United States v. Melvin Knight

981 F.3d 1095

• Court: Court of Appeals for the D.C. Circuit
• Decided: December 8, 2020
• Docket: 19-3016
• Status: Published

Opinion

United States Court of Appeals FOR THE DISTRICT OF COLUMBIA CIRCUIT

Argued January 17, 2020 Decided December 8, 2020

No. 19-3016

UNITED STATES OF AMERICA, APPELLEE

v.

MELVIN KNIGHT, APPELLANT

Consolidated with 19-3017

Appeals from the United States District Court for the District of Columbia (No. 1:13-cr-00131-2) (No. 1:13-cr-00131-1)

Howard B. Katzoff, appointed by the court, argued the cause for appellant. With him on the briefs was Mary E. Davis, appointed by the court.

Bryan H. Han, Assistant U.S. Attorney, argued the cause for appellee. With him on the brief were Jessie K. Liu, U.S. Attorney, and Elizabeth Trosman, John P. Mannarino, and Pamela S. Satterfield, Assistant U.S. Attorneys. 2 Before: ROGERS, WILKINS, and KATSAS, Circuit Judges. *

Opinion for the Court by Circuit Judge ROGERS.

Opinion dissenting in part by Circuit Judge KATSAS.

ROGERS, Circuit Judge: In 2013, Melvin Knight and Aaron Thorpe were arrested for armed robbery and kidnapping. They were charged by the U.S. Attorney’s Office in the D.C. Superior Court and offered a generous plea deal by the Assistant U.S. Attorney: plead guilty to a single count of assault with a dangerous weapon and no further charges stemming from these crimes would be filed. Under the D.C. Superior Court Sentencing Guidelines, the likely sentences would be between two and six years for each defendant. The plea offer was wired, however, so both Knight and Thorpe had to accept it or it would be withdrawn. Thorpe wanted to accept the plea offer, but Knight, who was erroneously advised by his counsel that the offer came with ten years in prison and never advised by his counsel of the sentencing consequences of rejecting plea the offer, did not. Once they declined the plea offer, the government dismissed the Superior Court charges and prosecuted Knight and Thorpe on a ten-count indictment in federal court. A jury found Knight and Thorpe guilty on all counts, and the U.S. district court sentenced Knight to more than 22 years’ imprisonment and Thorpe to 25 years’ imprisonment.

* Senior Judge Stephen F. Williams was a member of the panel at the time the case was argued and he participated in its consideration before his death on August 7, 2020. Judge Wilkins was randomly selected thereafter to serve as a member of this panel. 3 On direct appeal, Knight and Thorpe both argued that they had been denied effective assistance of counsel in violation of the Sixth Amendment to the U.S. Constitution. This court, concluding that their claims were “colorable,” United States v. Knight, 824 F.3d 1105, 1113 (D.C. Cir. 2016), remanded the case. Following an evidentiary hearing after remand, the district court denied relief. Although agreeing that Knight’s counsel’s performance was deficient, the court determined that Knight had suffered no prejudice. The court rejected Thorpe’s claim that his counsel was deficient and did not address prejudice. Knight and Thorpe appeal.

For the following reasons, we reverse in part. Knight satisfied his burden under both prongs of the standard for an ineffective assistance of counsel claim. See Strickland v. Washington, 466 U.S. 668, 687 (1984). First, as the government acknowledges, the performance by Knight’s counsel did not meet minimal professional standards. Second, the district court’s determination that Knight suffered no prejudice rested on subsidiary factual findings that ignored the direct effect of his counsel’s deficient performance on Knight’s ability to intelligently assess his options and therefore were clearly erroneous. Viewed properly, the contemporaneous evidence and Knight’s testimony at the evidentiary hearing sufficed to establish a reasonable probability Knight would have accepted the plea offer but for his counsel’s ineffective assistance. In contrast, we agree that Thorpe’s counsel was not ineffective and there was no violation of his Sixth Amendment rights. Accordingly, we affirm as to Thorpe and reverse the denial of Knight’s Sixth Amendment challenge, remanding his case to the district court to provide a remedy consistent with this opinion. 4 I.

In January 2013, Knight and Thorpe were involved in an armed robbery and kidnapping of Edmund Peters. They were arrested and each was charged with one count of armed kidnapping in the D.C. Superior Court and appointed separate counsel. Shortly after their arrest, the Assistant U.S. Attorney assigned to their cases sent an email to their counsel: If Knight and Thorpe would plead guilty to one count of assault with a dangerous weapon (“ADW”), then the government would agree not to bring additional and more serious charges, including two counts of armed kidnapping; two counts of possession of a firearm during a crime of violence; two counts of obstruction of justice; a second count of assault with a dangerous weapon, namely assault with a firearm; and one count of felon in possession of a firearm. The plea offer was wired, however, allowing the government to dispose of the charges against both defendants without a trial while preserving its right to prosecution by trial if both did not accept the plea offer, which would be withdrawn. In addition, the plea offer was contingent on Knight and Thorpe also agreeing “[n]ot to seek to modify the conditions of their release pending the plea,” meaning that they could not be released from custody before entering the plea. Id.

Knight’s counsel visited Knight in jail but did not mention the plea offer. The next day, February 1, 2013, the Assistant U.S. Attorney placed the plea offer on the record and the Superior Court judge continued the preliminary hearing until February 19, 2013, to give Knight and Thorpe time to consider whether to accept the plea offer. In fact, Knight’s counsel’s lone interaction with Knight about the plea offer was limited to misinformation. While still in court, Knight asked how much time the government wanted him to serve for the ADW charge, and Knight’s counsel told him “[t]en years.” Hearing Tr. 19 5 (May 24, 2017). Ten years was the statutory maximum for ADW, but the offense had no mandatory minimum and the sentencing range under the Superior Court Sentencing Guidelines was 24–72 months (2 to 6 years). Although counsel told Knight he would visit him in jail to discuss the plea offer further, he never did. Consequently, the brief and misleading exchange in open court was the extent of the advice that Knight received from counsel about the plea offer. Among other things relevant to the plea offer, Knight was never advised of the worst-case scenario were he to reject the plea offer, namely being indicted on additional charges with a greatly increased sentencing exposure in federal court.

Thorpe’s counsel, by contrast, advised his client of the plea offer immediately upon learning of it, prior to the appearance in the Superior Court for the scheduled preliminary hearing. Counsel also visited Thorpe in jail to discuss the terms of the plea offer. Their discussion covered the estimated sentencing range for the ADW charge; potential additional charges that Thorpe would face if he rejected the plea offer and the sentencing consequences; and the fact that the plea offer was wired. Thorpe’s counsel also alerted his client to the fact that he had learned from Knight’s counsel, as the result of a chance meeting in the Superior Court, that Knight was not expected to take the plea. He told Thorp that the trial prosecutor had refused to unwire the plea so Thorpe could plead separately.

At the February 19 preliminary hearing, Thorpe’s counsel stated Knight and Thorpe did not intend to accept the plea offer. The U.S. Attorney’s Office withdrew the plea offer and a trial date was set. Prior to trial, the Superior Court charges were dismissed and a federal grand jury returned a ten-count indictment charging both Knight and Thorpe with six D.C. Code felony offenses and the federal offense of being a felon in possession of a firearm. A jury found them guilty as 6 charged, and the district court sentenced Knight to 268 months’ imprisonment (22 years and 4 months) and Thorpe to 300 months’ imprisonment (25 years).

