OPINION
STRAS, Justice.The appellant, Lincoln Lamar Caldwell, challenged his conviction of first-degree *768premeditated murder for the benefit of a gang in his third petition for postconviction relief, in which he alleged that three witnesses presented false testimony at his trial. The postconviction court summarily denied the petition. On appeal, Caldwell argues that the court abused its discretion when it failed to grant him an evidentiary hearing in connection with his petition. Because we conclude that Caldwell has alleged facts that, if proven, would entitle him to relief, we reverse and remand to the postconviction court for an evidentiary hearing.
I.
Kirk Harrison (“Kirk”) shot and killed Brian Cole from an SUV driven by Caldwell. Both Caldwell and Kirk were members of the LL gang, which was engaged in an ongoing rivalry with the One-Nine gang. Cole was not a member of a gang, but he was standing near members of the One-Nine gang when Kirk shot him.
On the day of the murder, Caldwell drove an SUV with five passengers, including Kirk. According to the trial testimony of Kirk’s brother, Carnell Harrison (“Car-nell”), who was also a passenger in the SUV that day, Caldwell handed a gun to Kirk. Later, while Caldwell was driving the SUV in north Minneapolis, the group recognized members of the One-Nine gang in a crowd of 10 to 20 people. Caldwell drove around the block, parked the SUV, and then either Caldwell or Kirk said, “we can catch them over there.” Each member of the group, except Carnell, walked toward the crowd. Carnell assumed that Caldwell and Kirk were going to “shoot at [the One-Nines].” A few minutes later, the group of men returned to the SUV, at which point Caldwell and Kirk were arguing. Kirk said, “[y]ou should have took the shot,” and Caldwell responded, “it was too crowded.”
Caldwell then drove the SUV around the block again. Another passenger, William Brooks, testified that Caldwell and Kirk had remarked that members of the One-Nine gang had shot at them previously and that they “were going to get them.” According to Brooks, Caldwell then passed a gun to Kirk. Other witnesses testified that, at some point, a passenger in the SUV pointed out the One-Nines, who were standing in a group of people. Caldwell then drove toward the group. When the SUV was near the One-Nines, Kirk leaned out of the window and attempted to fire the gun. Initially, the gun did not fire because of the safety mechanism. Once Kirk disengaged the safety, he fired six or seven shots into the crowd, one of which hit Cole and killed him. Caldwell and his passengers fled the scene in the SUV.
An acquaintance of Caldwell, S.T., testified at trial that he encountered Caldwell at a friend’s house shortly after the shooting. According to S.T., Caldwell recounted “g[etting] down” — which S.T. understood as “fighting or shooting or [a] brawl or something” — with the One-Nines. Caldwell also explained that the “intended target” had been “Ill Will,” one of the leaders of the One-Nine gang. From Caldwell’s description of the gun — a “grey and black Smith & Wesson” 9mm semiautomatic pistol — S.T. recognized it as one that he had seen in Caldwell’s possession on several occasions.
Before Cole’s funeral, Caldwell encountered one of Cole’s friends, who wore a t-shirt printed with an image of the newspaper article reporting Cole’s death. Caldwell told the friend, “I’m the reason why you got that shirt.” Later, Caldwell told an associate that he originally thought that the victim of the shooting was a member of the One-Nine gang with whom Caldwell had a particularly strong rivalry. Referring to killing Cole, Caldwell stated, “we *769shouldn’t have did it.... Dude was a nobody.”
A grand jury indicted Caldwell on six counts of murder as both a principal and an accomplice. See Minn.Stat. § 609.05, subd. 1 (2012). A jury found Caldwell guilty of all six counts. The district court convicted him of the most serious offense-first — degree premeditated nlurder for the benefit of a gang, Minn.Stat. §§ 609.185(a)(1), 609.229 (2012) — apd sentenced him to life in prison without the possibility of release. Caldwell filed a direct appeal, which we stayed while he filed two petitions for postconviction relief. The postconviction court denied both petitions, after which we consolidated Caldwell’s three appeals. In the consolidated appeal, we affirmed Caldwell’s conviction and the denial of both petitions for post-conviction relief. State v. Caldivell, 803 N.W.2d 373, 377 (Minn.2011).
Caldwell subsequently filed a third petition for postconviction relief. In that petition, Caldwell alleged that Brooks, Cornell, and S.T. each testified falsely at his trial. He supported the petition with a statement by each witness and a signed and notarized affidavit in which the investigator who interviewed the witnesses affirmed that each statement was a true and correct transcription of his recorded interview with the witness. Caldwell also offered an undated handwritten note signed “William Brook.”
