Dissenting.
Ayhan Hakimoglu played his hand, and lost. Now we are being asked to make our own bet. Sitting in diversity, we must predict how the highest court of New Jersey would rule. See Robertson v. Allied Signal, Inc., 914 F.2d 360, 378 (3d Cir.1990). As the majority points out, we must make this prediction with little guidance from New Jersey law. But that is an incident — and a flaw — of the regime of diversity jurisdiction. I believe that the New Jersey Supreme Court would recognize a cause of action, in tort, allowing patrons to recover gambling debts from casinos that serve them alcohol after they are visibly intoxicated.1 This prediction *295is founded on long standing trends in New Jersey law recognizing new causes of action, even in areas pervaded by legislation.
In my view, the New Jersey Supreme Court is especially likely to create a cause of action where a defendant profits from conduct causing the foreseeable injury, and has the ability, in the exercise of due care, to prevent such injury at small cost to itself. Because this case presents these factors, and because I am unpersuaded by the majority’s arguments, I would reverse the judgment of the district court and remand for trial on the merits. I also write to underscore a crucial point mentioned by the majority: as New Jersey has no certification procedure, we are forced to make important state policy with little guidance. I therefore suggest that New Jersey, to serve its own interests and ours, enact a certification provision.
I.
In predicting the course of New Jersey law, we must focus on policies and trends in the jurisprudence of New Jersey.2 See McKenna v. Ortho Pharmaceutical Corp., 622 F.2d 657, 662 (3d Cir.1980) (in evaluating state law, “relevant state precedents must be scrutinized with an eye toward the broad policies that informed those adjudications and to the doctrinal trends which they evince”), cert. denied, 449 U.S. 976, 101 S.Ct. 387, 66 L.Ed.2d 237 (1980). The New Jersey Supreme Court has long been a leader in expanding tort liability. For example, it was one of the first courts to announce the doctrine of strict liability, applying it to automobiles. See Henningsen v. Bloomfield Motors, Inc., 32 N.J. 358, 161 A.2d 69 (1960). Moreover, the court’s recent cases show its continuing willingness to expand tort liability in a variety of contexts. See, e.g., Weinberg v. Dinger, 106 N.J. 469, 524 A.2d 366 (1987) (imposing a duty of care on water companies to ensure adequate water pressure for firefighters); T & E Indus., Inc. v. Safety Light Corp., 123 N.J. 371, 587 A.2d 1249 (1991) (recognizing a cause of action by the owner of contaminated property against a previous owner who allegedly caused the contamination); Hopkins v. Fox & Lazo Realtors, 132 N.J. 426, 625 A.2d 1110 (1993) (imposing a duty of care for the safety of visitors to open houses); Dunphy v. Gregor, 136 N.J. 99, 642 A.2d 372 (1994) (expanding bystander liability to include a fiancee). Most relevant for our purposes, the New Jersey Supreme Court has consistently imposed liability on providers of alcohol for foreseeable drinking-related injuries — even though the sale of alcoholic beverages has been intensely regulated for many years. See, e.g., Rappaport v. Nichols, 31 N.J. 188, 156 A.2d 1 (1959) (recognizing action for death and damages against tavern that sold alcohol to minor). The court has imposed common law tort liability upon tavern owners and restaurateurs for furnishing alcohol to intoxicated persons who subsequently cause injury through drunk driving. See Soronen v. Olde Milford Inn, Inc., 46 N.J. 582, 218 A.2d 630 (1966) (extending dram shop liability to patron’s own injuries), modified in part by Lee v. Kiku Restaurant, 127 N.J. 170, 603 A.2d 503 (1992). Importantly, under New Jersey law, *296the person who was served while intoxicated himself can sue the tavernkeeper even for damages to his car. See N.J.S.A. 2A:22A-5(a). Additionally, the court has extended this liability to social hosts, even though they, unlike tavern owners, do not profit from the transaction. It “makes little sense to say that [a licensed defendant] is under a duty to exercise care, but give immunity to a social host who may be guilty of the same wrongful conduct merely because he is unlicensed.” Linn v. Rand, 140 N.J.Super. 212, 356 A.2d 15 (1976); accord Kelly v. Gwinnell, 96 N.J. 538, 476 A.2d 1219 (1984) (extending dram shop liability to social hosts).