On direct appeal from their convictions, Knight and Thorpe contended, in part, that each had received ineffective assistance of counsel in the Superior Court regarding the plea offer. See Knight, 824 F.3d at 1109. This court determined that “Knight’s and Thorpe’s claims of ineffective assistance are colorable” and remanded them to the district court. Id. at 1113. After a three-day evidentiary hearing, at which Knight and Thorpe and their counsel testified, the district court denied relief. Although determining that Knight’s counsel had performed deficiently, the court concluded that Knight had not been prejudiced. The court noted the lack of contemporaneous evidence that Knight would have accepted the plea offer in view of evidence that Knight (1) was focused on obtaining pretrial release so that he could be with his wife for the birth of their child, which would not have been possible if he accepted the plea offer; (2) had told counsel he wanted to go to trial as he was hopeful that the victim of the crimes would not testify against him; and (3) had rejected a plea offer on the federal charges. The court also determined that Thorpe’s counsel’s performance was not deficient, rejecting the argument that a wired plea offer required his counsel to meet with Knight’s counsel to discuss the plea offer. The court did not address the issue of prejudice to Thorpe.

Knight and Thorpe appeal, and the court’s review of the district court’s denial of their ineffective assistance of counsel claims is de novo. United States v. Abney, 812 F.3d 1079, 1086–87 (D.C. Cir. 2016). The district court’s subsidiary factual findings are reviewed for clear error. See id. at 1087. 7 II.

“The Sixth Amendment guarantees a defendant the effective assistance of counsel at ‘critical stages of a criminal proceeding,’ including when he enters a guilty plea.” Lee v. United States, 137 S. Ct. 1958, 1964 (2017) (quoting Lafler v. Cooper, 566 U.S. 156, 165 (2012)). To succeed on a claim of ineffective assistance of counsel, a defendant must show that (1) his counsel’s performance was deficient, as judged against prevailing professional norms under the circumstances, and (2) the deficient performance was prejudicial. See Strickland, 466 U.S. at 687. To satisfy the prejudice prong, the defendant must show that there is a reasonable likelihood that the outcome would have been different had the defendant been adequately counselled. See Missouri v. Frye, 566 U.S. 134, 147 (2012).

More specifically, where a defendant maintains that his counsel’s inadequate assistance caused him to proceed to trial when he would otherwise have accepted a plea offer, prejudice means that “but for the ineffective advice of counsel there is a reasonable probability that . . . the defendant would have accepted the plea and the prosecution would not have withdrawn it . . . , that the court would have accepted its terms, and that the conviction or sentence, or both, under the offer’s terms would have been less severe than under the judgment and sentence that in fact were imposed.” Lafler, 566 U.S. at 164. A criminal defendant alleging ineffective assistance of counsel generally may not rely solely on post hoc testimony to show that he would have accepted the plea offer if he had been properly advised. Cf. Lee, 137 S. Ct. at 1967; United States v. Aguiar, 894 F.3d 351, 361–62 (D.C. Cir. 2018). At least where a defendant has pled guilty and is seeking to show a reasonable probability that he would have gone to trial but for counsel’s ineffectiveness, the Supreme Court has instructed that 8 “[j]udges should . . . look to contemporaneous evidence to substantiate a defendant’s expressed preferences.” Lee, 137 S. Ct. at 1967. Knight and Thorpe do not suggest that this principle is inapplicable in their circumstances. Nonetheless, although contemporaneous evidence of the defendant’s preferences may inform the prejudice inquiry, a defendant is not required to have hypothesized, at the time of the plea offer, that his attorney might be providing inadequate assistance and state that his decision whether or not to accept a plea offer would change if that were so. Aguiar, 894 F.3d at 362.

Further, when a plea offer is wired, a defendant attempting to show prejudice “must establish not only that he would have taken the plea offer if his counsel had advised him correctly, but also either that each of his co-defendants would have accepted their respective plea offers, or that the Government would have offered [him] an unwired plea.” United States v. Gaviria, 116 F.3d 1498, 1512 (D.C. Cir. 1997).

A.

The government does not dispute that Knight’s counsel’s performance was constitutionally deficient for failing to correctly inform him of the time he would serve for ADW if he accepted the plea offer. See United States v. Soto, 132 F.3d 56, 59 (D.C. Cir. 1997). Neither does the government dispute that counsel’s performance was constitutionally deficient for failing to advise Knight of the worst-case scenario of declining the plea offer. See Aguiar, 894 F.3d at 361. Nor does the government dispute that Knight’s co-defendant would have accepted the wired plea offer had his acceptance not been foreclosed by Knight’s rejection of the offer. Rather, the government disputes Knight’s contention that he was prejudiced by his counsel’s deficient performance. In the government’s view, Knight has not established Strickland 9 prejudice because he has presented no contemporaneous evidence to show a reasonable probability that he would have accepted the plea offer absent counsel’s errors and, in fact, the contemporaneous evidence that did exist indicated Knight would not have accepted the plea even if adequately counselled.

But the government, like the district court, overlooks the direct negative impact that counsel’s shortcomings had on Knight’s understanding of his circumstances at the time he was deciding whether to accept the plea offer. Counsel’s inaccurate appraisal of pleading to ADW and failure to alert Knight to the worst-case scenario of rejecting the plea offer left Knight unable to make an intelligent decision about whether to accept the plea offer. Knight’s statements and his preferences at the time he rejected the plea offer must be evaluated in view of his erroneous understanding of his circumstances. Further, the government, like the district court, ignores key contemporaneous evidence suggesting Knight may have accepted the plea offer had his counsel performed adequately. For these reasons, the district court’s factual findings underlying its determination that Knight suffered no prejudice are clearly erroneous.

First, the Superior Court plea offer was contemporaneous evidence of a plea offer whose generosity is self-evident from the prosecutor’s email to Knight’s counsel and Thorpe’s counsel. The email set forth the terms of the plea offer and the consequences of its rejection, suggesting that Knight would have accepted the offer had he understood how favorable it was to him and how unfavorable his sentencing exposure would be if he proceeded to trial. The limited exchange that Knight had with counsel at the time of the Superior Court plea offer, described below, also suggests that he might have been amenable to accepting the plea offer even when he wrongly 10 believed it carried a ten-year sentence. Combined with Knight’s after-the-fact testimony on remand about what he would have done had he been correctly and adequately advised by counsel regarding the plea offer, the contemporaneous evidence before the district court suffices to show a reasonable probability that he would have accepted the plea offer if he had been advised of its leniency and the sentencing exposure he would face as a consequence of rejecting it.

The generosity of the plea offer is underscored by the significant disparity in sentencing exposure between the plea offer on the Superior Court charge and the charges that Knight faced in federal court. The Superior Court plea offer required that Knight and Thorpe each plead guilty to only a single count having no statutory minimum sentence, a ten-year maximum, and a Superior Court Sentencing Guidelines range of two to six years. By proceeding to trial Knight risked a ten-count indictment in federal court, dramatically greater sentencing exposure, and an actual imposed sentence of more than twenty- two years. The prosecutor’s email to counsel forewarned of these consequences yet Knight’s counsel never shared that information with him. Even absent such forewarning, counsel is obligated to advise a client facing criminal charges of what the law, including sentencing guidelines, makes “clear” and is “‘easily determined’ by competent counsel.” Aguiar, 894 F.3d at 359 (quoting standard announced by Supreme Court in Padilla v. Kentucky, 559 U.S. 356, 368–69 (2010)); see also Padilla, 559 U.S. at 365, 370. Because Knight’s counsel did not render adequate assistance on key considerations before Knight, Knight was not in a position to appreciate the generosity of the plea offer or realistically evaluate the consequences of rejecting it. This significant sentencing disparity is contemporaneous evidence that Knight would have accepted the plea offer had counsel correctly apprised him of how favorable it was and of the sentencing exposure he would 11 face if he declined the offer and went to trial. See Gaviria, 116 F.3d at 1513. Indeed, both the government and the district court characterized the plea offer as “incredibly sweet.” Hearing Tr. 50 (May 25, 2017).