In his transcribed statement, Bfooks denied the following key facts from his trial testimony: that he heard Caldwell or Kirk say that the One-Nines had shot at them previously; that he heard Caldwell or Kirk say that they “were going to get” the One-Nines in retaliation for the previous incident; and that he saw Caldwell pass a gun to Kirk. As for the reason that Brooks’s trial testimony differed frpm his statement, Brooks explained that he feared criminal liability for his involvement in the shooting. He was also upset with the others because “[a]ll these babies and kids and dogs and all types of people” had been present during the shooting. According to Brooks, he recanted because he felt bad that Caldwell “g[ot] life for something he didn’t really do.” The handwritten note stated that Brooks lied at trial in an attempt to make a deal with the prosecutor and get out of jail, but in his transcribed statement, Brooks specifically denied the facts from the handwritten note.
In his statement, Carnell denied several key facts from his trial testimony: that he saw Caldwell pass a gun to Kirk; that he heard members of the group discuss trying to “catch” the One-Nines on foot; and that he remembered many of the details from the shooting. Like Brooks, Carnell explained that his account differed at trial because he felt pressure to avoid criminal charges and that he came forward to prevent Caldwell from remaining in prison for a crime that he did not commit.
S.T. declared in his statement that he invented his trial testimony. More specifically, S.T. denied speaking to Caldwell at a friend’s house shortly after the murder. In fact, S.T. stated that he did not see Caldwell at all in the aftermath of the shooting. S.T. also specifically denied seeing Caldwell with the gun Kirk used to shoot Cole, which contradicted S.T.’s testimony at trial that the gun belonged to Caldwell. S.T. explained that he testified falsely in exchange for a suspended sentence on an unrelated charge. He came forward because, in his view, the police took advantage of him.
The postconviction court was not reasonably well satisfied that Brooks’s and Car-nell’s trial testimony was false, nor was it reasonably well satisfied that S.T.’s testimony, even if it was false, might have affected the verdict. Accordingly, the *770court denied Caldwell’s petition without an evidentiary hearing.
II.
The question presented by this case is whether the postconviction court erred when it denied Caldwell’s request for an evidentiary hearing. We review the ultimate decision by the postconviction court to grant or deny an evidentiary hearing for an abuse of discretion. See Riley v. State, 819 N.W.2d 162, 167 (Minn.2012). In doing so, we review the postconviction court’s underlying factual findings for clear error and its legal conclusions de novo. Martin v. State, 825 N.W.2d 734, 740 (Minn.2013).
A postconviction petitioner is entitled to an evidentiary hearing “[u]nless the petition and the files and records of the proceeding conclusively show that the petitioner is entitled to no relief.” Minn. Stat. § 590.04, subd. 1 (2012). In the context of witness-recantation claims, we have interpreted the statutory rule to impose two requirements. First, the allegations in a petition must be “more than argumentative assertions without factual support.” Beltowski v. State, 289 Minn. 215, 217, 183 N.W.2d 563, 564 (1971). Stated differently, the allegations in the petition must have factual support that carries “sufficient indi-cia of trustworthiness,” Ferguson v. State (Ferguson II), 779 N.W.2d 555, 560 (Minn. 2010), to “justify the expense and risk” of an evidentiary hearing. State v. Ferguson (Ferguson I), 742 N.W.2d 651, 660 (Minn. 2007); see also Miles v. State, 800 N.W.2d 778, 784 (Minn.2011) (applying the indicia-of-trustworthiness standard from witness-recantation claims to claims involving newly discovered evidence). Second, the sufficiently trustworthy allegations must recite facts that would, if proven by a preponderance of the evidence, entitle the petitioner to a new trial. See, e.g., Martin, 825 N.W.2d at 740; see also Minn.Stat. § 590.04, subd. 3 (2012) (setting the burden and standard of proof at evidentiary hearings). We address each of these requirements in turn.1
A.
We begin with the first requirement, which is that Caldwell’s evidence must bear “sufficient indicia of trustworthiness” to justify an evidentiary hearing. In this case, Caldwell has submitted statements from Brooks, Carnell, and S.T. that recant portions of their trial testimony. A signed and notarized affidavit from the investigator who interviewed each witness accompanies the statements and affirms that the interviews occurred and that the transcriptions are true and correct.