The only two New Jersey cases to address the present issue, both from a federal district court, have predicted that New Jersey would recognize this cause of action. In GNOC Corp. v. Aboud, 715 F.Supp. 644 (D.N.J. 1989), Judge Cohen opined that “New Jersey has unambiguously communicated a strong policy against the noxious potential of excessive alcohol consumption in the twin contexts of common law dram shop liability and statutory/administrative regulation of casino alcoholic beverage service.” Id. at 653. In extending dram shop liability to the defendant casino, the court explained that its decision was “merely furthering the public policy goals underlying the Casino Control Act and the regulations promulgated thereunder.” Id. at 654.3
Then, in Tose v. Greate Bay Hotel And Casino, Inc., 819 F.Supp. 1312 (D.N.J.1993), the district court, following Aboud, held that the casino could be liable for losses flowing from its allowing an intoxicated patron to gamble. Id. at 1321-23. Tose subsequently tried his case to a jury, which rejected his claim. Greate Bay Hotel And Casino, Inc. v. Tose, 34 F.3d 1227, 1228 (3d Cir.1994). On appeal, which focused mainly on the question whether the Casino Control Commission had exclusive primary jurisdiction over gamblers’ claims against casinos, the judgment for the defendant was affirmed. Id. Because Tose lost his trial, we did not need to decide whether New Jersey would recognize this cause of action. Id. at 1232 n. 7.
II.
Analysis of this case under the principles of New Jersey tort law supports the conclusions of Aboud and Tose that New Jersey’s highest court would recognize Hakimoglu’s cause of action. In Hopkins v. Fox & Lazo Realtors, 132 N.J. 426, 625 A.2d 1110 (1993), the New Jersey Supreme Court set out its standard for determining when a tort duty, and thus a cause of action in negligence, exists. The inquiry, “ultimately a question of fairness,” requires the court to weigh (1) the relationship of the parties; (2) the nature of the risk; (3) the opportunity and ability to exercise care; and (4) the public interest in the proposed solution. Id. These factors support a cause of action in this case.
First, the relationship of the parties argues strongly for casino liability. Casinos, perhaps the ultimate for-profit institution, make their money from patrons’ losses. Gambling losses are the casino’s business. The casino and the gambler, therefore, are linked in an immediate business relationship much like that from which dram shop liability sprang— the tavern and the patron. See Rappaport, 31 N.J. at 188, 156 A.2d at 1. Like the tavern owner, the casino’s control over the environment into which the patron places himself, and its ability to open or close the alcohol spigot, imposes on the casino some concomitant responsibility toward that patron. Just as the tavern owner must make sure that drinking does not cause her patron to hurt himself or others, the casino should ensure that its alcohol service does lead its patron to hurt himself through excessive gambling.
*297Second, the nature of the risk — essentially a test of foreseeability — also points to casino liability. Gamblers come to the casino to gamble; the casino supplies free alcohol; the odds favor the casino. Losses are the natural result, if not the intent, of this situation. Unacceptable losses due to alcohol consumption are certainly foreseeable.
This foreseeability factor explains the inapplicability of contrary authority. The New Jersey Supreme Court’s recent limitations of dram shop liability, fairly read, all turn on a lack of foreseeability. See, e.g., Lombardo v. Hoag, 269 N.J.Super. 36, 634 A.2d 550 (App.Div.1993), certif. denied, 135 N.J. 469, 640 A.2d 850 (1994) (rejecting duty of passenger to stop owner of vehicle from driving because imposing such an “overbroad duty would open a Pandora’s box of potential liability and responsibility problems”); Jensen v. Schooley’s Mountain Inn, Inc., 216 N.J.Super. 79, 522 A.2d 1043 (App.Div.) certif. denied, 108 N.J. 181, 528 A.2d 11 (1987) (tavern not liable for intoxicated customer’s death after he climbed to top of tree, fell, and drowned in river); Griesenbeck v. Walker, 199 N.J.Super. 132, 488 A.2d 1038 (App.Div.1985), certif. denied, 101 N.J. 264, 501 A.2d 932 (1985) (no cause of action against social host for physical injuries from a fire at guest’s residence which occurred after the guest returned intoxicated).
Lack of foreseeability also explains why the New Jersey courts and legislature have never extended liability for tavern owners and social hosts (as opposed to casinos) beyond physical injuries and property damage. See Griesenbeck, 199 N.J.Super. at 141, 488 A.2d at 1043 (App.Div.1985) (observing that the court has never extended liability for servers of alcohol beyond injuries related to drunk driving, barroom accidents and barroom brawls); see also N.J.S.A. 2A:22A-1 et seq. (1987) (codifying liability for physical injury and property damage for “licensed alcoholic beverage server[s]”).4 Casinos, on the other hand, can plainly foresee large and unacceptable losses from patrons they help get drunk. And the New Jersey Supreme Court has made clear that tort is an appropriate basis for liability (possibly in addition to a contract theory, see supra n. 1), even if no physical damage occurs, when the losses are foreseeable. See, e.g., People Express Airlines, Inc. v. Consolidated Rail Corp., 100 N.J. 246, 495 A.2d 107 (1985) (allowing airline to recover economic damages in tort when defendant’s tank car accident required it to vacate its offices).
