POSNER, Chief Judge.

Between 1984 and 1989, Fujisawa Pharmaceutical Company, a large manufacturer of drugs, purchased stock in Lyphomed, Inc., a manufacturer of both proprietary and generic drugs that was controlled by its principal shareholder and executive, John Kapoor. By 1986, Fujisawa had displaced Kapoor as Lyphomed's largest shareholder, owning 30 percent of its shares; and in 1990 Fujisawa merged Lyphomed into a wholly owned subsidiary of Fujisawa that is, along with Fujisawa itself, a plaintiff in this case (we shall refer to both plaintiffs, collectively, as Fujisawa). By this time Fujisawa had sunk some $800 million into the purchase of Lyphomed stock. On August 17, 1992 (and the date is not a detail), Fujisawa brought this suit against Kapoor under the federal securities laws and RICO, charging that between 1983 and 1990 Kapoor had concealed from Fujisawa the fact that between 1980 and 1986 Lyphomed had filed a large number of fraudulent "ANDAs" with the Food and Drug Administration.

An "ANDA" (Abbreviated New Drug Application) is a request to be allowed to produce a generic drug upon the expiration of the patent covering the proprietary original. 21 U.S.C. § 355(j); 21 C.F.R. secs. 314.92-.99. The abbreviation lies in the fact that, since the proprietary version had been approved upon a showing that it was safe and effective (the showing required for approval of an application for a new drug), the producer of the generic version need only show that it is equivalent to the proprietary version. 21 U.S.C. § 355(j)(2)(A). On February 4, 1991, the FDA, suspicious of the accuracy of some of the data in Lyphomed's ANDAs, began an investigation that eventually resulted in the withdrawal of a number of its generic drugs from the market and a temporary ban on the submission of new ANDAs by the company. Both measures hurt Fujisawa-Lyphomed's generic-drug division.

The suit claims that Kapoor, by concealing Lyphomed's troubles with the FDA from Fujisawa while the latter was buying stock in Lyphomed, much of it from him personally, committed securities frauds in violation of section 10(b) of the Securities Exchange Act of 1934, 15 U.S.C. § 78j(b), and the SEC's Rule 10b-5 thereunder, 17 C.F.R. sec. 240.10b-5, and made false statements in reports required by the Act, in violation of section 18, 15 U.S.C. § 78r. The district judge granted summary judgment for Kapoor on the securities counts on the ground that they were barred by the one-year statute of limitations applicable to them. Fujisawa later tried to amend its complaint to add a claim under section 20A of the '34 Act, 15 U.S.C. § 78t-1 (a provision, added in 1988, directed against insider trading), which has a five-year statute of limitations. sec. 78t-1(b)(4). The judge held section 20A inapplicable to the facts alleged by Fujisawa, and so denied as futile the motion to amend. She dismissed the RICO count on the ground that there was no evidence of the required "pattern" of racketeering activity, 18 U.S.C. § 1962(c), and relinquished jurisdiction over Fujisawa's supplemental state-law claims. 936 F. Supp. 455 (N.D. Ill. 1996).

The one-year statute of limitations applicable to suits under Rule 10b-5 begins to run not when the fraud occurs, and not when the fraud is discovered, but when (often between the date of occurrence and the date of the discovery of the fraud) the plaintiff learns, or should have learned through the exercise of ordinary diligence in the protection of one's legal rights, enough facts to enable him by such further investigation as the facts would induce in a reasonable person to sue within a year. Law v. Medco Research, Inc., 1997 WL 253023, at *4 (7th Cir. May 15, 1997), and cases cited there. This is the doctrine known as "inquiry notice." The parties put opposite untenable glosses on it. Fujisawa contends that the statute of limitations doesn't begin to run until the victim has in hand all the facts he needs in order to bring suit immediately, and that this wasn't until shortly before it brought its suit, or perhaps even after — for it had not yet completed its investigation when it sued, so that it had to move to amend its complaint later. On this view, the potential plaintiff can complete his investigation, draft his complaint, and put the complaint in a drawer to be taken out in a year and filed if the price of the stock has fallen. Kapoor contends that the statute of limitations begins to run as soon as the victim has access to the facts that would show the fraud, which he contends means no later than 1990, when Fujisawa acquired Lyphomed and with it custody of copies of all the questionable applications that the FDA's investigation later brought to light.