Often, a postconviction petitioner will submit an affidavit from a recanting witness in support of a petition, see, e.g., Ferguson II, 779 N.W.2d at 558, but we have recognized that other forms of evidence, including a sworn affidavit from a third party, can provide sufficient indicia of trustworthiness to justify an evidentiary hearing. In Dobbins v. State, for example, we held that a petitioner was entitled to an evidentiary hearing based on an affidavit from a third party who allegedly had heard the prosecution’s key witness confess to the crime of which the petitioner *771had been convicted. 788 N.W.2d 719, 732-34 (Minn.2010). The factual support for the petition in Dobbins was sufficiently trustworthy because the petitioner “submitted a sworn affidavit from” the third party. Id. at 734; see also Ferguson v. State, 645 N.W.2d 437, 446 (Minn.2002); cf. Bobo v. State, 820 N.W.2d 511, 514-15, 520 (Minn.2012) (remanding for a postconviction evidentiary hearing based on newly discovered evidence in the form of third-party affidavits stating that an alleged alternative perpetrator had confessed to committing the crime). Similarly, in Op-sahl v. State, we remanded for an eviden-tiary hearing based, in part, on “affidavits of individuals who claim[ed] to have heard certain state witnesses recant trial testimony.” Opsahl v. State (Opsahl I), 677 N.W.2d 414, 419, 424 (Minn.2004). We treated the third-party affidavits identically to first-party affidavits in which witnesses had recanted their own testimony. Id. at 419, 423-124.
There is no logical or legal reason to treat Caldwell’s evidence any differently than the sworn affidavit of a third party who has heard a recantation. Similar to the affidavits in Dobbins and Opsahl I, the sworn affidavit of the third party in this case — an investigator who interviewed all three witnesses — affirms, under the penalty of perjury, that the investigator heard each of the witnesses recant their trial testimony. To be sure, each of the recantations occurred in a more formal — and thus potentially more intimidating — setting here than in Dobbins, but we have never suggested that the trustworthiness of a recantation depends on the formality of the setting in which it occurs. Nor are third-party affidavits that summarize a witness’s statement, such as those in Dobbins or Opsahl I, any more trustworthy than a contemporaneous recording and transcription that perfectly reproduces the recanting witness’s actual statement. On the contrary, the type of evidence that Caldwell submitted with his petition in this case is likely to be more trustworthy than the third-party affidavits in Dobbins and Opsahl I, cases in which it was necessary to assume that an intermediary accurately reported the words of another person.
The handwritten note, by contrast, lacks sufficient indicia of trustworthiness. It is unsworn and undated, does not state which parts of Brooks’s trial testimony were false, and twice misspells his name. Although it offers an explanation for Brooks’s allegedly false trial testimony — to make a deal with the prosecutor to get out of jail — Brooks’s transcribed statement denied the explanation in the note. Such a denial, in a more recent, detailed, and trustworthy statement, confirms the lack of trustworthiness of the note. We therefore disregard the allegations contained in the note in considering whether Caldwell’s allegations entitle him to an evidentiary hearing.
B.
We turn now to the second requirement, which is that Caldwell’s petition must allege facts that would, if proven by a preponderance of the evidence, entitle Caldwell to a new trial. See Martin, 825 N.W.2d at 740. Whether Caldwell is entitled to an evidentiary hearing turns on the legal standard governing the particular type of claim alleged in the petition. See id.; Bobo, 820 N.W.2d at 516-517 (explaining that “[t]o receive an evidentiary hearing on a postconviction claim of ineffective assistance of appellate counsel, a defendant is required to allege facts that ... would satisfy the two-prong test announced in Strickland v. Washington /466 U.S. 668, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984) ],” while “[t]o receive an evidentiary hearing on a ... postconviction claim of newly discovered evidence, a defendant is required to allege facts that ... would *772satisfy the four-prong test set forth in Rainer v. State ”). Here, Caldwell alleges that three witnesses testified falsely, which is the type of allegation that we evaluate under the Larrison standard. See State v. Caldwell, 322 N.W.2d 574, 584-87 (Minn. 1982) (adopting the test set forth in Larrison v. United States, 24 F.2d 82, 87-88 (7th Cir.1928)); see also Pippitt v. State, 737 N.W.2d 221, 227 (Minn.2007) (explaining that the Larrison test applies to witness recantations and, “more generally, ... ‘when a court reviews an allegation that false testimony was given at trial’ ” (quoting Dukes v. State, 621 N.W.2d 246, 257 (Minn.2001))).