Other circuits have recognized that a disparity in sentencing exposure may suffice to show prejudice under the second prong of Strickland. See United States v. Herrera, 412 F.3d 577, 581 (5th Cir. 2005); Griffin v. United States, 330 F.3d 733, 737–38 (6th Cir. 2003); United States v. Day, 969 F.2d 39, 45–46 (3d Cir. 1992). Although those decisions predate the Supreme Court’s decision in Lee, the Supreme Court did nothing to undermine the commonsense conclusion that a disparity in sentencing exposure is relevant to the prejudice inquiry. Indeed, even after Lee, our sister circuits have continued to view a severe disparity between the plea offer and sentence faced by proceeding to trial as compelling evidence that the defendant would have accepted a plea offer but for counsel’s constitutionally deficient performance. See, e.g., Dodson v. Ballard, 800 F. App’x 171, 181 (4th Cir. 2020); Byrd v. Skipper, 940 F.3d 248, 259 (6th Cir. 2019). And it is telling here, for purposes of establishing a reasonable probability, that Knight’s similarly situated wired co-defendant who was advised by his counsel of the generosity of the plea offer and his potential sentencing exposure if he rejected it, wanted to accept the plea offer. This, too, is contemporaneous evidence that Knight would also have been inclined to accept the offer had he not been misinformed and inadequately informed about the plea offer and the enhanced sentencing exposure he would face by going to trial and instead received the assistance of counsel to which the Constitution entitled him. While our dissenting colleague cites Lee for the proposition that defendants often weigh differently the respective risks of pleading and going to trial, Lee does not question that one co- defendant’s willingness to accept a plea offer may tend to show 12 that another co-defendant would have taken the same plea offer. The record shows that Knight and Thorpe’s criminal history categories were only one level apart. See Govt’s Supp. Sent. Mem. 2–3 (Nov. 8, 2013). That Thorpe may have received a slightly higher sentence than Knight by going to trial does not make Thorpe’s willingness to accept the plea offer irrelevant to whether there is a reasonable probability that Knight would have accepted the offer as well.

Other evidence contemporaneous to when the plea offer was pending in the Superior Court indicates that Knight may have been amenable to accepting the plea offer had he been properly advised by counsel. Knight explained on remand at the ineffective assistance hearing that when his counsel informed him the government had extended a plea offer, his first question was “how much time do[es] [the government] want for that?” Hearing Tr. 19 (May 24, 2017). That question suggests that his decision whether to accept the plea offer was calibrated to the sentence that he would receive as a result of pleading guilty. Knight’s circumstances do not otherwise indicate that he was dead-set on going to trial no matter its risks and consequences, and he may well have responded to the plea offer differently had counsel correctly advised him of its sentencing consequences. Instead, he made the decision to reject a two-to-six-year sentence plea offer based on the understanding that his sentence would be ten years.

The remainder of Knight’s contemporaneous exchange with counsel further indicates a reasonable probability that Knight would have accepted the plea offer had counsel apprised him of the consequences of declining it. When counsel told him that the guilty plea would require ten years’ imprisonment and Knight responded that he was “not copping to that,” counsel interjected: “Well, just hold up, I’m going to come over to the jail and talk to you.” Id. That statement 13 indicated there was more Knight should consider before deciding whether to accept the plea offer. Knight’s reply, “okay,” id., indicates that although his initial reaction upon learning he would have to serve ten years in prison was to reject the plea offer, he was amenable to further discussion and possibly to changing his mind even under the mistaken impression that he would face ten years if he pled guilty. Given this exchange, had Knight’s counsel visited him in jail, corrected his earlier erroneous advice, and adequately counselled him on the sentencing exposure he faced if he rejected the plea offer, there is a reasonable probability that Knight would have changed his mind, especially given the magnitude of the disparity in that exposure as compared to accepting the government’s initial offer. After all, he was expecting to have a child shortly enter his life and presumably would not have preferred to be in prison during the entirety of his child’s youth.

Despite this record evidence, the government insists that the only evidence Knight would have accepted the plea offer was his after-the-fact testimony at the ineffective assistance hearing. In the government’s view, what contemporaneous evidence did exist of Knight’s preferences at the time of the plea offer suggests that he would not have accepted it because Knight told counsel that he wanted to be released to attend his child’s birth. Under the terms of the plea offer, he would not have been able to be present. Also, the government notes, Knight was hopeful that the victim of the crimes would not testify against him at a trial.

Admittedly, these are considerations that would have weighed in favor of Knight rejecting the plea offer and proceeding to trial. Because one of the terms of that offer was that he would be unable to seek to modify the conditions of his presentence detention, accepting the offer would have meant 14 he would not have been able to obtain immediate release for the birth of his child. And his reported optimism that a key witness would not testify against him at trial might have tempted him to take his chances in the hope of obtaining an acquittal. Yet none of this evidence precludes there being a reasonable probability that Knight, upon being properly advised by counsel, including a realistic assessment of whether a key government witness would not testify at trial, would have accepted the generous plea offer. Knight was under the erroneous impression that accepting the plea offer came with ten years’ imprisonment, and counsel failed to advise him of the worst-case scenario consequences of declining the plea offer. Given that the Supreme Court has acknowledged that “the possibility of even a highly improbable result may be pertinent to the extent it would have affected [a defendant’s] decisionmaking,” Lee, 137 S. Ct. at 1967, Knight’s considerable misunderstanding of his circumstances, caused by counsel’s omissions and misinformation, is relevant to whether Knight would have accepted the plea offer but for counsel’s ineffectiveness.

Furthermore, to the extent some evidence suggests that Knight was not inclined to accept the Superior Court plea offer, it is of limited value because it is infected by counsel’s deficient performance. What the evidence shows is that under what he mistakenly understood to be the circumstances, Knight, unlike his co-defendant Thorpe, did not want to accept a generous plea offer. Yet the record also shows that Knight’s understanding, unlike his co-defendant’s, diverged significantly from his actual situation. The priority that Knight placed on being present at his child’s birth arose in the context of thinking the plea offer required ten years’ imprisonment. His assessment of what was in his best interests could well have changed had he been correctly advised of the consequences of accepting and of rejecting the plea offer. Given the severity of the charges that 15 Knight faced, along with the fact that his alleged commission of the offense while using a firearm and on supervised release for a prior federal drug conviction weighed in favor of detention, the prospect that he would be able to obtain pretrial release was likely illusory, as adequate counsel could have discussed with him. See, e.g., United States v. Smith, 79 F.3d 1208, 1210–11 (D.C. Cir. 1996); United States v. Peralta, 849 F.2d 625, 626 (D.C. Cir. 1988). So too, given the government’s interest in having Peters’ testimony as the victim of the crimes, was Knight’s speculation that Peters would not testify against him at trial. Had Knight received proper advice from his counsel at the time of the plea offer, he would have learned these were unlikely prospects.

In short, that Knight was focused on obtaining release says little about what he would have done had he been adequately advised of the consequences of declining the plea offer. In analyzing whether a defendant had suffered prejudice from his attorney’s failure to inform him of his sentencing exposure if he declined the government’s plea offer, this court has aptly observed: “[T]he choices that [the defendant] actually made do not necessarily shed any useful light on the choices that he would have made if he had been properly advised.” United States v. Thompson, 27 F.3d 671, 677 (D.C. Cir. 1994). The same is true here.