Fujisawa's argument merges inquiry notice into discovery of the violation. We tried to clarify the distinction in Law v. Medco Research, Inc., supra, a case that, as it happens, arose indirectly from Kapoor's activities here alleged as fraud. (Medco, a licensor of Fujisawa-Lyphomed, was sued by investors for concealing Lyphomed's problems with the FDA that caused the agency to delay its approval of Medco's most important drug.) Inquiry notice, we emphasized in Law, must not be construed so broadly that the statute of limitations starts running too soon for the victim of the fraud to be able to bring suit within a year. The facts constituting such notice must be sufficiently probative of fraud — sufficiently advanced beyond the stage of a mere suspicion, sufficiently confirmed or substantiated — not only to incite the victim to investigate but also to enable him to tie up any loose ends and complete the investigation in time to file a timely suit. Cf. Tregenza v. Great American Communications Co., 12 F.3d 717, 721 (7th Cir. 1993).

But the facts that put the victim of the fraud on notice can fall short of actual proof of fraud. How short may depend on the victim's access to the information that he will need in order to be able to plead a reasonably well substantiated and adequately particularized case of securities fraud, bearing in mind that before he files his suit he will not have the aid of compulsory process. The better his access, the less time he needs. "Suspicious circumstances, coupled with ease of discovering, without the use of legal process, whether the suspicion is well grounded, may cause the statute of limitations to start to run before the plaintiffs discover the actual fraud." Law v. Medco Research, Inc., supra, 1997 WL 253023 at *4.

Kapoor fastens on ease of access to the necessary information. All of it was in documents that were in the possession of the victim itself, Fujisawa, as the controlling shareholder and later sole owner of Lyphomed. But more than bare access to necessary information is required to start the statute of limitations running. There must also be a suspicious circumstance to trigger a duty to exploit the access; an open door is not by itself a reason to enter a room. We reject the suggestion that the defrauded purchaser of a company is presumed to be on notice of everything in the company's files, so that the statute of limitations begins to run at the moment of the acquisition. FDIC v. Manatt, 688 F. Supp. 1327, 1331 (E.D. Ark. 1988), aff'd, 922 F.2d 486 (8th Cir. 1991).

How suspicious the circumstance need be to set the statute of limitations running — how close, in other words, it needs to be to the proof that one would have to have in hand in order to be able to file suit on the spot without running afoul of Rule 9(b), which requires that fraud be pleaded with particularity, and without courting Rule 11 sanctions — will depend on how easy it is to obtain the necessary proof by a diligent investigation aimed at confirming or dispelling the suspicion. If, as Fujisawa argues, the very first suspicious circumstance was the opening of the FDA's investigation in February of 1991, then it is possible although unlikely that the statute of limitations would not have begun to run until sometime later than that, when the investigation brought to light the fraudulent applications. It is unlikely because Fujisawa had both better access to the relevant documents than the FDA and a greater incentive (unless it wanted to play ostrich) to find in them evidence that Kapoor had concealed information about Lyphomed's ANDAs that was highly material to the value of a company that Fujisawa had paid hundreds of millions of dollars to acquire. So the commencement of the FDA's investigation would have been the last possible signal to Fujisawa to loose the hounds on the trail of a fraud. Fujisawa should not have waited until the FDA's investigation hit pay dirt before beginning its own investigation.

In any event the FDA's investigation that began in February 1991 was not the first event that should have aroused Fujisawa's suspicions. The investigation was not a bolt from the blue; the facts that gave rise to it had long been known to Fujisawa; it should have begun suspecting fraud many years earlier. Its counsel have done an elaborate dance around the facts known to their client before August of 1991, arguing with great ingenuity that every single one of them can be explained without hypothesizing fraud by Kapoor. That misses the point. Of course a suspicion may turn out to be unfounded; that doesn't mean that it would not induce a careful person to inquire whether it was unfounded. As early as 1987 — by which time Fujisawa was already Lyphomed's largest shareholder, was represented on the board of directors, and had employees working at Lyphomed's facilities — the FDA cited Lyphomed for a number of violations of FDA regulations, including selling drugs that had not been properly tested or did not comply with specifications. The FDA even seized Lyphomed products at one of its plants and placed the company on an "Alert List" that limited its ability to sell new products. The price of Lyphomed's stock (which was publicly traded) fell and Fujisawa expressed concern. Employees of Lyphomed, and Fujisawa employees working at Lyphomed, warned Fujisawa that Lyphomed was rushing out ANDAs too fast. Lyphomed promised the FDA to clean up its act, and in 1988 was removed from the Alert List, but the FDA continued to refuse to approve new ANDAs submitted by the company.