Finally, the presence of foreseeability rebuts the casinos complaint that recognizing liability in this case would lead to unfair and extreme results. A restaurant located near a casino would be held liable, the casinos argue, if it served alcohol to a patron who became intoxicated, entered the casino, and lost money. This, they imply, would be unfair. That may be so. But because foreseeability is lacking in the casinos’ hypothetical, the analogy to the present case does not withstand scrutiny. The restaurant and its customer, in the casinos’ hypothetical, do not stand in a similar posture to a casino and its gambling patron. The restaurant is not in the gambling business and does not necessarily know whether the dining patron would later be gambling. The loss involved, therefore, is too remote to fairly and rationally hold the restaurant accountable. By contrast, in a casino setting with gambling as the primary activity, there is no difficulty in foreseeing that the patron will engage in that activity and the high chance that he will suffer financial losses under a state of intoxication.
The third factor — the opportunity and ability to exercise care — further suggests liability here. To a much greater degree than tavern owners, casino operators can readily protect themselves against the type of liability sought to be imposed here. Unlike most tavern owners, restauranteurs or social hosts, casinos generally have huge staffs and sophisticated surveillance cameras. Gamblers, particularly high rollers, are constantly monitored by a dealer, floor persons, a pit boss, hidden cameras, and sometimes even officials of the New Jersey Casino Control *298Commission. See Tose, 819 F.Supp. at 1320. When the line is crossed, the casino need only refuse to serve more alcohol.5
Of course, the patron is also in a position to exercise care by not getting drunk. But this does not undermine my argument. New Jersey has made it clear that if the intoxicated person sues for injuries to himself, he may be charged with contributory negligence. See Kiku, 127 N.J. at 170, 603 A.2d at 503. Imposing contributory negligence is not a retreat from the policy underlying dram shop liability; rather, it is best explained as an effort to fairly apportion the loss among all who bear some responsibility. See Fisch v. Bellshot, 135 N.J. 374, 387, 640 A.2d 801, 807 (1994) (“[Pjublic policy is best served by limiting a licensee’s dram shop liability through the application of comparative negligence rather than by eliminating such liability altogether.”). This holding also ensures, from the standpoint of deterrence, that both parties in a position to avert the harm take steps to prevent it.
Finally, the public interest in the proposed solution also leads to the conclusion that New Jersey would recognize this cause of action. Throughout its history, New Jersey has exercised strict control over various types of gambling. See Tose, 819 F.Supp. at 1319. Indeed, only by a constitutional provision or amendment can any type of gambling be lawfully conducted in this state, subject to approved “restrictions and control.” N.J. Const., Art. IV, § VII, par. 2. In an environment where gambling has been regarded as “an activity rife with evil,” the state’s general ban on casino gambling should be no surprise. See Petition of Soto, 236 N.J.Super. 303, 314, 565 A.2d 1088, 1094 (App.Div.1989), cert. denied, 496 U.S. 937, 110 S.Ct. 3216, 110 L.Ed.2d 664 (1990).
Concern for the struggling city’s welfare drove New Jersey citizens to allow casino operations, with strict controls, in Atlantic City. See Tose, 819 F.Supp. at 1319. The 1977 Casino Control Act establishes a comprehensive and elaborate regulatory framework for the casino industry, reflecting a concern that casinos be restrained in order to protect the public. See N.J.S.A. 5:12-1 to 190; see also Knight v. City of Margate, 86 N.J. 374, 380, 431 A.2d 833, 836-37 (1981). The Act typically regulates the gambling operators rather than penalizing the individual gamblers. For instance, casinos, rather than an underage gambler, are held liable when the latter enters a casino. See N.J.S.A. 5:12-119; see also Department of Law & Public Safety v. Boardwalk Regency, 227 N.J.Super. 549, 548 A.2d 206 (App.Div.1988) (holding casino responsible for allowing two underage persons to gamble).
When it passed the Act, the New Jersey legislature recognized that casinos — with their concentration of wealth — have disproportionate power over the political process. See Petition of Soto, 236 N.J.Super. at 313, 565 A.2d at 1093-94. As expressed in the Act, it is New Jersey’s pronounced policy to regulate casinos “with the utmost strictness to the end that public confidence and trust in the honesty and integrity of the State’s regulatory machinery can be sustained.” Id. (emphasis added). The historical background reveals that New Jersey recognizes an important public interest in protecting gamblers. From New Jersey’s perspective, requiring casinos to protect gamblers from losses flowing from their excessive service of alcohol would probably also be in the public interest.
The most plausible objection to my position is that torts of negligence generally seek to deter and compensate for the destruction of wealth, while the tort in this case is arguably merely allocative. In other words, a typical economic tort would redress negligence that shut down a factory, causing a loss in production, while in this case the alleged tortfeasor casino coaxes the money from the gambler and then retains it. Society is no worse off; different parties just *299possess the wealth. But allocative economic torts, at least for intentional acts of conversion, are no stranger to New Jersey law. See, e.g., Atlantic Northern Airlines v. Schwimmer, 12 N.J. 298, 96 A.2d 652 (1953); Charles Bloom & Co. v. Echo Jewelers, 279 N.J.Super. 372, 652 A.2d 1238 (App.Div.1995); Lombardi v. Marzulli, 230 N.J.Super. 205, 553 A.2d 67 (Law Div.1988).