Under Larrison, a petitioner is entitled to a new trial based on false trial testimony if: (1) the court is reasonably well satisfied that the testimony given by a material witness was false; (2) without the false testimony, the jury might have reached a different conclusion; and (8) the petitioner was taken by surprise when the false testimony was given and was unable to meet it or did not know that the testimony was false until after trial. Caldwell, 322 N.W.2d at 584-85. The first two prongs of the Larrison standard are compulsory. See Ferguson, 645 N.W.2d at 444. The third prong is relevant but is not an “absolute condition precedent” to granting relief. Id. at 445.
Caldwell does not need to satisfy the Larrison standard at this stage of the proceeding. Rather, to determine whether Caldwell is entitled to an evidentiary hearing, we assume the truth of his allegations that bear sufficient indicia of trustworthiness and determine whether those allegations would be legally sufficient to entitle him to relief if they were proven at a hearing. See Martin, 825 N.W.2d at 740. If so, then Caldwell is entitled to a hearing. See id.; Minn.Stat. § 590.04, subd. 1 (requiring a postconviction court to grant an evidentiary hearing “[ujnless the petition and the files and records of the proceeding conclusively show that the petitioner is entitled to no relief’). If not, then the postconviction court did not abuse its discretion when it denied Caldwell’s petition without an evidentiary hearing. See State v. Nicks, 831 N.W.2d 493, 511 (Minn.2013). Accordingly, in a case involving allegedly false trial testimony, a petitioner is entitled to an evidentiary hearing if, assuming that the trustworthy allegations contained in the petition, files, and records are true, a court would conclude that a material witness’s trial testimony was false and that the false testimony might have affected the verdict.2
1.
The first prong of the Larrison standard focuses on whether a material witness’s testimony at trial was false. An evidentiary hearing is ordinarily required to determine the credibility of a recantation because a postconviction court must assume the truth of the allegations in a petition when it determines whether to grant an evidentiary hearing. See Ferguson II, 779 N.W.2d at 559-60. Stated differently, an evidentiary hearing is the means by which a court generally must *773determine the credibility of a withess. See Bobo, 820 N.W.2d at 516.
In this case, the postconviction court concluded that it was “not reasonably well satisfied by a preponderance of the evidence” that Brooks’s or Carnell’s trial testimony was false. Though the court recited the correct legal standard for evaluating Caldwell’s request for an evi-dentiary hearing, it then failed to apply that standard correctly when it assessed the credibility of Caldwell’s witnesses without the benefit of an evidentiary hearing. See Martin, 825 N.W.2d at 743. The dissent points out that the postconviction court denied making credibility determinations, as if that fact is somehow determinative. The problem, however, is that what the postconviction court did is not the same as what it said. It is not sufficient, as the dissent seems to imply, for a court simply to state the legal standard borfectly. It must also apply that standard correctly. Here, after accdrately stating the standard for an evidentiary hearing, the postconviction court “conflated the requirements for a postbonviction evidentia-ry hearing with the requirements for a new trial” when it evaliiated the credibility of Caldwell’s witnesses without the benefit of an evidentiary hearing. Id. (emphases in original omitted).
To be sure, a postconviction court has discretion to deny an evidentiary hearing on grounds that are unrelated to a recanting witness’s credibility — for example, when an alleged recantatioh does not actually contradict a witness’s trial testimony. Cf. Opsahl v. State (Opsahl II), 710 N.W.2d 776, 782 (Minn.2006) (holding that relief was unavailable in part because some recanting witnesses “reversed any recantation they may have made” and others “claim[ed] merely that they did not give a full explanation of their testimony at trial”). In such a situation, a petitioner would not be entitled to relief, regardless of what occurs at an evidentiary hearing. However, we have been clear that it is impermissible for a court to deny an evi-dentiary hearing in a witness-recantation case based on nothing more than its own speculation about whether the recantation is credible. See Opsahl I, 677 N.W.2d at 423-24. In this case, a review of each witness’s trial testimony and alleged recantation indicates that such speculation, rather than the legal insufficiency of Caldwell’s allegations, drove the court’s decision to deny an evidentiary hearing.
At least two elements of Brooks’s trial testimony — both of which he recanted in his statement — provided material evidence of Caldwell’s guilt as an accomplice. First, Brooks testified at trial that Caldwell or Kirk said the One-Nines had shot at them earlier and that they “were going to get” them in retaliation. Second, Brooks testified at trial that Caldwell passed a gun to Kirk in the moments before the shooting. Brooks’s statement directly contradicts his trial testimony on these two key facts.