There were other significant developments in 1988. Lyphomed's board, on which Fujisawa had a seat, reduced Kapoor's role in the company. And a securities fraud class action was brought on behalf of the shareholders of Lyphomed, including Fujisawa, against Lyphomed — and Kapoor. The suit charged among other things that Lyphomed had concealed discrepancies and alterations in its laboratory records, had concealed (other) violations of FDA regulations, and had falsely touted its ability to submit timely, well-prepared ANDAs. Speed in obtaining approval for a generic drug is essential to profitability, and Kapoor had bragged about Lyphomed's speed in churning out ANDAs. The class action, an offshoot of which was before us in Harman v. Lyphomed, Inc., 945 F.2d 969 (7th Cir. 1991), was settled for $10 million.

There is more, but we have said enough to show that well before February 4, 1991, Fujisawa had reason to believe that Kapoor had deliberately concealed from it serious problems with ANDAs that might result in the withdrawal of the FDA's approval of a number of Lyphomed's products, materially reducing the value of the stock that Fujisawa had bought in blissful ignorance of these problems. Fujisawa should have begun to investigate, and had it done so it would not have been caught by surprise on February 4 and could easily have obtained enough information to sue within a year of that time.

So the Rule 10b-5 claim is time-barred, and the section 18 claim as well. It is true that section 18 has its own limitations provision, 15 U.S.C. § 78r(c), with the same one-year statute of limitations and three-year statute of repose that is applied to Rule 10b-5 claims, and the provision does not include the "should have known" language that is the source, or at least a source, of the duty to investigate suspicious circumstances. But as we explained in Tregenza, with specific reference to another securities law statute of limitations (indeed the one, section 9(e) of the Securities Exchange Act of 1934, 15 U.S.C. § 78i(e), that is borrowed for use in Rule 10b-5 suits) — but with a glance at section 18(c) as well — Congress's omission of the "should have known" language appears not to have been deliberate and has not deterred the courts from interpreting the discovery rule in securities fraud cases as imposing a duty of inquiry even when the statute does not impose such a duty in so many words. Tregenza v. Great American Communications Co., supra, 12 F.3d at 721-22, and cases cited there. This was done for section 18(c) in Menowitz v. Brown, 991 F.2d 36, 41-42 (2d Cir. 1993) (per curiam).

It would be a mistake in any event to place too much weight on the "should have known" language. A discovery rule is a standard graft on statutes of limitations, and the question, which Congress has seemed minded to leave to the courts, is its precise meaning in particular statutory settings. In most fields it refers to discovery just of the plaintiff's injury. E.g., Cada v. Baxter Healthcare Corp., 920 F.2d 446, 450 (7th Cir. 1990). In the securities field it has a broader, a more generous meaning. Law v. Medco Research, Inc., supra, 1997 WL 253023 at *4. But its generosity is not unlimited. A person who is put on notice of a problem and fails to investigate to determine whether it exists can be thought to have "discovered" it by analogy to the "ostrich" doctrine in the criminal law, where a person who takes deliberate efforts to avoid discovering that he is assisting in a criminal act is treated just as if he had discovered it. United States v. Giovannetti, 919 F.2d 1223, 1227-28 (7th Cir. 1990). It would be highly undesirable if, suspecting securities fraud, an investor could sit back and wait out the entire three years of the repose period before suing, so that he could see how the price of his stock behaved in the interim.

The ostrich analogy is not perfect; and we are mindful that the grafting of inquiry notice onto statutes of limitations in the securities laws has been said to be inconsistent with dicta in the Supreme Court's opinion in Lampf, Pleva, Lipkind, Prupis Petigrow v. Gilbertson, 501 U.S. 350 (1991). See Lewis D. Lowenfels Alan R. Bromberg, "SEC Rule 10b-5 and Its New Statute of Limitations: The Circuits Defy the Supreme Court," 51 Bus. Law. 309, 315-17, 332-34 (1996). But the tension dissolves when inquiry notice is defined, in our recent Law decision as in this case, to require more than merely suspicious circumstances — to require that the suspicious circumstance place the potential plaintiff in possession of, or with ready access to, the essential facts that he needs in order to be able to sue. Law v. Medco Research, Inc., supra, 1997 WL 253023 at *4.