For all the foregoing reasons, application of the Hopkins criteria, see supra at 296-97, counsels us to recognize this cause of action under New Jersey law, particularly when analyzed against the background of New Jersey tort doctrine.
III.
The arguments of the majority and the district court do not compel a different result. These arguments, drawn largely from Judge Irenas’s footnote in Tose,6 rest on two main assertions: (1) that Miller v. Zoby, 250 N.J.Super. 568, 595 A.2d 1104 (App.Div.), cert. denied, 127 N.J. 553, 606 A.2d 366 (1991), undermined Aboud; and (2) that the intense legislative regulation of gambling precluded the court from finding this cause of action. In the present case, Judge Siman-dle relied on both assertions, see Hakimoglu v. Trump Taj Mahal Associates, 876 F.Supp. 625, 630-31, 633 (D.N.J.1994), and the majority places most of its stock in the legislative “scope preemption” argument. Both of these assertions are incorrect.
First, it is untrue that the Appellate Division’s decision in Zoby undermined Aboud. In Zoby, the court denied an implied cause of action against a casino for violating credit regulations. Id. 595 A.2d at 1104. But this court in Tose carefully distinguished Zoby as involving the availability of an implied right of action under the Casino Control Act— analytically a very different issue from the applicability of common law tort liability. Greate Bay, 34 F.3d at 1232 n. 7. Like the case at bar, neither Aboud nor Tose was based on an implied cause of action under the Casino Control Act or its regulations. Rather, all involve common law causes of action, which I believe the New Jersey Supreme Court would recognize.
Second, the argument that legislative regulation of casinos precludes this common law cause of action both misapprehends New Jersey jurisprudence and overstates its own force. This scope preemption argument, which forms the bulk of the majority opinion, mistakes New Jersey jurisprudence by viewing this issue through the lens of federal court interpretive assumptions, including great deference to legislative bodies. If this case presented an issue of federal law, a federal court might view the extensive legislative regulation of casinos as precluding it from properly recognizing this cause of action. But many factors might fundamentally affect how a state supreme court would interpret and make the law. To reiterate, this case requires us to predict what the New Jersey Supreme Court would do if presented with this situation. See Robertson, 914 F.2d at 378.
In my view, as explained above, New Jersey’s jurisprudence differs from that of the federal courts: New Jersey is likely to recognize a cause of action when the Hopkins factors are present, even where, because of extensive legislative regulation, federal courts would not. New Jersey’s high court has made clear that tort liability, historically a judicial matter, falls squarely in its bailiwick. “[W]e do not agree that the issue addressed in this case is appropriate only for a legislative resolution. Determination of the scope of duty in negligence cases has traditionally been a function of the judiciary.” Gwinnell, 96 N.J. at 552, 476 A.2d at 1226; accord Hopkins, 132 N.J. at 439, 625 A.2d at 1116 (“[DJetermining the scope of tort liability has traditionally been the responsibility of the courts.”). In Dunphy, the court’s most recent expansion of tort liability, the court stated:
*300We have recognized, in numerous settings, that traditional principles of tort liability-can be adapted to address areas in which recognition of a cause of action and the imposition of a duty of care are both novel and controversial.
136 N.J. at 109, 642 A.2d at 376-77 (citations omitted).
As I have explained, the New Jersey Supreme Court has long been hospitable to the recognition of liability for drinking-related injuries. See, e.g., Soronen, 46 N.J. at 582, 218 A.2d at 630 (extending dram shop liability to patron’s own injuries); Gwinnell, 96 N.J. at 538, 476 A.2d at 1219 (extending dram shop liability to social hosts). This willingness to define the scope of liability exists even where the conduct at issue is the subject of legislative or administrative regulation. See, e.g., Kiku, 127 N.J. at 170, 603 A.2d at 503 (creating contributory negligence defense in dram shop action against restaurant despite intense legislative regulation of alcohol, restaurants, and codification of dram shop liability).
Indeed, even the authority cited by the casinos as “indicative of the firm efforts of the New Jersey courts to limit the liability of a server of alcohol for a plaintiffs injuries” acknowledges that the state supreme court is free to recognize new causes of action. In Lombardo, 269 N.J.Super. at 36, 634 A.2d 550, the court reversed the trial court’s decision to extend dram shop liability. In doing so, it noted “that it is generally not considered the function of a trial court to create an exception to an established rule of law. Such a function is generally reserved for the Supreme Court or the legislature.” Id. at 48, 634 A.2d 550 (citations omitted). To emphasize again, our task here is to determine what the New Jersey Supreme Court — not a trial court — would do in this situation. See Pennsylvania Glass Sand Corp. v. Caterpillar Tractor Co., 652 F.2d 1165, 1167 (3d Cir.1981).
