dissenting.
I respectfully disagree with the court’s holding that there was probable cause to search Mr. Martinez’s Buick without the drug dog’s positive indication because, if we disregard the evidence provided by Radar, there is no evidence linking the car with drug trafficking. Radar’s alert to one side, the warrant application indicated only that Martinez was suspected of dealing drugs and that *402the police found drug paraphernalia in his apartment and marijuana in his wife’s car. Under the established law of this circuit, none of this makes it probable that Mr. Martinez’s car contained evidence of drug dealing.
In United States v. Hogan, 25 F.3d 690, 693-94 (8th Cir.1994), we held that an affidavit containing substantially identical information did not provide probable cause to search a car because it failed to establish a connection between that ear and drug trafficking. (It indicated that the defendant was transporting drugs in his pickup, that the police found drug paraphernalia and a small amount of drugs in his house, and that he left the house in the car that was searched.) The court distinguishes Hogan on the ground that in that ease the police seized the car “at a point on a public highway some three to five miles from the defendant’s home, whereas Martinez’s ear was parked near his residence.” This distinction is illusory because, in both cases, the police first encountered the relevant vehicle at the defendants’ residences. The fact that the police permitted the defendant in Hogan to drive the car away and then radioed a backup unit to stop him seems to me to be of no legal consequence.
The court’s opinion expands probable cause beyond the boundaries established by our prior decisions. Previous similar cases have always involved evidence directly connecting the searched car with drugs, see, e.g., United States v. Wagner, 884 F.2d 1090, 1094 (8th Cir.1989), cert. denied 494 U.S. 1088, 110 S.Ct. 1829, 108 L.Ed.2d 958 (1990) (suspects were loading a footlocker suspected to contain drug paraphernalia into pickup); United States v. Wentz, 686 F.2d 653, 655 (8th Cir.1982) (suspect repeatedly drove automobile to known drug house); United States v. McGlynn, 671 F.2d 1140, 1146 (8th Cir.1982) (suspect retrieved thick envelope from ear immediately before participating in drug deal), and I think that the court goes one step too far by holding that no such connection is required. Today’s decision essentially gives authorities license to search any car that is owned by a person suspected of dealing drugs.
I also disagree with the court’s suggestion that Martinez’s car might have been subject to search under the authority of the warrant issued for his residence. Although a “vehicle found on a premises ... is considered to be included within the scope of a warrant authorizing a search of that premises,” United States v. Reivich, 793 F.2d 957, 963 (8th Cir.1986), I do not think that Martinez’s Buick, which was parked in the apartment building parking lot, was “on the premises” of Martinez’s apartment. In my opinion, the principle relied on in Reivich can reasonably extend only to the garage of a residence, or, perhaps, in a proper case, to its driveway.
The government argues that even if the warrant was improperly issued, the evidence is nevertheless admissible under the principles announced in United States v. Leon, 468 U.S. 897, 104 S.Ct. 3405, 82 L.Ed.2d 677 (1984). The Court in that case held that suppression of the evidence is not called for if the officers executing an invalid search warrant reasonably believed that it was in fact valid. But the ease for the objective reasonableness of the officers’ actions in the present circumstances cannot be made. That is because the application for the warrant should have indicated that Radar was not a particularly good drug dog and the police knew it: In fact, on the day in question, he was one for twelve. Before alerting on the Buick, Radar had already alerted twelve times, and, although he did detect a small amount of marijuana in a suitcase, he also alerted on (among other things) money, a shotgun, an empty cabinet, a kitchen chair, some papers marked “tax info”, a sofa, a pillow, and a television. No drugs were found in any of these places.
We have held, it is true, that “there is no legal requirement that the affidavit [supporting a warrant application] specify the number of times the dog previously has sniffed out drugs.” United States v. Maejia, 928 F.2d 810, 815 (8th Cir.1991). In all our previous cases, however, the application did include at least some indication of the dog’s skill. See United States v. Delaney, 52 F.3d 182, 188 (8th Cir.1995) (“The dog ... had alerted *403more than fifty times in situations where drugs were found.”); Maejia, 928 F.2d at 815 (“The dog had been used in past with successful results.”); United States v. Moore, 911 F.2d 140, 145 (8th Cir.1990) (noting “positive alert for narcotics given by the certified canine drug detection unit.”).
But whatever may be the law with respect to whether a warrant application must always describe a dog’s training or reliability (this is the rule in the Ninth Circuit, see United States v. Lingenfelter, 997 F.2d 632, 689 (9th Cir.1993)), authorities emphatically do have a duty to inform the magistrate if a drug dog is unreliable. See, e.g., United States v. Ludwig, 10 F.3d 1523, 1528 (10th Cir.1993) (suggesting a poor record might lead magistrate to find no probable cause). In this ease, the police knew that Radar had an extraordinarily bad track record, yet they neglected to mention that very pertinent fact in the warrant application.
In United States v. Jacobs, 986 F.2d 1231, 1235 (8th Cir.1993), we invalidated a search warrant because the police failed to tell the magistrate relevant information about a drug sniff. The police said that a drug dog “exhibited an interest” in a package, but they did not tell the magistrate that the dog failed to alert on the package. Id. at 1233. We held that the police violated Franks v. Delaware, 438 U.S. 154, 155-56, 98 S.Ct. 2674, 2676-77, 57 L.Ed.2d 667 (1978), because the omission was made with at least a reckless disregard for whether it made the warrant application misleading, and we further held that the evidence had to be suppressed because the application did not establish probable cause when it was supplemented with the omitted material. Id. at 1234-35.
I believe that the same flaw infects the warrant issued in this case. When the police decided not to include information about Radar’s unreliability, they exhibited, as a matter of law, a reckless disregard for whether the omission made the application misleading; and suppression is required because when the omission is supplied the application does not support a finding of probable cause. The knowledge that the police had of Radar’s unreliability, moreover, made it objectively unreasonable for them to believe that the warrant was valid, and thus the government cannot avail themselves of the principle established in Leon. See Leon, 468 U.S. at 923, 104 S.Ct. at 3420-21; see also Jacobs, 986 F.2d at 1235. While an argument might be made that Leon protects evidence if a reasonable officer could have believed that the warrant application, excluding the information about Radar’s activities, supported a finding of probable cause, the government does not in fact make this argument, and I would in any case be hesitant to parse the officers’ motives quite so finely in circumstances that support a finding of willful deception on their part. “Good faith is not a magic lamp for police officers to rub whenever they find themselves in trouble.” United States v. Reilly, 76 F.3d 1271 (2d Cir.1996) (Calabresi, J.).
I would therefore reverse the district court’s denial of Mr. Martinez’s motion to suppress.