The judge was right to refuse Fujisawa permission to amend its complaint to add a claim based on section 20A of the '34 Act in order to get around the one-year statute of limitations applicable to Rule 10b-5. The amendment would have been futile. The section, although broadly worded — it makes a violator liable to "any person who, contemporaneously with the purchase or sale of securities that is the subject of such violation, has purchased . . . securities of the same class," 15 U.S.C. § 78t-1(a) — has no application to this case. See Lampf, Pleva, Lipkind, Prupis Petigrow v. Gilbertson, supra, 501 U.S. at 361-62; Ceres Partners v. GEL Associates, 918 F.2d 349, 362-63 (2d Cir. 1990). It was added in 1988 to give standing to additional victims of insider trading. Suppose that A, an insider, buys stock from B, an outsider, knowing that the price of the stock will rise when the inside information becomes public, and at approximately the same time ("contemporaneously") outsider C sells shares of the same stock ("securities of the same class") to D, another outsider, presumably at the same low price at which B sold to A. Section 20A makes A in our hypothetical case liable to C as well as to B. James D. Cox, Robert W. Hillman Donald C. Langevoort, Securities Regulation: Cases and Materials 811-12 (2d ed. 1997). We do not know why Congress fixed so long a statute of limitations for this class of cases.

Fujisawa does not argue that its case fits the paradigm that we sketched. It does not seek the profits that Kapoor made by selling to others on the basis of his inside information, because Kapoor sold only to it; it is seeking his profits on his sales to it. 15 U.S.C. § 78t-1(b)(1). It thus is seeking the identical remedy under section 20A as under Rule 10b-5, for Kapoor's profit is equal to Fujisawa's loss. This shows that Fujisawa reads section 20A to reach any case in which a securities fraud involves trading on inside information. This interpretation would amount to saying that Congress, in attempting to provide additional relief for victims of insider trading, had inadvertently enacted a five-year statute of limitations applicable in effect to a vast number of Rule 10b-5 cases — an implausible suggestion not compelled by the statutory text that we quoted. It is true that Kapoor had and traded on inside information, but, to repeat, the purpose of section 20A was to extend the protections of the existing insider-trading prohibition to persons not in privity with the insider, which Fujisawa very much was.

We come last to the RICO count. Although Congress as part of its recent reform of private securities litigation has deleted securities frauds from the list of "racketeering" activities that trigger the application of RICO, the amendment is inapplicable to suits, such as this one, that were pending when the new law was passed, provided that the suit arose under the securities laws. Private Securities Litigation Reform Act of 1995, Pub.L. No. 104-67, sec. 107, 109 Stat. 737, 758 (codified at 18 U.S.C. § 1964(c)); id., sec. 108. Kapoor argues that the suspension of the amendment is applicable only to securities claims, and not to RICO claims. The argument is hard to fathom. The new law deals only with securities litigation, and with RICO only insofar as a RICO claim might be part of an action brought under the securities laws as well — like this suit. It is only in cases such as this that RICO claims can no longer be brought and hence only in cases such as this that were pending on the date of enactment that the right to maintain a RICO claim is preserved.

The courts have spilled oceans of ink trying to define the word "pattern" in RICO. E.g., H.J. Inc. v. Northwestern Bell Tel. Co., 492 U.S. 229, 236-49 (1989); GICC Capital Corp. v. Technology Finance Group, Inc., 67 F.3d 463, 465-69 (2d Cir. 1995); Menasco, Inc. v. Wasserman, 886 F.2d 681, 683-85 (4th Cir. 1989). The definitional complexities can be elided by using the concept of prototypes or family resemblance to identify behaviors sufficiently close to the unquestioned core of RICO to justify the application of the law. A core RICO offense would be a criminal takeover of a legitimate business which the criminal then used to launder his illegal earnings or commit further illegalities behind a facade of legitimate enterprise. United States v. Turkette, 452 U.S. 576, 591 (1981); Cenco Inc. v. Seidman Seidman, 686 F.2d 449, 457 (7th Cir. 1982). Slightly outside the core, and because only slightly outside still within the prohibitory scope of the statute, would be the case of a person who controlled a company and used the mails to progressively fleece an elderly widow — first soliciting an investment of her savings in a Ponzi scheme sponsored by the company, then urging her to sell her furniture and invest her proceeds in the scheme, then inducing her to take out a second mortgage on her house to enable a further investment in the scheme, the whole campaign unfolding over several years and leaving her at the end destitute. See Liquid Air Corp. v. Rogers, 834 F.2d 1297, 1304-05 (7th Cir. 1987); Uniroyal Goodrich Tire Co. v. Mutual Trading Corp., 63 F.3d 516, 524 (7th Cir. 1995); Gagan v. American Cablevision, Inc., 77 F.3d 951, 962-64 (7th Cir. 1996); cf. Burdett v. Miller, 957 F.2d 1375, 1379 (7th Cir. 1992).