Nor does the absence of unequivocal contemporaneous evidence that Knight affirmatively wanted a plea deal mean that he cannot show a reasonable probability that he would have accepted the plea offer if he had been provided the effective assistance of counsel. As this court explained, “[t]he Supreme Court did not suggest in Lee that a defendant must hypothesize his counsel’s advice might be erroneous and state contemporaneously that his plea decision would differ if that were so.” Aguiar, 894 F.3d at 362 (discussing Lee, 137 S. Ct. 16 at 1967–68). Here, as in Aguiar, “[t]he gravamen of [Knight’s] claim is that because of [his] counsel’s deficiency, he had no reason to suspect he needed to make such a statement, and thus did not know the full consequences of his decision to reject the plea.” Id. At the time of the Superior Court plea offer, with the misinformation and insufficient information he had received, Knight could not intelligently assess whether to accept the offer. To meet his burden, Knight was not required to show either that he wanted to accept the plea offer but was dissuaded by counsel, or that he certainly would have accepted the offer but for counsel’s ineffectiveness. Instead, he needed to show only that there was a reasonable probability that he would have accepted the plea offer were it not for his counsel’s inadequate assistance.

The government also maintains that Knight’s subsequent rejection of a plea offer in his federal case shows that he was not amenable to any kind of plea deal and thus would not have accepted the Superior Court plea offer even if properly advised by counsel. It is debatable whether this evidence is contemporaneous because the government made the federal court plea offer six months after Knight and Thorpe rejected the Superior Court plea offer. At most it sheds only limited light on whether Knight would have accepted the Superior Court plea offer had his counsel provided proper assistance because the plea offer on the federal charges was considerably less attractive. In federal court, Knight would have been required to plead guilty to three counts carrying a mandatory minimum sentence of five years and a cumulative maximum sentence of 45 years. The Superior Court plea offer required Knight to plead guilty to a single count that carried no mandatory minimum and a ten-year maximum sentence, with a Sentencing Guidelines range of two to six years. The government, therefore, puts too much weight on his rejection of the federal plea in arguing that because Knight declined this 17 plea offer, he would not have accepted any plea offer. Moreover, Thorpe’s willingness to accept the plea offer in Superior Court counsels against reading too much into Knight’s rejection of the subsequent plea offer in his federal case, for Thorpe also rejected the federal court offer. That Thorpe did so, and that he would have accepted the Superior Court offer, suggests that Knight might have had good reason for rejecting the federal court offer regardless of whether he would have accepted the Superior Court offer had he been properly advised by his counsel.

In sum, it is undisputed by the district court and the government that Knight’s counsel’s performance was deficient. Those deficiencies distorted Knight’s understanding of his circumstances, rendering it impossible for him to make an intelligent decision about whether to accept a generous plea offer. Consequently, a proper evaluation of the evidence of Knight’s interests and desires to go to trial and to be present at the birth of his child required the district court to consider the effect of his counsel’s failings. The evidence before the district court sufficed to establish a reasonably probability that Knight, like Thorpe, would have accepted the plea offer. Because the district court’s subsidiary findings regarding Knight’s desire to be at his child’s birth and to go to trial failed to account for the direct impact of his counsel’s deficient performance, those findings, to the extent they were the basis for the district court’s determination that Knight failed to show prejudice, are clearly erroneous. And because the government has never suggested that it would have rescinded the offer, or that the Superior Court would not have accepted the offer, those arguments are forfeited. See Carducci v. Regan, 714 F.3d 171, 177 (D.C. Cir. 1983). Upon review of the prejudice determination, we therefore reverse the district court’s denial of Knight’s Strickland claim. 18 B.

In contrast, we agree with the district court that Thorpe did not receive ineffective assistance of counsel. Unlike Knight’s counsel, Thorpe’s counsel met with his client at least twice in jail prior to the February 19 preliminary hearing to discuss the plea offer. During these conversations, Thorpe’s counsel provided Thorpe with all of the information necessary to make an intelligent decision whether or not to accept the plea offer, including the sentencing range for the ADW charge, the wired nature of the plea offer, and the possibility of federal charges with substantially greater sentencing exposure if he rejected the plea offer. Further, Thorpe’s counsel informed Thorpe that he had learned from Knight’s counsel that Knight was unlikely to accept the plea offer. And, consistent with the practice of the D.C. Public Defenders Service where we worked, Thorpe’s counsel asked the government to unwire the plea offer. Thus, because this conduct “falls within the wide range of reasonable professional assistance,” Strickland, 466 U.S. at 689, Thorpe did not receive constitutionally deficient performance from his counsel. It follows that Thorpe cannot establish a violation of his Sixth Amendment rights.

III.

“Sixth Amendment remedies should be ‘tailored to the injury suffered from the constitutional violation and should not unnecessarily infringe on competing interests,’” Lafler, 566 U.S. at 170 (quoting United States v. Morrison, 449 U.S. 361, 364 (1981)), and there is considerable discretion to fashion such a remedy, see id. at 171. Although the remedy for a Sixth Amendment violation should not “grant a windfall to the defendant or needlessly squander the considerable resources the State properly invested in the criminal prosecution,” it “must ‘neutralize the taint’ of [the] constitutional violation.” 19 Id. at 170 (quoting Morrison, 449 U.S. at 365). When a Sixth Amendment deprivation causes a defendant to reject “an offer . . . for a guilty plea to a count or counts less serious than the ones for which [he] was convicted” at trial, “resentencing alone [based on the convictions at trial] will not be full redress for the constitutional injury.” Id. at 171. “In these circumstances, the proper exercise of discretion to remedy the constitutional injury may be to require the prosecution to reoffer the plea proposal.” Id.

This is such a case, for the appropriate remedy calls upon the government to reoffer the original plea deal to Knight. While the district court has some discretion to accept or reject the plea, see Lafler, 566 U.S. at 172, the Supreme Court in Lafler declined to define the boundaries of that discretion, id., and so do we here. But we do note that there is nothing in Lafler to indicate that the breadth of the district court’s discretion is as great as our dissenting colleague suggests. See Dis. Op. at 14–15. Rather, the Court in Lafler indicated that the trial court’s discretion derives from the court rule governing its acceptance or rejection of plea agreements. 566 U.S. at 174 (citing Mich. Ct. Rule 6.302(C)(3) (2011)). Here, that rule is Rule 11 of the Federal Rules of Criminal Procedure, under which “a district court lacks authority to reject a proposed agreement based on mere disagreement with a prosecutor’s underlying charging decisions.” United States v. Fokker Servs. B.V., 818 F.3d 733, 745 (D.C. Cir. 2016) (citing United States v. Ammidown, 497 F.2d 615, 622 (D.C. Cir. 1973)). And Lafler certainly did not indicate that the impossibility of restoring to the government the costs of trying a defendant is grounds alone to deny that defendant any remedy for the violation of his Sixth Amendment rights. Indeed, such a rule would threaten to render the remedy articulated in Lafler a nullity because in virtually every Lafler-type case the government expends significant resources at trial as a result of the defendant’s 20 counsel’s incompetent performance. To the contrary, the Court in Lafler, 556 U.S. at 172, noted that in fashioning its remedy the trial court should “find[] a remedy that does not require the prosecution to incur the expense of conducting a new trial,” even though “[t]he time continuum makes it difficult to restore the [parties] to the precise positions they [previously] occupied.” In doing so, the trial court “must weigh various factors,” id. at 171, and may “consult” the respective positions occupied by the defendant and prosecution before the rejection of the plea as a “baseline,” id. at 172. Nor would enforcement of the plea agreement on remand “dramatically” benefit Knight to the detriment of the government, Dis. Op. at 15, for although the government incurred the expense of proceeding to trial as a result of Knight’s counsel’s ineffective assistance, that must be balanced against the fact that Knight’s term of imprisonment has already exceeded the upper bound of the Superior Court Sentencing Guidelines range for the ADW charge set forth in the plea offer.