The dissent suggests that Brooks’s failure to say definitively, “Caldwell did not pass the gun to the shooter,” undermines our conclusion that Brooks’s recorded statement constituted a recantation of the second point. We disagree. Brooks testified at trial that he saw “[Caldwell] pass[ ] Kirk the gun.” In his recorded statement, when asked if he saw Caldwell pass a gun to Kirk, Brooks answered, “I couldn’t really see anything. Cause if you going towards the door and the chair way you can’t even see him pass him nothing.” If Brooks’s recorded statement is true and Brooks could not actually see what Caldwell was doing, then he could not have seen Caldwell pass the gun to Kirk. In other words, both statements cannot be true, and the contradiction between Brooks’s statement and his trial testimony *774necessitates further inquiry into the credibility of his recantation.3
Moreover, in contrast to cases in which we have upheld a postconviction court’s denial of an evidentiary hearing, here Brooks provided reasons for his recantation: he had lied at trial because he was upset that Kirk fired toward innocent bystanders, and he recanted because he was distressed that Caldwell was in prison for a crime that he did not commit. While Brooks’s explanations for the recantation may be improbable and his statement expresses some uncertainty about the events surrounding the shooting, the postconviction court’s decision to deny the petition was premature.
Caldwell’s allegations regarding Car-nell’s and S.T.’s recantations further support our conclusion that the postconviction court denied the petition prematurely. Like Brooks’s statement, Carnell’s statement repudiates his testimony that he saw Caldwell pass a gun to Kirk on the day of the shooting. Other details in Carnell’s statement — that he either never saw a gun or did not recall seeing a gun and that he did not remember details about the shooting when he testified at trial — are in tension with his trial testimony, but, even if true, would not necessarily imply that he had lied at trial. S.T.’s statement also denies key facts from his trial testimony, including that he saw Caldwell in the aftermath of the shooting and that the gun used to kill Cole was one that belonged to Caldwell. Like Brooks, both witnesses explained their recantations: S.T. said that he had lied at trial in exchange for a suspended sentence on an unrelated charge but later decided that the police had taken advantage of him, and Carnell said that he recanted because he was upset that Caldwell had been convicted for a crime that he did not commit.
The dissent, based largely on its observation that neither witness stated “that they did not testify truthfully at trial,” concludes that Brooks and Carnell did not actually recant their trial testimony. In the dissent’s view, if a witness testified at trial, “A occurred,” but then later stated that “A did not occur, but B did,” we cannot conclude that the first statement is false unless the witness adds, “and by the way, I lied at trial.” Fortunately, the dissent’s overly formalistic view of witness recantations is not the law in Minnesota. See, e.g., Dobbins, 788 N.W.2d at 734-35 (concluding that the first Larrison requirement could be satisfied by a witness’s alleged statement that directly contradicted his trial testimony but did not explicitly admit to testifying falsely). For one thing, witnesses may be understandably reluctant to admit that they testified falsely under oath because perjury is a crime. See MinmStat. § 609.48, subd. 1(1) (2012). Evasive answers by a witness do not necessarily mean that the witness did not testify falsely at trial or recant his or her testimony, as the dissent suggests. Rather, evasiveness is potentially relevant to whether a recantation is credible — a matter that the postconviction court is free to consider at an evidentiary hearing. See Bobo, 820 N.W.2d at 516 (describing an evidentiary hearing as “the means for evaluating the credibility of a witness” (emphasis added)).