Moreover, even given federal jurisprudential assumptions, the scope preemption argument is overstated. As I will explain, the logical extension of this argument would lead to an absurd result: namely, absolving casinos for liability when patrons they have continued to serve kill others in drunk driving accidents. As the majority points out, the Casino Control Act closely regulates casino operation. In particular, the casinos, like the taverns, are not permitted to serve visibly and obviously intoxicated persons. See N.J.S.A. 5:12-103(d) & (f)(2); N.J.A.C. 13:2-23.1(b). The other act on which the majority leans for scope preemption, the Licensed Alcoholic Beverage Server Act, N.J.S.A. 2A:22A-1 et seq., imposes liability for physical and property injury when licensed alcoholic beverage servers (not including casinos) serve already intoxicated persons. Under the majority’s approach, because of this lacuna, casinos would not be liable for deaths caused by a gambler that it nevertheless continues to serve after the gambler is obviously intoxicated. In passing these two acts, the legislature must have thought of this possibility, and yet made no provision for it. It cannot be the case, however, that a host is liable for injuries that his guest sustained after drinking at a dinner party while casinos are absolved from liability for drunk driving accidents, even to third parties, under New Jersey law. The scope preemption argument is thus flawed for this reason as well.
IV.
While our job is not to make policy for New Jersey (and we should be careful not to do so accidently by interpreting New Jersey law under assumptions of legislative deference it does not share), policy rationales would, in fact, guide the state’s high court in appraising this putative cause of action. The district court enumerated seven problems to recognizing this cause of action. I will set out and rebut these objections below.
The first objection is essentially that the aggrieved gambler, as such, had no inhibitions that alcohol could overcome. Hakimoglu, 876 F.Supp. at 636. The second objection is that the gambler, seeking risk, got just what he came for. Id. These two objections seem to state the same point; hence I deal with them together. The point is that, inhibitions or not, the gambler got much more than he came for. Tavern patrons, of course, come to drink, but injury results if they become so drunk that they hurt them*301selves or someone else, and dram shop liability attaches. It is surely true, as Judge Simandle observed, that sober gamblers can lose big and intoxicated gamblers can win big. However, ex ante, gambling is a form of consumption because the odds favor the casino. In other words, because the casino wins in the long run, statistically the patron is paying to gamble. The patron is consuming a nondurable good like someone dining out, taking a vacation — or drinking liquor. Alcohol, by impairing judgment and lowering inhibitions, can lead the patron to consume more than he would if he were sober. As with drinking itself, excessive consumption in the form of alcohol-induced gambling can cause quite severe injury — just ask the spouse of a gambler whose house is foreclosed and kids withdrawn from their schools because the gambler lost the family’s savings. And the damage can be wrought in an instant. This is more than even the most uninhibited person bargains for.
The third, also related objection is that alcohol cannot interfere with responsible gambling because gambling requires no particular skill. Id. To begin with, the latter part of this assertion is wrong. Although slot machines require no particular ability, many others games do require skill in counting cards and making strategic choices. See Tose, 819 F.Supp. at 1319 n. 9 (discussing how card counting improves a bettor’s odds). As a whole, this assertion is also beside the point. Even if a gambler plays a game requiring no skill, alcohol can have a critical effect on his judgment about when to stop playing.
The fourth objection is that problems of proof — principally regarding proximate causation — would confound fact-finders in evaluating such claims. Hakimoglu, 876 F.Supp. at 636. Once again, the relevant concern here is not, as the casinos suggest, whether the intoxicated gambler reduces his odds of winning because of his inability to play “prudently.” Rather, the issue is whether the intoxication impairs the gambler’s ability to determine when to stop gambling when his losses grow beyond a level which he can afford. And proof of the nexus between the intoxicated status of a high roller such as Hakimoglu and his losses seems less difficult than determining the causation of cancer in many medical malpractice and toxic tort cases. Regardless, it does not furnish a basis on which to conclude that New Jersey would shy away from recognizing a cause of action. See People Express Airlines v. Consolidated Rail, 100 N.J. 246, 254, 495 A.2d 107, 111 (1985) (explaining that an “asserted inability to fix crystalline formulae for recovery on the differing facts of future cases simply does not justify the wholesale rejection of recovery in all cases”).
