and I credited that person in the notes. Afterward, that person became a regular source of mine. That is how sources are built: through respect, not through exploitation.\n\n## What happens next\n\nAs I look at the rest of this transfer window, I see three variables I will track closely.\n\nThe first variable is the contract structure of major deals. I will not read the total number, but the fine print: term, bonuses, protection clauses. That is where the truth lies.\n\nThe second variable is the rest patterns of key players. I will track their minute distribution through the first half of the season. If a player with an injury history starts the season with a spike in minutes, that is a signal I need to watch.\n\nThe third variable is the gap between market value and data value. I will look for undervalued players, and I will note the reasons. If my data is right, the market will adjust. If my data is wrong, I will learn something.\n\n## Final thought\n\nThere is one thing I want to leave with the reader of this piece.\n\nThe transfer market is a noisy place. Every day brings hundreds of rumors, thousands of numbers, tens of thousands of opinions. In an environment like that, clarity does not come from reading more, but from verifying more carefully.\n\nI do not watch minute 90; I watch minute 1 through 90. I do not read the final number; I read how that number was produced. And I do not trust crowd consensus; I trust what I verify by hand.\n\nIf you read a number during this transfer window, ask yourself: where did it come from, what is its definition, how large is its sample, and who benefits if you believe it. If you cannot answer those four questions, set the number aside for now.\n\nDo not rush to believe. Rewind first. That is not advice about basketball. It is a way of working. And in a market built on noise, that way of working may be your greatest asset.\n\nPeople see a mistake and laugh; I see a mistake and look for the source. That story about Zion and the wrong 9 rebounds changed my career. It taught me that the truth is not in the number that was published, but in the number you counted yourself. And in this transfer window, while everyone races to report fastest, the one who counts most carefully will be the one who is most right — not today, but a year from now, when all the noisy numbers have settled and only the truth remains.\n\nGive me the source, then we can talk. That is all I ask of a market where everyone wants to speak, but few are willing to count.