Moreover, even if the dissent’s strict definition of what constitutes a “recanta*775tion” were correct, the Larrison standard applies broadly to all allegations of false trial testimony, not just to witness recantations, Pippitt, 737 N.W.2d at 227, so the dissent’s metaphysical discussion of recantations is ultimately beside the point. Thus, no matter how we label Caldwell’s allegations-as allegedly false testimony or as recantations — the Larrison standard requires us to examine the substance of the witnesses’ statements, rather than how the witnesses characterize their statements, to determine whether they may have testified falsely at trial. See id. (holding, following an evidentiary hearing, that the first Larrison prong was not satisfied when a witness claimed that he had testified falsely at trial but failed to provide any details or otherwise demonstrate which parts of his testimony were false). It is the substance of the conflicting statements of Carnell, Brooks, and S.T. that convinces us that Caldwell has made a sufficient showing that each of the three witnesses may have testified falsely at trial.4
The dissent nevertheless argues that mere contradictions between a witness’s trial testimony and a later statement are insufficient to qualify as a recantation because we ultimately affirmed the denial of postconviction relief in both Pippitt and Opsahl II. Pippitt, 737 N.W.2d at 224; Opsahl II, 710 N.W.2d at 782. The dissent’s argument, however, overlooks the fact that those cases involved credibility determinations by the postconviction court after an evidentiary hearing.5 Pippitt, 737 *776N.W.2d at 224; Opsahl II, 710 N.W.2d at 780-81. Indeed, prior to Opsahl II, we had reversed the postconviction court’s summary denial of Opsahl’s petition because his allegations “m[et] the minimal standard for an evidentiary hearing.” Op-sahl I, 677 N.W.2d at 423-24. Thus, the procedural history of Pippitt and Opsahl II, both of which were decided following an evidentiary hearing, supports our conclusion that the postconviction court denied Caldwell’s petition prematurely. Moreover, not only does the dissent’s argument misinterpret Pippitt and Opsahl, it also conflicts with a number of other decisions in which we have remanded for an eviden-tiary hearing based on allegations that “if true, would suggest that [a witness’s] trial testimony was false.” Ferguson II, 779 N.W.2d at 559 (emphasis added); see also Dobbins, 788 N.W.2d at 735 (remanding for an evidentiary hearing after observing that, “[b]ecause [the witness’s] trial testimony and his subsequent statements to [an informant] are inconsistent, one or the other is necessarily false”); Wilson v. State, 726 N.W.2d 103, 107 (Minn.2007) (“In this case, an evidentiary hearing is appropriate because it is difficult if not impossible to test [the witness’s] conflicting statements without examining [the witness] under oath.”).
On this record, therefore, because we conclude that Caldwell’s allegations, if proven, would be sufficient to satisfy the first prong of the Larrison standard, the postconviction court abused its discretion when it denied an evidentiary hearing to Caldwell based in part on its conclusion that the recantations from Brooks and Carnell were not credible.
2.
The second prong of the Larrison standard focuses on the impact of the false testimony on the verdict. Again, in the context of a request for an evidentiary hearing, we assume the truth of the petitioner’s allegations, and therefore consider whether “the jury might have reached a different conclusion” absent the allegedly false testimony.6 Martin, 825 N.W.2d at 740, 743. By “might,” we mean “something more than an outside chance although much less than ... would probably.” Caldwell, 322 N.W.2d at 585 n. 8 (internal quotation marks omitted). Rather than isolating each instance of allegedly false trial testimony, as under the first Larrison prong, the second prong examines the combined impact on the verdict of all of the allegedly false trial testimony. See, e.g., Martin, 825 N.W.2d at 744 (considering the evidence without the testimony of two recanting witnesses); Opsahl I, *777677 N.W.2d at 424 (remanding for an evi-dentiary hearing because allegations that “challenge[d] the truth or believability of five out of seven witnesses” “call[ed] into question ... a significant part of the state’s ... case”).
The dissent repeatedly characterizes the changes in Brooks’s and Cainell’s testimony from the trial to their recorded statements as “minor.” The dissent’s characterization is unsupported by the record,7 and more importantly, reflects a fundamental misunderstanding of the latv of accomplice liability. To return a guilty verdict, the jury had to find, among other things: “(1) that [Caldwell] ‘knew that [Kirk] w[as] going to commit a criine,’ and (2) that [Caldwell] ‘intended his presence or actions to further the commission of that crime.’ ” State v. Milton, 821 N.W.2d 789, 805 (Minn.2012) (quoting State v. Mahkuk, 736 N.W.2d 675, 682 (Minn. 2007)). The critical evidence proving those two elements was the testithony of Brooks and Carnell that Caldwell passed a gun to Kirk before the shooting. Given the critical nature of their testimony and the dearth of other comparable evidence bearing on Caldwell’s mental state, Caldwell’s allegations, if proven, would be sufficient to determine that the jury may well have reached a different conclusion in the absence of the allegedly fdlse testimony.
In addition to characterizing the changes in testimony as “minor,” the dissent also claims that the testimony of other witnesses provided “ample” and even “abundant” evidence that Caldwell was guilty as an accomplice. Once again, the dissent misunderstands the law of accomplice liability, which required the State to prove beyond a reasonable doubt that Caldwell knew that Kirk was going to commit a crime and that Caldwell intended for his presence or actions to assist Kirk in the commission of that crime, see Milton, 821 N.W.2d at 805. The testimony relied upon by the dissent shows Caldwell’s general involvement in the events before and after the shooting, but it bears only indirectly, and tenuously, on his mental state with regard to the shooting itself.