I acknowledge that it is not always easy to determine when a gambler is intoxicated. But this matter is before us on a motion under Fed.R.Civ.P. 12(b)(6), and we must take the well pleaded facts as true. Aid when we do, the alleged (mis)conduct of the defendant casinos, i.e., that they continued to serve a visibly intoxicated gambler who was losing millions of dollars, should be sufficient to state a claim for relief under the federal rules. I do not mean to pin a medal on the gambler, especially one such as Hakimoglu, who knows full well what he is doing when he goes repeatedly to the casino and loses big. The jury may have no sympathy for him and find him contributorily negligent, or find the casino not liable at all, as it did in the Tose case. But that is a question for the jury: whether as the result of the casino’s (mis)conduet, the gambler has lost his ability to make a reasonable judgment as to whether to continue. In my view, such an individual, at such a time, appears to be within the class that the New Jersey jurisprudence protects.
The district court’s fifth objection is that recognizing this tort could open the floodgates to fraudulent claims. Hakimoglu, 876 F.Supp. at 637. A typical dram shop claim, defendants argue, will follow an accident at which the police will be called, the blood alcohol content of the driver examined and witnesses interviewed. But that is not always the case. Many a dram shop claim— and they are recognized by New Jersey without any prerequisite of prompt investigation — are filed “out of the blue” from the vantage point of the defendant. Moreover, lawsuits such as Hakimoglu’s are both ex*302tremely costly.to pursue and quite risky, and lawyers will not undertake them except in the rare case where losses are substantial. The “floodgates” argument, therefore, is unconvincing. And since the high rollers who are both losing and drinking big are surely identified at the time of their losses, the surveillance cameras can be concentrated on them and the tapes can be specially marked and preserved. In other words, the casinos can protect themselves.
The sixth objection is that sufficient deterrence already exists because casinos cannot enforce credit markers entered into by drunk patrons. Id. A remedy in the marker situation, however, does nothing to deter losses in the many cases when, as here, the loss in question was not on credit. This objection could be restyled as one against over deterrence (and its corresponding inefficiency), which is always a potential problem for torts. If the casino had little to gain and much to lose from its behavior — as it might if, say, New Jersey law allowed large recoveries for minor physical injuries sustained in a casino — it might take overly zealous steps to prevent this occurrence. Overdeterrence is not likely to be problematic here, however, because the casinos would be liable only up to the amount that they had gained by their tortious conduct. They have much to gain and little to lose from continuing to serve intoxicated gamblers, even if this tort were recognized. If anything, underdeterrence probably would remain the biggest problem: only in some percentage of cases will the gambling losers claim and win their money back.7
Finally, the district court argues that the court should not recognize this cause of action because New Jersey’s casino regulators have never required “a casino to refund such gaming losses allegedly incurred by an intoxicated patron at any time in sixteen years of casino gambling in New Jersey.” Id. In addition to undermining its sufficient deterrence argument,8 this objection also misses the point. The legal authority and policy choices of the casino commission have no bearing on how the New Jersey Supreme Court, as a matter of common law, might choose to regulate this situation.
For all of the foregoing reasons, I am satisfied that the New Jersey Supreme Court would recognize Hakimoglu’s cause of action.
V.
This case is its own best evidence, as the majority observes, of the utility of a certification procedure; I respectfully urge New Jersey to adopt one.9 The lack of a certification procedure disadvantages both New Jersey and the federal judiciary. Especially in cases such as this where little authority governs the result, the litigants are left to watch the federal court spin the wheel. Meanwhile, federal judges, by no means a high-rolling bunch, are put in the uncomfortable position of making a choice.10 In effect, we are forced to make important state policy, in contravention of basic federalism principles. See Dolores K. Sloviter, A Federal Judge Views Diversity Jurisdiction Through the Lens of Federalism, 78 VaL.Rev. 1671 (1992). The possibility that federal courts may make interpretive assumptions that differ from those of the state court further complicates this process. States like New Jersey lacking certification procedures face the threat that federal courts will misanalyze the state’s law, already open to varied interpretations, by inadvertently viewing it through the lens of their own federal jurisprudential assumptions.
The mischief created by the lack of a certification procedure was demonstrated by Judge Sloviter when she catalogued some of *303the Third Circuit’s missteps in interpreting the law of Pennsylvania, which also lacks a certification procedure:
[W]e have guessed wrong on questions of the breadth of arbitration clauses in automobile insurance policies (we predicted they would not extend to disputes over the entitlement to coverage [Myers v. State Farm Ins. Co., 842 F.2d 705 (3d Cir.1988)], but they do [Brennan v. General Accident Fire & Life Assurance Corp., 524 Pa. 542, 574 A.2d 580 (1990) ], the availability of loss of consortium damages for unmarried cohabitants (we predicted that they would be available) [Bulloch v. United States, 487 F.Supp. 1078 (D.N.J.1980) ], but they are not [Leonardis v. Morton Chem. Co., 184 N.J.Super 10, 445 A.2d 45 (App.Div.1982) ], and the “unreasonably dangerous” standard in products liability cases (we predicted the Restatement would not apply) [Beron v. Kramer-Trenton Co., 402 F.Supp. 1268 (E.D.Pa.1975), aff'd, 538 F.2d 318 (3d Cir.1976) ], but it does [Azzarello v. Black Bros. Co., Inc., 480 Pa. 547, 391 A.2d 1020 (1978) ].