Substitute Fujisawa for elderly widow and you have our case, always assuming that Fujisawa can actually prove the allegations of its complaint. Kapoor controlled Lyphomed and, if those allegations are believed, used his control to conduct the business of Lyphomed through a pattern of racketeering activity extending over six years and consisting of innumerable mail and wire frauds successfully designed to lure Fujisawa into investing more and more of its money in Lyphomed, much of which went into Kapoor's pocket. If the successive transactions between Kapoor and Fujisawa had merely been installments in a sale of the company, the requirement of a pattern would probably not have been satisfied, Skycom Corp. v. Telstar Corp., 813 F.2d 810, 818 (7th Cir. 1987), because the reality would have been that there was only a single fraud. But this was not an installment deal; Fujisawa did not commit itself, when it began buying stock in Lyphomed, to buying the entire company.

RICO has a four-year (borrowed) statute of limitations, Agency Holding Corp. v. Malley-Duff Associates, Inc., 483 U.S. 143, 156 (1987), and it is at present an unsettled question, on which the Supreme Court has however recently granted certiorari, Klehr v. A.O. Smith Corp., 117 S. Ct. 725 (1997), and heard argument, whether it runs, in effect, from the first or the last predicate act (more precisely, from discovery of either). In this circuit, in a case such as this in which the predicate acts are fraud, it runs from the discovery of the injury but can be tolled until the plaintiff has enough information to sue, Bontkowski v. First National Bank, 998 F.2d 459 (7th Cir. 1993); McCool v. Strata Oil Co., 972 F.2d 1452, 1465-66 (7th Cir. 1992), so that the net effect is similar to the running of the statute of limitations in a Rule 10b-5 case, cf. Law v. Medco Research, Inc., supra, 1997 WL 253023 at *4, the difference being the length of the limitations period. Neither the district judge nor the parties have discussed the RICO statute of limitations; it remains for consideration on remand.

Fujisawa asks us, if we remand the case (as we are doing), to direct that it be assigned to another judge because Judge Bucklo on her own initiative, without notice to the parties, reversed the ruling of the district judge previously assigned to the case denying Kapoor's motion to dismiss the RICO claim for want of a pattern of racketeering activity. The doctrine of law of the case requires the second judge in a case in which there has been a reassignment to abide by the rulings of the first judge unless some new development, such as a new appellate decision, convinces him that his predecessor's ruling was incorrect. Best v. Shell Oil Co., 107 F.3d 544, 546 (7th Cir. 1997); Williams v. Commissioner, 1 F.3d 502, 503 (7th Cir. 1993). The qualification is essential, as the failure to revisit the original judge's ruling in such a case will doom the ultimate decision to reversal. Judge Bucklo thought this was such a case, and though we disagree with her that Uniroyal Goodrich Tire Co. v. Mutual Trading Corp., supra, on which she relied, changed the circuit's law on the requirements for proving a pattern, she can hardly be thought to have been acting improperly. More questionable is her having dismissed the RICO claim without notice to the parties, to give them an opportunity to show that she was wrong to think that her predecessor's ruling had been incorrect on the basis of the new decision. Stewart Title Guaranty Co. v. Cadle Co., 74 F.3d 835, 836-37 (7th Cir. 1996). But all judges make mistakes; mistake is not a ground for our ordering a case reassigned on remand — that would make the exception swallow the rule, since almost always when a case is remanded it is because of an error in the proceeding under review.

The judgment of the district court is affirmed in part and reversed in part, and the case remanded for further proceedings consistent with this opinion.

AFFIRMED IN PART, REVERSED IN PART, AND REMANDED.