The dissent places particular emphasis on testimony that Caldwell’s gang was engaged in a conflict with the One-Nines, that Caldwell was driving the SUV at the time of the shooting, that he drove toward the One-Nines and then drove away after Kirk fired the gun, and that Caldwell later took responsibility for Cole’s death. The testimony establishes that Caldwell was a gang member, he was present for the shooting, and even that he may have bragged to others about his gang exploits. It does not establish, however, that Caldwell knew that Kirk would commit a crime or that he intended his actions to further the commission of that crime, except through several layers of speculation. Brooks’s and Carnell’s testimony, by contrast, provided a direct connection between Caldwell’s actions (and by extension, his mental state) and the actual crime that Kirk committed. None of the evidence relied upon by the dissent can provide such a connection.
In fact, without the allegedly false trial testimony of the three witnesses, the jury would have heard no evidence that Caldwell provided a gun to the shooter. The dissent apparently considers it significant that “one [could] reasonably infer” that *778Caldwell passed a gun to Kirk based on the events leading up to the shooting. But as we have explained, “the second Larrison prong does not ask whether the evidence was sufficient to convict the defendant in the absence of the recanted testimony. The question instead is whether the jury might have found the defendant not guilty if the recanting witness had not testified [falsely].” State v. Tumage, 729 N.W.2d 593, 599 (Minn.2007) (emphasis added); accord Martin, 825 N.W.2d at 744. Therefore, regardless of whether the dissent’s inference is reasonable, the important fact is that the jury would have needed to rely on such inferences, rather than the direct evidence provided by Brooks’s and Carnell’s testimony, to conclude that Caldwell was even aware that Kirk had a gun. Compare Dobbins, 788 N.W.2d at 735-36 (“Because the State had little other direct evidence against [the petitioner], and [the petitioner’s] theory of the case was credible, we conclude the jury might have reached a different conclusion had [the witness’s] alleged false testimony not been admitted.” (citation omitted) (internal quotation marks omitted)), with Hooper v. State, 680 N.W.2d 89, 96 (Minn.2004) (concluding that a recantation regarding a confession did not satisfy the second prong of the Larrison test because two other witnesses also heard the defendant confess, a witness who was present during the crime testified about the defendant’s actions, and several pieces of physical evidence directly connected the defendant to the crime).
The allegedly false trial testimony thus provided a basis for the jury’s conclusion that Caldwell “knew that [Kirk] w[as] going to commit a crime” and “intended his presence or actions to further the commission of that crime.” Mahkuk, 736 N.W.2d at 682. We therefore conclude that the jury might have reached a different conclusion regarding Caldwell’s guilt if it had not heard the allegedly false testimony from the witnesses at trial.
3.
The final, noncompulsory prong of the Larrison standard focuses on the petitioner’s knowledge of the alleged falsity of the testimony at the time of trial. Because Caldwell was indisputably present for the events that formed the basis for the allegedly false testimony, he does not argue that he was taken by surprise by the false testimony at trial, that he was unable to counter it at trial, or that he did not know of its falsity until after trial. Nevertheless, we have held that a petitioner’s failure to allege facts that would satisfy Lar-rison’s third prong does not prevent a petitioner from receiving an evidentiary hearing or a new trial. See, e.g., Ferguson II, 779 N.W.2d at 561-62. Thus, Caldwell’s failure to allege such facts does not “conclusively show that [he] is entitled to no relief,” and does not justify the postcon-viction court’s denial of his request for an evidentiary hearing. MinmStat. § 590.04, subd. 1.
III.
For the foregoing reasons, we reverse the decision of the postconviction court and remand for an evidentiary hearing on Caldwell’s witness-recantation claim.
Reversed and remanded.
. While a postconviction petitioner must satisfy both requirements to be entitled to an evidentiary hearing, we do not mean to suggest that a court must address both requirements in order to deny an evidentiary hearing. For example, a postconviction court may deny an evidentiary hearing if the petitioner's allegations are legally insufficient without first addressing whether the petitioner’s allegations bear sufficient indicia of trustworthiness. Conversely, a postconviction court may deny an evidentiary hearing if the petitioner's allegations do not bear sufficient indicia of trustworthiness without also addressing whether the petitioner’s allegations are legally sufficient.
. According to the dissent, we "contend” that appellate courts must independently review the record and decide de novo if an alleged recantation justifies an evidentiary hearing. Of course, we contend no such thing. Instead, as our cases require, we evaluate the postconviction court’s decision to deny an evidentiary hearing to Caldwell under an abuse-of-discretion standard. In this case, the postconviction court made an obvious error and thus abused its discretion when it concluded that the allegations in Caldwell’s petition and the accompanying materials were insufficient to justify an evidentiary hearing.