According to Thorpe, even if his counsel was not deficient, the government must nonetheless reoffer the plea to both defendants, essentially because the generous plea offer in the Superior Court was wired. As Thorpe sees it, despite receiving constitutionally adequate counsel, he has suffered a Sixth Amendment injury “identical” to Knight because the ineffective assistance of Knight’s counsel prevented him from obtaining the benefits of the plea offer. Reply Br. 18. But although Thorpe expressed his desire to accept the plea offer from the outset, he knew that the plea offer was conditioned on both defendants accepting it. Thorpe’s ability to accept the wired plea offer was thwarted by Knight’s uninformed decision to reject it. He was also thwarted by the government’s refusal to unwire the defendants so he could accept the plea offer. Both defendants were convicted by a jury in federal court, and their convictions were affirmed on direct appeal, save for the remand 21 on their ineffective assistance of counsel claims. In these circumstances, where Thorpe’s Sixth Amendment rights were not violated, the court is unaware of any precedent granting relief to one defendant because a co-defendant received the ineffective assistance of counsel. Nor does it seem appropriate to order the government to reoffer a wired plea in order to restore Knight to his original position because were this a different case and Knight’s co-defendant had been acquitted at trial, he would certainly refuse to accept the reissued wired plea, and Knight’s constitutional injury would not be remedied at all.

The appropriate remedy for a defendant who received a wired plea offer but was prevented from taking it solely by his counsel’s ineffectiveness is simply to order the government to extend the offer to that defendant again, without regard to whether his co-defendant would be presently willing to accept the offer. Although this court cannot order that it do so, the government has the discretion to ameliorate any injustice that would result from permitting the inadequately counseled defendant to accept the original plea offer but not the co- defendant whose counsel’s performance was adequate. Even now, the prosecution may seek dismissal of some or all of the charges against Thorpe under Rule 48(a) of the Federal Rules of Criminal Procedure. See, e.g., Rinaldi v. United States, 434 U.S. 22 (1977).

IV.

Our dissenting colleague would resolve this appeal by creating a novel legal framework making it more difficult than current law requires for a defendant to prove the denial of the constitutional right to the effective assistance of counsel. This is accomplished mostly by three means: misreading Supreme Court precedent, creating new law out of whole cloth contrary 22 to precedent, and ignoring on-point precedent of this court. These deviations from the applicable law render useless the usual comparative analysis between the opinion of the court and the dissent but do not obviate the need to respond.

First, our dissenting colleague states that Lee “strongly suggests” that the court’s prejudice analysis may not take account of the generosity of the plea deal in its prejudice analysis, Dis. Op. at 4. Nothing in Lee implies that disparity in sentencing exposure can never qualify as contemporaneous evidence. In Lee, 137 S. Ct. at 1963, the defendant pled guilty to a charge that, unbeknownst to him because of his counsel’s ineffective assistance, would result in his mandatory deportation. There was “no question” that deportation was the paramount consideration for Lee in deciding whether to plead guilty. Id. at 1967. He sought to vacate the plea and proceed to trial, even though he had “no viable defense” to the charge, faced near-certain conviction by a jury, and would thereafter face deportation on top of a likely longer prison sentence. See id. at 1966–67. In deciding whether to accept the plea offer had he been properly advised, Lee therefore would have faced a choice between “certainly” being deported if he pled guilty and “[a]lmost certainly” being deported if he went to trial. Id. at 1968. Thus, as to the “determinative issue” in Lee’s decision whether to accept the plea offer, id., there was barely any disparity at all — only the small difference between certainty and almost-certainty. In Lee, therefore, the Court had no occasion to consider whether a disparity in outcomes between accepting a plea offer and proceeding to trial could bear on the Strickland prejudice analysis.

Second, contrary to our dissenting colleague, reliance on testimony adduced at the evidentiary hearing does not run afoul of Lee. See Dis. Op. at 6–7. At most, Lee, 137 S. Ct. at 1967, proscribed courts from relying “solely” on “post hoc assertions 23 from a defendant about how he would have pleaded.” The Supreme Court in Lee did not, however, impose a blanket ban on considering testimony adduced at an evidentiary hearing about temporally contemporaneous events. The Supreme Court has observed that a trial record is “not developed precisely for the object of litigating or preserving [an ineffective assistance] claim and thus [is] often incomplete or inadequate for this purpose.” Massaro v. United States, 538 U.S. 500, 504–05 (2003). As a result, in deciding an ineffective assistance claim, the court “may take testimony from witnesses for the defendant and the prosecution and from the counsel alleged to have rendered the deficient performance.” Id. at 505. Nor, as the dissent suggests, is such properly considered testimony limited to the defendant’s production of a contemporaneous “transcript, letter, or recording,” should he be lucky enough to have one. Dis. Op. at 6. Such a test is nowhere required or suggested or even hinted at in Lee, much less in supporting authority. Id. Our dissenting colleague protests that he has been misunderstood, noting he would not exclude plea generosity evidence or evidence adduced at an evidentiary hearing. Dis. Op. at 7 n.1. But this is to no avail for he still views such not to be “contemporaneous evidence” as he defines it and so insufficient to show prejudice.

Third, our dissenting colleague acknowledges that this court reviews the district court’s prejudice decision de novo but finds no clear error in the district court’s factual finding that Knight offered no contemporaneous evidence, a subject that this court also reviews de novo. Abney, 812 F.3d at 1087; see also id. at 1093–94; United States v. Toms, 396 F.3d 427, 432 (D.C. Cir. 2005). Our colleague either ignores the evidence before the district court (or belittles it as “snippets,” Dis. Op. at 12), or redefines “contemporaneous evidence” as limited to physical evidence or requires statements by a defendant at the time that this court has held are not required. Dis. Op. at 6–7. 24 The flawed logic on which the district court proceeded, see Thompson, 27 F.3d at 677, is highlighted when the dissent too points to Knight’s refusal to enter a plea to the federal charges, Dis. Op. at 10. In applying common sense in the government’s favor, see id. at 13, while refusing to consider Knight’s actions contextually, see id. at 11–12 — that is, in the context of his ignorance of his circumstances as a result of his counsel’s deficient advice — our colleague relies on generalizations that can only be considered anecdotal absent record support.

Accordingly, we affirm the order denying Thorpe’s Sixth Amendment challenge but reverse the denial of Knight’s Sixth Amendment challenge and remand his case to the district court to provide a remedy consistent with this opinion. KATSAS, Circuit Judge, dissenting in part: After a jury convicted him of serious crimes, Melvin Knight claimed that bad legal advice had caused him to reject a favorable plea offer. Following an evidentiary hearing on this claim, the district court found no reasonable probability that Knight would have accepted the plea offer had he received adequate advice. That finding was not clearly erroneous, and it establishes that Knight did not receive ineffective assistance of counsel.

I

On January 28, 2013, Knight and Aaron Thorpe violently kidnapped Edmund Peters and Luttitia Fortune. Knight and Thorpe assaulted their victims outdoors, fired a gunshot, forced their way into Peters’s apartment, tied up the victims, and beat Peters while attempting to steal his money and drugs. They promised to kill Peters, and Thorpe placed the barrel of his gun against Peters’s head. When police surrounded the apartment, Knight and Thorpe untied the victims, concocted a story of friendly sparring, and told the victims to play along. Peters complied out of fear that Knight would further harm him. Knight told Peters that he would not go to prison over the kidnapping, which Peters took as another threat. The police were not fooled.

