EASTERBROOK, Chief Judge.

Michael Sanders arrived in the United States from Nigeria with 3.6 kilograms of heroin in his luggage. When caught, he claimed to be a courier with no interest in the drugs apart from a $3,000 fee for his services; he agreed to participate in a controlled delivery to the next people in the chain, who were to collect the heroin at a bus station in Chicago. Several conversations in Yoruba with "Baba," Sanders's contact, preceded his arrival. Eventually Taofiq Afonja drove up and told Sanders to put his luggage in the trunk of his car. Sanders asked, in Yoruba, whether Afonja was "that person" or "the one" (and surely was not referring to Neo in The Matrix). Afonja replied (in translation) that he was, and that "[t]hey have spoken to us. It is them they are talking to on that phone." Afonja then took the suitcase but before he and his passenger, Folashade Moore, could leave, all three were arrested. Another car, presumably carrying Baba, got away; he is a fugitive.

Sanders pleaded guilty to conspiring to possess the heroin with intent to distribute it and has been sentenced to 120 months' imprisonment, the statutory minimum. Moore confessed that she had gone to the bus station to pick up a drug courier for Baba, her boyfriend. Nonetheless she pleaded not guilty. A jury convicted her of attempted possession of the heroin. She has been sentenced to 121 months' imprisonment. Afonja, who did not confess, was tried separately to avoid Bruton problems and convicted of conspiracy and attempt. His sentence, too, is 121 months. All three have appealed, but counsel for Moore and Sanders have filed Anders briefs. Sanders does not want to withdraw his guilty plea and received the lowest available sentence; he has no conceivable appellate issue. Moore proposes to contest the admissibility of her confession on the ground that the agents did not give Miranda warnings, but no motion to exclude the confession was made before or during trial, so this argument has been forfeited. What's more, three agents testified without contradiction that Miranda warnings had been given. Other potential arguments likewise would be unavailing, as Moore's lawyer concluded. We dismiss Sanders's and Moore's appeals as frivolous.

Afonja has a non-frivolous argument: that a witness testifying as an expert for the prosecution did not satisfy the requirements of Fed.R.Evid. 702. Afonja maintained that he didn't know what was in Sanders's suitcase. Robert Coleman, a police officer employed by Will County, Illinois, and assigned to a drug task force, testified for the prosecution as an expert about drug transactions. One of the questions he addressed was whether innocent persons participate in drug transactions. Over Afonja's objection, Coleman testified that, except for children, only "people that are involved in the drug deal" will be present — and by "involved" Coleman meant people who "have knowledge as to what's taking place, the illegal activity". Afonja maintains that the district judge should have prevented Coleman from giving this testimony.

The district judge concluded that Coleman's training and experience make him an expert on drug transactions. The prosecutor repeats this theme, and we may assume that Coleman indeed knows much more about these transactions than do jurors and so is well situated to provide information about them. But Rule 702 does not say that any testimony within the scope of a witness's expertise is admissible. It provides:

If scientific, technical, or other specialized knowledge will assist the trier of fact to understand the evidence or to determine a fact in issue, a witness qualified as an expert by knowledge, skill, experience, training, or education, may testify thereto in the form of an opinion or otherwise, if (1) the testimony is based upon sufficient facts or data, (2) the testimony is the product of reliable principles and methods, and (3) the witness has applied the principles and methods reliably to the facts of the case.

The district judge did not address any of the Rule's three questions: (1) whether Coleman's view "is based upon sufficient facts or data"; (2) whether it is "the product of reliable principles and methods"; and (3) whether the "witness has applied the principles and methods reliably to the facts of the case."

Both the judge and the prosecutor stopped with the proposition that Coleman is an expert; for its part, the defense also bypassed the Rule's requirements in favor of the assertion that an expert should not be allowed to testify unless his experience includes a transaction just like this one (presumably, one in which several Yoruba-speaking people exchange a suitcase outside a bus station in Chicago). The defense position is anti-intellectual and has nothing to do with Rule 702; the point of good data and reliable analysis is to find patterns that transcend details such as which bus station is used or what language people speak.

Both the judge and the prosecutor supposed that decisions in this circuit make it unnecessary to address the questions posed by Rule 702. We have held that an agent's field experience can provide "specialized knowledge" that supports expert testimony. See, e.g., United States v. Ceballos, 302 F.3d 679, 686-88 (7th Cir. 2002); United States v. Allen, 269 F.3d 842, 846 (7th Cir. 2001). And we have twice held that district judges did not err in admitting testimony of the kind that Coleman gave here. See United States v. Garcia, 439 F.3d 363, 367-68 (7th Cir. 2006); United States v. Love, 336 F.3d 643, 646-47 (7th Cir. 2003). But neither Garcia nor Love dealt with Rule 702. Garcia held that testimony (by Coleman himself) did not deprive the accused of the presumption of innocence, and Love that testimony about the probability of innocents participating in drug deals did not violate Fed.R.Evid. 704(b), which forbids expert testimony about whether the defendant had the mental state required for conviction. Neither Garcia nor Love holds that district judges must admit testimony of the sort that Coleman proffered or excuses judges from conducting the inquiry required by Rule 702 every time any expert proposes to testify. See Kumho Tire Co. v. Carmichael, 526 U.S. 137, 153-57, 119 S.Ct. 1167, 143 L.Ed.2d 238 (1999).