Sloviter, 78 Va.L.Rev. at 1679-80.11
New Jersey, in failing to adopt a certification procedure, is in a small minority. At present, forty-three state supreme courts, the court of last resort in Puerto Rico, and the Court of Appeals of the District of Columbia can answer certified questions of law from federal circuit courts. See American Judicature Society (“AJS”), Certification of Questions of Law: Federalism in Practice 15-17 (1995).12 Granting certification power is also supported by the federal judiciary’s Long Range Plan for the Federal Courts. Recommendation 8 of that Plan states: “The states should be encouraged to adopt certification procedures, where they do not currently exist, under which federal court (both trial and appellate) could submit novel or difficult state law questions to state supreme courts.” Committee on Long Range Planning, Judicial Conference of the United States, Proposed Long Range Plan for the Federal Courts 32 (March 1995).13 Certification is not a panacea, and can inflict delay on litigants. See Geri Yonover, A Kinder, Gentler Erie: Reining in the Use of Certification, 47 Ark.L.Rev. 305 (1994). But this is an argument for exercising the authority wisely — not for denying it altogether.
Fifty-four percent of United States Circuit judges responding to the AJS survey indicated they were “willing” or “very willing” to certify questions, AJS Report, supra, at 43, and eighty percent of state supreme court justice said they were “willing” or “very willing” to answer these questions. AJS Report, supra, at 46. Ninety-five percent of the United States Circuit Judges and ninety percent of the United States District Judges were either “very satisfied” or “somewhat satisfied” with the certification process in their most recent certified case. Id. at 42. In terms of overall satisfaction, eighty-seven percent of the state court justices said they were either “very satisfied” or “somewhat satisfied” with their most recent certification experience. Id. at 43.
While this is not a forum for drafting a certification statute, I believe that a federal court should be authorized to certify a ques*304tion of law to the state court when: (1) the issue is one of importance; (2) it may be determinative of the litigation; and (3) state law does not provide controlling precedent through which the federal court could resolve the issue. This is a textbook case for certification. The issue is determinative of the litigation; important public policy issues are at stake; and little authority guides our decision. Moreover, neither the casinos nor Hakimoglu, with all their resources, require immediate resolution of the matter. Yet, alas, New Jersey lacks a certification procedure, and still we must “predict.”14
VI. CONCLUSION
The majority fairly observes that this case is a difficult one and that reasonable arguments support either side. Nevertheless, I believe that the better arguments should lead us to predict that New Jersey would find a cause of action here, subject to the defense of contributory fault. The New Jersey Supreme Court has been highly hospitable to recognizing causes of action, even in areas where the legislature has acted, for foreseeable injuries. The four factors the court uses for evaluating whether a duty exists — (1) the relationship of the parties; (2) the nature of the risk; (8) the opportunity and ability to exercise care; and (4) the public interest — all point toward finding a cause of action here. And the policy objections of the majority and the litigants either miss the point or are overstated. For all of the foregoing reasons, I believe the New Jersey Supreme Court would recognize a cause of action, in tort, allowing patrons to recover gambling debts from casinos that serve them alcohol after they are visibly intoxicated. I therefore respectfully dissent.
. In addition to the tort theory Hakimoglu has pursued, a gambler in his position may have a claim in contract. The gambler's obvious intoxication, one might argue, voided the gambling contract. See, e.g., Feighner v. Sauter, 259 N.J.Super. 583, 590, 614 A.2d 1071, 1075 (App. Div.1992) (listing grounds for contract rescission, including intoxication); Onderdonk v. Presbyterian Homes of New Jersey, 85 N.J. 171, 183, 425 A.2d 1057, 1062 (1981) (every contract has “implied covenant of good faith and fair dealing”). The district court seemed to doubt the existence of this "so-called gambling ‘contract’ ” because "there is no mutuality.” Hakimoglu v. Trump Taj Mahal Associates, 876 F.Supp. 625, 633 n.7 (D.N.J.1994). "The patron does not negotiate the terms of his relationship with the casino," the court explained, "nor can the patron or the casino vary the rules of the game, the odds, or the payoffs.” Id.; see also Tose v. Greate Bay Hotel and Casino, Inc., 819 F.Supp. 1312, 1317 n. 8 *295(D.N.J. 1993) ("[B]ecause every aspect of the relationship between the gambler and the casino is minutely regulated by the state[,] there is little freedom of contract in the usual sense.”). But the patron retains the choice whether to play, and how much to bet. Thus, this situation is little different from most sales contracts. Purchasing a hair dryer, for example, forms a contract even though the price is set and the characteristics of the good are heavily regulated. On what other basis is the casino legally able to keep the gambler’s money after he loses? Moreover, the pervasive regulation of the gambling relationship does not nullify its contractual nature. New Jersey courts have held that gambling on credit markers forms a contract between the casino and the patron, see Lomonaco v. Sands Hotel, 259 N.J.Super. 523, 614 A.2d 634 (Law Div.1992), and that the Casino Control Act did not abrogate traditional common law contract defenses such as intoxication. See id. However, Hakimoglu has declined to press a contract claim and hence we do not decide the question.