Knight and Thorpe initially were charged with armed kidnapping in D.C. Superior Court. On January 31, 2013, the government offered Knight and Thorpe a wired plea deal—one that required acceptance by both defendants. If each defendant would plead guilty to one count of assault with a dangerous weapon, the government would forgo various other, more serious charges. While the offer was outstanding, Knight’s counsel advised Knight that if he accepted the plea offer, he would likely face ten years of imprisonment. In fact, ten years was the statutory maximum for assault with a dangerous weapon, while the recommended sentencing guideline range would have been two to six years. Counsel also failed to advise 2

that Knight would face substantially greater exposure if he rejected the plea offer and the government chose to pursue the further charges. Knight rejected the offer, which prevented Thorpe from accepting it.

The government dismissed the case in Superior Court and obtained a federal indictment. Knight received a different counsel for district court. Knight and Thorpe each was charged with one count of possessing a firearm as a felon, two counts of possessing a firearm during a crime of violence, two counts of armed kidnapping, one count of armed burglary, one count of assault with a dangerous weapon, one count of obstruction, and one count of conspiracy. The jury convicted on all counts, and Knight and Thorpe received prison sentences of 268 and 300 months, respectively. On direct review, we rejected various challenges to the convictions and to Thorpe’s sentence, but we remanded the case for factual development of claims that each defendant’s counsel had provided ineffective assistance during the plea negotiations in Superior Court. United States v. Knight, 824 F.3d 1105 (D.C. Cir. 2016).

On remand, the district court held three days of evidentiary hearings on the ineffective-assistance claims. Knight and Thorpe testified at length, as did their respective Superior Court counsel and a custodian of D.C. jail records. The court ordered production of the initial plea offer, transcripts memorializing the plea discussions in Superior Court and district court, and other documents bearing on the contested representations. Knight also introduced an ethics complaint that he had filed against his Superior Court counsel.

After reviewing all this evidence, the district court rejected the claims of both defendants. United States v. Thorpe, No. 13- cr-131, 2019 WL 1117197 (D.D.C. Mar. 11, 2019). The court found that Knight’s counsel performed deficiently in 3

connection with the plea offer, but it found no reasonable probability that Knight would have accepted the offer had he received adequate advice. Id. at *8–10. The court also found that Thorpe’s counsel did not perform deficiently. Id. at *11.

II

The Sixth Amendment confers upon criminal defendants a right “to have the Assistance of Counsel.” The Supreme Court has held that “Assistance” means effective assistance. Strickland v. Washington, 466 U.S. 668 (1984). To establish a violation of this right, the defendant must prove both that his counsel performed deficiently and that this caused prejudice. Id. at 687. To prove prejudice, the defendant must show a reasonable probability that the deficient performance changed the result of the proceeding. See id. at 694. Thus, for claims that deficient advice caused the defendant to reject a plea offer, the defendant must show a reasonable probability that he would have accepted the offer had he received adequate advice. Lafler v. Cooper, 566 U.S. 156, 164 (2012).

This case turns on whether Knight established a reasonable probability that he would have accepted the Superior Court plea offer if he had received proper advice. For two reasons, I would affirm the district court’s conclusion that Knight failed to meet his burden of proof on this question.

A

In Lee v. United States, 137 S. Ct. 1958 (2017), the Supreme Court imposed a high evidentiary hurdle for defendants seeking to undo their plea decisions through claims of ineffective assistance. The Court held that the defendant must adduce evidence contemporaneous with the plea decision: “Courts should not upset a plea solely because of post hoc 4

assertions from a defendant about how he would have pleaded but for his attorney’s deficiencies. Judges should instead look to contemporaneous evidence to substantiate a defendant’s expressed preferences.” Id. at 1967. Although Lee involved a defendant seeking to undo a prior plea acceptance, the Court’s reasoning fully applies to defendants seeking to undo a prior plea rejection. We have recognized that Lee applies in both contexts. See, e.g., United States v. Aguiar, 894 F.3d 351, 361– 62 (D.C. Cir. 2018).

No contemporaneous evidence suggests that Knight would have accepted the plea offer had he received proper advice. In this Court, Knight’s current counsel acknowledged that the contemporaneous evidence was “nearly useless” to show prejudice. Oral Arg. at 4:45. Yet my colleagues invoke two categories of evidence that they say are both contemporaneous and weighty enough to show prejudice.

First, my colleagues reason that the “generosity” of the Superior Court plea offer was itself contemporaneous evidence of prejudice. Ante, at 9. But Lee strongly suggests otherwise. There, the Supreme Court did not make its own abstract assessment of how favorable the disputed plea was to the defendant. Nor did the Court rest on the defendant’s undisputed testimony, at a post-conviction hearing, that he would have rejected the plea offer had he known that it would lead to mandatory deportation. See 137 S. Ct. at 1967–68. Instead, the Court insisted on corroborating evidence “contemporaneous” with the plea itself—there, the defendant’s specific statements during his plea colloquy that any risk of deportation would have affected his plea decision. See id. at 1968. And the Court discounted objective evidence that the accepted plea was favorable to the defendant given the very high likelihood of a conviction. On that point, the Court stressed that defendants assess trial risks differently, and even 5

a defendant “almost certain” to be convicted could rationally reject a plea. Id. at 1968 (cleaned up).

Moreover, there are good reasons for insisting on contemporaneous evidence beyond an assessment of how generous a plea offer appears after-the-fact. Plea deals secure important benefits for the government. They eliminate the time and expense of developing and trying cases, which would otherwise overwhelm a judicial system in which almost 98 percent of convictions are secured through guilty pleas. See Statistical Tables for the Federal Judiciary, tbl. D-4 (2019). They also eliminate the otherwise inescapable risk of outright acquittals. It is hardly surprising that the government offers significant benefits in return for guilty pleas—and, therefore, that defendants who “take their case[s] to trial and lose receive longer sentences” than those who plead guilty. Missouri v. Frye, 566 U.S. 134, 144 (2012) (quotation marks omitted). Perhaps this plea offer was unusually generous, but Knight has not made that case. And if any large disparity in exposure qualifies as contemporaneous evidence of prejudice, then we have opened the floodgates, all but eliminating prejudice as an independent element for Lafler claims. That is precisely the opposite of what the Supreme Court sought to accomplish in Lee, which stressed that “[s]urmounting Strickland’s high bar is never an easy task,” 137 S. Ct. at 1967 (quotation marks omitted), and which imposed a contemporaneous-evidence requirement to keep it that way.

My colleagues cite out-of-circuit cases for the proposition that a “disparity in sentencing exposure” may show prejudice. Ante, at 11. But three of those cases were decided before Lee established the requirement of contemporaneous corroborating evidence. United States v. Herrera, 412 F.3d 577, 582 (5th Cir. 2005); Griffin v. United States, 330 F.3d 733, 739 (6th Cir. 2003); United States v. Day, 969 F.2d 39, 45–47 (3d Cir. 1992). 6

A fourth rested on the defendant’s “long history of entering into plea agreements in prior cases.” Dodson v. Ballard, 800 F. App’x 171, 180–81 (4th Cir. 2020). In a fifth, the defendant “specifically asked” his counsel about pleading guilty, yet counsel promised that going to trial would be a “home run.” Byrd v. Skipper, 940 F.3d 248, 258–59 (6th Cir. 2019). Nothing like that happened here.