On what "facts or data" does Coleman's opinion rest? Are his inferential methods reliable? Coleman did not describe either the facts he considered or the methods of analysis used to get from facts to a conclusion. All he said is that drug dealers don't want extra witnesses, which leads them to include in any given transaction only a circle of knowledgeable operatives. That a priori proposition may be sound, but an incentive ("avoid conducting drug deals where people who may talk to the police can see you") operates at the margin. The existence of this incentive implies that there will be fewer strangers at drug deals than at sales of antique chairs or swaps of baseball cards, but it does not imply that the number of innocent adults on the scene will be zero. And there are contrary incentives. Dealers may use stooges because they are cheap and reliable; people who know that they are picking up drugs must be paid for the risk that they are taking, and they may be tempted to steal the drugs. Dupes have no reason to demand payment or make off with the inventory. How do these contrary incentives balance out? Coleman did not even mention the potential benefits to drug dealers of using ignorant participants.

How these and other incentives play out cannot be determined by a priori reasoning. Facts are essential to testimony based on "specialized knowledge" as well as to scientific and technical expertise. Yet Coleman did not describe any data, and his evaluation does not seem to be falsifiable. Coleman is certain that every adult involved in every drug transaction knows what is going on. Thus if Afonja protests ignorance of what was in the suitcase, Coleman will not believe him. He will treat this as one more "proof that only knowledgeable participants come to an exchange of drugs. Coleman does not have — or at least did not explain — any way to avoid the GIGO problem. (Garbage in, garbage out.) He assumes that everyone present is culpable and uses that assumption as the "proof of culpability. That's not a reliable way to proceed.

Now maybe Coleman has done some data collection and evaluation, or maybe such work has been done by others — though we could not find any published literature on the subject. Perhaps Coleman knows what portion of people found at drug transactions can be identified as culpable on the basis of evidence other than their presence at the transactions, and what portion can be ruled out as knowledgeable participants (again on the basis of other evidence). But no one — not the prosecutor, not defense counsel, and not the judge — asked whether such empirical work had been conducted. No one tried to apply the three criteria in Rule 702 to Coleman's testimony. Defense counsel's attention was elsewhere: on the theme that Coleman had not encountered a drug transaction exactly like this one.

The prosecutor's brief and oral argument rest on the proposition that testimony by any genuine expert is admissible under Rule 702. That's not so. Most junk science is the work of people with Ph.D. degrees and academic positions. For example, in Emerald Investments L.P. v. Allmerica Financial Life Insurance Annuity Co., 516 F.3d 612 (7th Cir. 2008), a professor of finance at a respectable university gave irresponsible testimony that "demonstrated a willingness to abandon the norms of his profession in the interest of his client." In other cases we have excluded analysis that scholars would not have accepted in their undergraduates' term papers. See, e.g., Zenith Electronics Corp. v. WHTV Broadcasting Corp., 395 F.3d 416 (7th Cir. 2005); Mid-State Fertilizer Co. v. Exchange National Bank, 877 F.2d 1333 (7th Cir. 1989). Good credentials may be a necessary condition for expert testimony but are not a sufficient condition.

Because the right questions were not asked, we cannot know whether Coleman should have been allowed to testify. It is difficult to say that the district judge abused his discretion — that's the standard, see General Electric Co. v. Joiner, 522 U.S. 136, 118 S.Ct. 512, 139 L.Ed.2d 508 (1997) — when the information on which a sound exercise of discretion depends was never placed before the judge. A judge is not obliged to look into the questions posed by Rule 702 when neither side either requests or assists. So there was no error; the judge answered correctly the only question that the parties posed (whether Coleman qualified as an expert).

Afonja contends that the evidence was insufficient to support his conviction, but the conversation between Afonja and Sanders is damning. Sanders asks "[a]re you that person" and explains that he needs to know "[s]o that I don't go and give the thing to somebody else." Afonja answers that he is the one, that "[i]t is us. It is us." There is more in the same vein. Afonja's statements are not those of a person who was just coming along to spare Moore the need to move a friend's heavy suitcase. Moreover, the car bearing Afonja and Moore (and the second car believed to contain Baba) had been at the station before Sanders's bus arrived. When a police cruiser arrived serendipitously, these two cars hightailed it out of there and did not return until Sanders called to announce his arrival. (They tried to get Sanders to use a taxi and appeared themselves only when he balked.) This behavior bespeaks guilty knowledge.

Afonja's conviction is affirmed. The appeals of Sanders and Moore are dismissed as frivolous, and we grant their lawyers' motions to withdraw.