. Precedent from Nevada, the only other state in which casino gambling is legal, provides no help, for Nevada does not recognize dram shop liability at all. See Hamm v. Carson City Nugget, Inc., 85 Nev. 99, 450 P.2d 358 (1969). The federal government has virtually complete authority over Native American Indian reservations, see James J. Belliveau, Casino Gambling Under The Indian Gaming Regulatory Act: Narragansett Tribal Sovereignty Versus Rhode Island Gambling Law, 27 Suffolk U.L.Rev. 389 (1994), but there is no federal law in this area.
. The holding of Aboud is actually broader than necessary for Hakimoglu: "In sum, a casino has a duty to refrain from knowingly permitting an invitee to gamble where that patron is obviously and visibly intoxicated and/or under the influence of a narcotic substance.” Id. at 655. This raises the interesting question about the scope of putative liability: is it liability for continuing to serve the intoxicated gambler (essentially a dram-shop theory), or liability for failure to stop him from gambling (essentially an invitee theory). While in practical terms there may be little difference between the two, doctrinally I think that the proper issue is liability for continuing to serve. At all events, Hakimoglu does not present a claim for failing to stop him from gambling.
. Because casinos are not “licensed alcoholic beverage server[s]” under the act, this law does nothing to limit casino liability directly.
. Again, on the theory that Hakimoglu pursues (based on dram shop liability), the casino presumably would need only to stop serving the patron alcohol after he became obviously and visibly intoxicated. It would not need to bar him from further gambling, though hopefully the refusal to serve might serve as a "wake-up call.” On the broader theory articulated in Aboud, however, the casino might have to keep a patron from gambling, even if he had become drunk elsewhere.
. The complicated procedural posture of the Tose case served as a sounding board for both sides of this debate. Judge Rodriguez, to whom the case was originally assigned, elected to follow Aboud. The case was then reassigned to Judge Irenas, who acknowledged that he was bound by Aboud as the law of the case, but noted his disagreement with that case in a footnote. Tose, 819 F.Supp. at 1316 n. 8.
. For more analysis of the economics of a gambling tort, and the connection between gambling and alcohol, see Jeffrey C. Hallam, Note, Rolling the Dice: Should Intoxicated Gamblers Recover Their Losses, Nw.U.L.Rev. 240 (1990).
. How can there be sufficient deterrence in the nonmarker situation when enforcement has been totally lacking?
. Judges Nygaard and Alito have expressed their agreement with the recommendations contained in this part of my opinion.
. I am, needless to say, not the first to make this observation. See, e.g., McKenna, 622 F.2d at 661 ("Although some have characterized this assignment as speculative or crystal-ball gazing, nonetheless it a task we may not decline.”).
.For additional examples of our difficulty predicting state law, and a call for the State of Pennsylvania to adopt a certification procedure, see Stella L. Smetanka, To Predict or To Certify Unresolved Questions of State Law: A Proposal for Federal Court Certification to the Pennsylvania Supreme Court, Temp.L.Rev. 725 (1995). In particular, Smetanka describes the Third Circuit’s troubles in assessing the scope of Pennsylvania’s public-policy exception to at-will employment in the wake of Geary v. United States Steel Corp., 456 Pa. 171, 319 A.2d 174 (1974). For recent examples of this difficulty, see Borse v. Piece Goods Shop, Inc., 963 F.2d 611 (3d Cir.1992), and Smith v. Calgon Carbon Corp., 917 F.2d 1338 (3d Cir.1990).
. [hereinafter "AJS Report”]. The states vary widely on whether the source of this authority is a constitutional provision, statute, court rule, or a combination of the three. Id. The states also differ considerably on their standard for accepting certified questions. Eleven states require that the certified question be determinative of the litigation; twenty-six states, Puerto Rico, and the District of Columbia require only that the question may be determinative; and six others require that there be — or appear to be — no controlling precedent or authority. AJS Report, supra, at 18-20.
. Recommendation 8 was adopted by the Judicial Conference as part of the approved long range plan on September 19, 1995.
. In order to bring this proposal to the attention of the appropriate New Jersey authorities, I request that the Clerk mail copies of this opinion, referencing Part V of the dissent, to the Chief Justice of the New Jersey Supreme Court, the Director of the Administrative Office of New Jersey Courts, the Chair of the Judiciary Committees of the New Jersey House and Senate, and the Attorney General of New Jersey.