My colleagues cite one consideration specific to the plea offer in this case—that Thorpe wanted to accept it. Ante, at 11– 12. That does not count for much, as defendants often weigh differently the respective risks of pleading and going to trial. See Lee, 137 S. Ct. at 1968–69. Moreover, Thorpe had “a more significant criminal history than Knight,” which substantially increased his downside risk at trial. Knight, 824 F.3d at 1111. Thorpe’s preferences thus do not shed much light on Knight’s.

Second, my colleagues conclude that Knight’s testimony at the 2017 ineffective-assistance hearing qualifies as contemporaneous evidence. Ante, at 12–13. It does not. At that hearing, conducted years after Knight had been convicted and sentenced, Knight testified about conversations with his lawyer during the 2013 plea negotiations. This may be evidence about events contemporaneous with the plea offer. But it is not “contemporaneous evidence” as opposed to “post hoc assertions from a defendant about how he would have pleaded but for his attorney’s deficiencies.” Lee, 137 S. Ct. at 1967. “Contemporaneous evidence” would be something akin to the statements made by Lee “at his plea colloquy,” which sufficed to corroborate his later post-conviction testimony. See id. at 1968. Here, Knight presented no evidence generated contemporaneously with the plea negotiations—such as a transcript, letter, or recording—to support his later contentions about his preferences at the time. His 2017 testimony, about the 2013 plea offer, was not “contemporaneous evidence.” 7

To downplay the need for contemporaneous evidence, my colleagues invoke Aguiar. Ante, at 15–16. There, we held that a defendant does not need contemporaneous evidence to secure “an evidentiary hearing to prove his claim.” 894 F.3d at 361– 62. But we expressly declined to address how the defendant could “satisf[y] his ultimate burden of proof.” Id. We also observed that Lee does not require a defendant to “hypothesize his counsel’s advice might be erroneous and state contemporaneously that his plea decision would differ if that were so.” Id. at 362. Perhaps not, but Lee does require the defendant to build a prejudice case consisting of more than just post-conviction testimony—given after the defendant has taken his shot at acquittal—plus the near truism that plea deals produce much lower sentences than do convictions after trial.

Because Knight presented no contemporaneous evidence that he would have accepted the Superior Court plea offer but for bad legal advice, we should reject his post hoc attempt to undo his plea decision.1

1 My colleagues characterize this dissent as saying that courts may not consider “the generosity of the plea deal” or “testimony adduced at an evidentiary hearing” post-conviction. Ante, at 22–23. To the contrary, I have simply explained that post-conviction testimony is not contemporaneous evidence. And because it cannot suffice to show prejudice under Lee, then neither can the formula embraced by my colleagues: post-conviction testimony plus the truism that the defendant would have received a much shorter sentence had he accepted the plea offer. 8

B

Even overlooking the lack of contemporaneous evidence, the district court permissibly concluded that Knight had failed to show prejudice.

We review the ultimate question of prejudice de novo, United States v. Abney, 812 F.3d 1079, 1086–87 (D.C. Cir. 2016), but “the district court’s factual findings made in the course of judging an ineffective assistance of counsel claim may be set aside only if clearly erroneous,” United States v. Mathis, 503 F.3d 150, 151 (D.C. Cir. 2007). Strickland itself makes clear that district-court “findings of fact made in the course of deciding an ineffectiveness claim” are “subject to the clearly erroneous standard” of review. 466 U.S. at 698.

According to the district court, Knight failed to show a reasonable probability that he would have accepted the plea offer had he received proper advice. 2019 WL 1117197, at *10. This was a finding of fact, not a legal statement about what constitutes Strickland prejudice. In United States v. Thompson, 27 F.3d 671 (D.C. Cir. 1994), we held that a district court’s determination “whether there was a ‘reasonable probability’ that [the defendant], if properly advised, would have pleaded guilty” in time to qualify for a sentencing reduction was a “factual finding” to be reviewed for clear error. Id. at 677 (quoting Strickland, 466 U.S. at 694). Likewise, we have treated as factual, and subjected to clear-error review, determinations whether the government would have offered a better plea deal but for defense counsel’s deficient performance, see Mathis, 503 F.3d at 152, and whether a defendant accepting a plea deal would have been convicted had he gone to trial, see United States v. Del Rosario, 902 F.2d 55, 9

58 (D.C. Cir. 1990), abrogated on other grounds by Padilla v. Kentucky, 559 U.S. 356 (2010). Other circuits have held that the question whether a defendant would have accepted a plea offer if properly advised is a factual one. See, e.g., Johnson v. Genovese, 924 F.3d 929, 938–39 (6th Cir. 2019); United States v. Scribner, 832 F.3d 252, 258 & n.4 (5th Cir. 2016); Merzbacher v. Shearin, 706 F.3d 356, 366–68 (4th Cir. 2013). The latter cases arose on collateral review, but the distinction between direct and collateral review has no bearing on whether the question at issue is legal or factual.

Common sense reinforces this view. The question whether Knight would have accepted the plea offer had he been properly advised has no impact on other cases. It involves no normative judgments. And it rests on “the credibility of witnesses and therefore turns largely on an evaluation of demeanor.” Miller v. Fenton, 474 U.S. 104, 114 (1985). The only possible justification for reviewing this finding de novo is that the finding effectively controls the question of Strickland prejudice. But it is “well established” that “an issue does not lose its factual character merely because its resolution is dispositive of the ultimate constitutional question.” Id. at 113. For these reasons, I would review the district court’s determination only for clear error.

“A finding is ‘clearly erroneous’ when although there is evidence to support it, the reviewing court on the entire evidence is left with the definite and firm conviction that a mistake has been committed.” United States v. U.S. Gypsum Co., 333 U.S. 364, 395 (1948). We apply this deferential standard because district courts are “best suited to developing the facts” bearing on ineffective-assistance claims. Massaro v. United States, 538 U.S. 500, 505 (2003). We owe even 10

“greater deference” when factual findings rest on credibility determinations, “for only the trial judge can be aware of the variations in demeanor and tone of voice that bear so heavily on the listener’s understanding of and belief in what is said.” Anderson v. Bessemer City, 470 U.S. 564, 575 (1985).

The critical finding here—that Knight was unlikely to accept the plea offer even if he had received adequate advice— is not clearly erroneous.

First, at a June 4, 2013 status hearing in district court, the prosecutor memorialized the defendants’ firm desire to go to trial: “I have talked to defense counsel in this case. It appears that the Defendants are not amenable to even discussing a non trial disposition. I would like that to be reflected on the record . . . because the Defendants have indicated they want to go to trial.” J.A. 465 (emphasis added). Knight’s district-court counsel—who is not alleged to have been ineffective—was present at the hearing and made no objection to this representation. The fact that Knight was “not amenable to even discussing” a plea, even after having received effective assistance of counsel and having been indicted for all of his crimes, strongly suggests that he would not have accepted a Superior Court plea deal had he then been advised that significant further charges were possible.

My colleagues focus on a draft plea agreement apparently offered by the government on June 6, 2013. Ante, at 16–17. But the defendants’ rejection of that offer only tends to confirm the prosecutor’s statement that they were “not amenable to even discussing” a plea. And despite my colleagues’ suggestion to the contrary, that sweeping statement remains significant even though the June plea offer was less favorable to the defendants than the January one had been. 11

Second, Knight’s behavior in February 2013, while the Superior Court plea offer was pending, tracks what the prosecutor later said in June. The district court found that Knight “did not want any additional time to consider or discuss the plea [in Superior Court], but instead pushed [his attorney] to ‘move forward.’” 2019 WL 1117197, at *10. To be sure, Knight probably assumed that the offer on the table would entail ten years in prison. But plea bargains involve a “give- and-take negotiation.” Bordenkircher v. Hayes, 434 U.S. 357, 362 (1978) (quotation marks omitted). And if Knight had been open to a plea deal along the lines of what the government had actually proposed, with a recommended guideline sentence of up to six years, one might have expected him at least to consider the possibility of further negotiations—especially given his experience with two prior guilty pleas. Instead, Knight pressed his attorney to “move forward” as quickly as possible, to take his chances at trial.