. The fact that Brooks substantively changed his testimony between the trial and his recorded statement distinguishes this case from those in which witnesses have merely forgotten facts from their testimony. It is one thing for a witness to forget key facts from his or her trial testimony, but it is entirely different for a witness to contradict key facts from his or her trial testimony, as our cases acknowledge. See Dobbins, 788 N.W.2d at 735 ("Because [the witness's] trial testimony and his subsequent statements to [an informant] are inconsistent, one or the other is necessarily false." (emphasis added)).
. In addition to holding a misguided view of what constitutes a recantation, the dissent ignores relevant portions of the record when it argues that neither Carnell nor Brooks admitted that his trial testimony was false. For example, when the investigator asked Carnell if his testimony had been truthful, Carnell said "some of it was but like some of it wasn’t.” (Emphasis added.) Similarly, Brooks agreed with the investigator that the interview was scheduled only after Brooks "indicat[ed] that the testimony that [he] gave in Lincoln Caldwell’s case may have been inaccurate.” Brooks also stated that he had "change[d] [his] mind” about some of his testimony at trial.
. In fact, the postconviction court’s credibility determination was central to our decision in Pippitt. In Pippitt, the postconviction court relied on a witness’s temporary inability to remember portions of his trial testimony at an evidentiary hearing to conclude that the witness, and thus his recantation, was not credible. Pippitt, 737 N.W.2d at 228-29. We affirmed the postconviction court’s decision because the witness’s selective recall at the hearing was a sufficient reason to doubt the credibility of the recantation. Id. Thus, far from constituting a "distinction without a difference,” as the dissent contends, the fact that the postconviction court held an eviden-tiary hearing — which is "the means for evaluating the credibility of a witness,” Bobo, 820 N.W.2d at 516 (emphasis added) — was essential to Pippitt.
Nevertheless, the dissent insists that Pippitt held that "a recantation requires more than a mere inability to remember or a simple statement contradicting earlier trial testimony.” Notably, both parts of the dissent's characterization of Pippitt are at odds with the reasoning of the decision. After all, the witness in Pippitt ultimately did remember and testified at the evidentiary hearing, so his initial failure to remember was only relevant to his credibility and there would have been no basis on which to adopt a categorical rule of the type articulated by the dissent. See Til N.W.2d at 228 ("[The witness] initially stated at the hearing that he could not remember whether the door even had a dead bolt. When shown a picture of the door, he reported his memory being refreshed, and confirmed that the door had a dead bolt and that his mother could lock it.” (emphasis added)). And Pippitt's abbreviated discussion of the witness’s contradictory statement at the evidentiary hearing was nothing more than a reiteration of the straightforward principle that a witness’s uncertainty and hazy memory during his or her testimony is a sufficient basis to doubt the witness's credibility. See id. at 228-29 ("When we compare the level of detail [the witness] provided about the front door during his trial testimony with his inability to re*776member, at the postconviction hearing, whether the door even had a dead bolt, we cannot be reasonably 'well-satisfied' that his trial testimony was false_”). Our discussion cannot be read, as the dissent contends, as establishing a categorical rule that a contradiction between a witness’s later statements and his or her trial testimony cannot constitute a recantation. See id. at 228 ("While that portion of [the witness's] post-conviction testimony conflicts with his trial testimony, the conflict does not necessarily indicate that his trial testimony was false for purposes of the Larrison test.” (emphasis added)).
. In addressing the impact on the verdict, Caldwell argues that we should also consider new or alternative testimony from a witness. We disagree. We limit our analysis under Larrison’s second prong to the impact of the allegedly false testimony on the verdict, excluding from consideration the impact of any new or alternative testimony by a witness. See, e.g., Caldwell, 322 N.W.2d at 585. Indeed, the limited focus of the Larrison standard distinguishes it from the Rainer standard, which applies to claims based on other types of newly discovered evidence, including new or alternative testimony from a witness. Cf. Rainer v. State, 566 N.W.2d 692, 695 (Minn. 1997). Caldwell does not seek relief under the Rainer standard.
. Nowhere is the speculation of the dissent more apparent than in footnote 22, which provides a long and convoluted explanation, full of assumptions and speculation, of how Brooks’s testimony and his recorded statement may be less inconsistent than we suggest. A far simpler — and more sensible— explanation for Brooks's two contradictory accounts of the shooting is that either Brooks is not telling the truth now or he was not telling the truth at trial.