Third, Knight had a powerful incentive to avoid any conviction. When he kidnapped Peters in 2013, Knight was still serving a five-year term of supervised release following his 2001 guilty plea and ten-year sentence for distributing more than 50 grams of cocaine base. 2019 WL 1117197, at *4 n.5; see United States v. Knight, No. 01-cr-00016 (D.D.C. Sept. 14, 2001). Knight admitted knowing that his supervised release would be revoked if he were convicted. And his attorney testified that Knight “was concerned about getting that additional time if he took the plea.” J.A. 340. As it turns out, Knight was sentenced to 21 months of imprisonment for violating the terms of his supervised release, running concurrently with the sentence imposed in this case, after the judge in the drug case accepted Knight’s request for leniency based on the length of the sentence imposed here. See United States v. Knight, No. 01-cr-00016 (D.D.C. July 1, 2014). 12

Fourth, Knight hoped that Peters—a key prosecution witness—would not testify against him. 2019 WL 1117197, at *11. Knight had good reason to be hopeful, for his threats already had induced Peters to lie to the police on the night of the arrest. And Peters agreed to testify against Knight only in exchange for his own favorable plea deal in a separate case— which was reached long after Knight had declined his Superior Court plea offer. In sum, Knight’s own threats against Peters supported his hope that Peters would not testify against him.

Fifth, on May 15, 2013, Knight filed an ethics complaint against his Superior Court counsel with the D.C. bar. By then, Knight already had been indicted in federal court and appointed new counsel, whom Knight does not contend was ineffective. In the bar complaint, Knight raised a host of allegations against his former counsel—most prominently that counsel, in obtaining a three-week continuance for Knight to consider the plea offer, did not push the case forward quickly enough. Nowhere in that complaint did Knight raise the alternative, contradictory allegation that his counsel should have spent more time attempting to negotiate a better plea deal or advising Knight of the risks of an expanded indictment.

My colleagues point to snippets of contrary evidence from Knight’s testimony at the ineffective-assistance hearing. For instance, Knight claims to have asked his attorney “how much time do they want” for the plea, and to have responded “okay” when his counsel asked to discuss the plea issue further. J.A. 61; see ante, at 12–13. But the district court had ample reasons for taking Knight’s testimony “with a grain of salt.” 2019 WL 1117197, at *9. Among other things, Knight was a repeatedly convicted felon, and his offenses in this case included a scheme to escape responsibility by coercing the victims of his crimes “to lie to the police afterwards.” Id. at *10 n.6. Moreover, Knight and his former counsel gave conflicting testimony on 13

whether Knight hoped that Peters would not testify against him, and the district court resolved that dispute by concluding that Knight had testified untruthfully. See id.

Finally, I am skeptical of my colleagues’ view that the plea offer here was unusually generous. In the abstract, there is little surprise that the initial offer—made three days after the crimes, before the government had conducted much of an investigation, before the primary victim had agreed to testify, and months before the trial—was substantially more favorable than the sentences imposed after a full trial and guilty verdict. As noted above, that is how pleas normally work. And this plea offer may have fairly reflected evidentiary uncertainty and Peters’s unwillingness to cooperate at the time, rather than an act of gratuitous generosity. But in any event, Lafler asks only whether the defendant would have accepted the plea offer, not whether an objectively reasonable person would have done so. See 566 U.S. at 164. Thus, our own assessment of the offer must yield to Knight’s subjective reasons for rejecting it.2

2 If the plea offer were unusually generous, that would highlight a further problem with my colleagues’ disposition of this appeal. To show prejudice under Lafler, the defendant must establish a reasonable probability that but for inadequate legal advice (1) the defendant would have accepted the plea offer, (2) the government would not have withdrawn it, (3) the court would have accepted the plea, and (4) the sentence under the plea would have been less severe than the sentence actually imposed. See 566 U.S. at 164. In this case, the district court found that Knight had failed to prove the first element of prejudice, so it did not address the others. An unusually generous plea offer would simply highlight the need to determine whether the Superior Court would have accepted it. On my colleagues’ own reasoning, then, we should remand for the district court to resolve that question. 14

The district court committed no clear error in finding that Knight was unlikely to have accepted the plea offer had he received adequate legal advice. Knight’s ineffective-assistance claim thus fails for lack of any prejudice.

III

On the question of remedy, my colleagues order the government to re-extend its original plea offer to Knight. They acknowledge that the district court retains discretion to consider whether to accept or reject this plea deal, while also noting that the discretion has limits. Ante, at 19–20.

Lafler governs this remedial inquiry. The Supreme Court noted that, if ineffective assistance causes the defendant to reject a plea offer, “the proper exercise of discretion to remedy the constitutional injury may be to require the prosecution to reoffer the plea proposal.” 566 U.S. at 171. But “[o]nce this has occurred, the [trial] judge can then exercise discretion in deciding whether to vacate the conviction from trial and accept the plea or leave the conviction undisturbed.” Id. (emphasis added). The Court thus held that the “correct remedy” in Lafler itself was simply “to order the State to reoffer the plea agreement,” and it vacated a Sixth Circuit decision that had further “ordered specific performance of the original plea agreement.” Id. at 174. In so doing, the Supreme Court explained that the trial court on remand could “exercise its discretion in determining whether to vacate the convictions and resentence [the defendant] pursuant to the plea agreement, to vacate only some of the convictions and resentence [the defendant] accordingly, or to leave the convictions and sentence from trial undisturbed.” Id.

In this case, several considerations favor rejecting the reoffered plea agreement. The original plea offer was made 15

only three days after Knight committed his crimes—before the government spent years building a case against him, prosecuting him, and defending against his appeal and post- conviction claims. Seven years after-the-fact, the plea offer would give Knight most of what the government originally offered to him, in the form of substantially lower sentencing exposure. But it would give the government none of what it demanded in return—avoiding the cost of prosecuting this case and the risk of an acquittal. Moreover, after Knight rejected the plea offer, the government offered Peters a favorable plea agreement to secure his testimony against Knight, thus narrowing it options for seeking a lawful punishment of Peters. And another court imposed a lenient sentence on Knight for his supervised-release violations because of his conviction and long sentence in this case. In short, intervening events have made it impossible to restore the parties to the respective positions that they would have held had Knight accepted the plea offer in 2013. And enforcing the plea agreement now would dramatically skew its benefits and burdens in favor of Knight and against the government.

My colleagues fairly note the competing interest in affording some remedy for ineffective assistance in this context, and I have no quarrel with the proposition that this interest must be “balanced against” the government interests noted above. Ante, at 20. In my view, such balancing would occur if the district court on remand were to reject the plea agreement and then impose a sentence taking account of both the ineffective assistance found by my colleagues and the changed circumstances noted above. That approach would recognize the impossibility of restoring the parties to the “precise positions they occupied prior to the rejection of the plea offer.” Lafler, 566 U.S. at 171–72. And it would rest not on judicial disagreement with the government’s initial charging decision, cf. United States v. Fokker Servs. B.V., 818 F.3d 733, 16

745 (D.C. Cir. 2016), but instead on a judgment that intervening events have made it impossible, seven years later, to give both parties the full benefit of their bargain.3

3 I agree with my colleagues that Thorpe received effective assistance of counsel and is entitled to no remedy.