Michael Riolo v. United States

38 F.4th 956

Opinion

[PUBLISH]

In the

United States Court of Appeals For the Eleventh Circuit

____________________

No. 20-12206

____________________

MICHAEL RIOLO, Petitioner-Appellant,

versus UNITED STATES OF AMERICA,

Respondent-Appellee.

____________________

Appeal from the United States District Court for the Southern District of Florida D.C. Docket Nos. 9:11-cv-81028-KAM, 9:09-cr-80058-KAM-1

____________________

2 Opinion of the Court 20-12206

Before JORDAN, JILL PRYOR, and MARCUS, Circuit Judges. JILL PRYOR, Circuit Judge:

Michael Riolo appeals the district court’s denial of his 28 U.S.C. § 2255 motion to vacate his 293-month prison sentence and convictions.1 Riolo argued to the district court that his trial counsel , Theresa Van Vliet, provided ineffective assistance of counsel. Specifically, Riolo asserted that Van Vliet told him if he pled guilty to five counts of mail fraud, he would serve no more than 10 years in prison because she had a deal with the government that his sentencing range would be 97–121 months’ imprisonment under the Sentencing Guidelines. Based on Van Vliet’s representations, Riolo maintained, he pled guilty when he otherwise would have proceeded to trial.

The district court held an evidentiary hearing on Riolo’s § 2255 motion. After reviewing the evidence, the district court found that Van Vliet never represented to Riolo that she had a deal with the government about his guideline range. More than that, the district court found that before the change-of-plea hearing, Van Vliet thoroughly apprised Riolo of the federal sentencing process, explaining that the United States Probation Office would make a recommendation to the district court about his guideline range and

1 Riolo’s 293-month sentence is an amalgam of five sentences, all for mail fraud convictions pursuant to 18 U.S.C. § 1341. For clarity, we refer to the five sentences as a single “sentence” like the parties do.

20-12206 Opinion of the Court 3

that the district court would ultimately determine his guideline range for itself. Based on its factual findings, the district court denied Riolo’s § 2255 motion.

On appeal, Riolo urges us to take another look at the evidence . If we do, he argues, we will find that several of the district court’s factual findings are clearly erroneous. In the light of his version of the facts—that Van Vliet assured him she had a deal with the government about his guideline range—Riolo urges us to conclude that she provided ineffective assistance of counsel. Even leaving aside the disputed facts, he adds, Van Vliet underestimated his guideline range by more than 100 months. That alone, he argues, constitutes ineffective assistance.

After careful review, and with the benefit of oral argument, we find no clear error in the district court’s factual findings. We therefore conclude that Van Vliet did not provide Riolo with ineffective assistance by telling him she had an agreement with the government about his guideline range. We also conclude that Van Vliet did not provide ineffective assistance by underestimating Riolo ’s guideline range. The district court’s judgment is affirmed.

I. BACKGROUND

We begin by summarizing Riolo’s offense conduct, Van Vliet’s representation of him, the change-of-plea hearing, the sentencing hearing, and Riolo’s direct appeal. Next, we review the events that followed Riolo’s § 2255 motion, including two previous 4 Opinion of the Court 20-12206

denials, appeals, and remands of his case, as well as the district court’s final order and judgment on appeal. A. Offense Conduct From August 1999 to December 2008, Riolo convinced more than 80 people to invest money in two corporations for which he was the sole officer and employee: Sterling Wentworth Currency Group, Inc. (“Sterling”) and LaSalle International Clearing Corporation (“LaSalle”). Riolo told investors that he would invest their money in foreign currency trading. Investors sent personal checks or wire transfers to the Sterling and LaSalle bank accounts .

But Riolo did not invest the funds he received. With signatory authority over the Sterling and LaSalle bank accounts, he withdrew investors’ funds and used their money for other purposes. He spent at least some of the money to pay for a home, multiple automobiles , a boat, jet skis, and other luxury items. If an investor wanted to withdraw funds, Riolo would send the money he received from other investors to cover the disbursement. Using this technique, he disbursed over $29.5 million to withdrawing investors , claiming they were receiving their principal investment and returns when most, if not all, of the payments were made with other investors’ funds. To perpetuate this Ponzi scheme, Riolo prepared sophisticated profit and loss statements for the investors— statements which falsely reflected that their money had been invested and was earning substantial returns. Over the nearly 10-year period, Riolo collected more than $44 million from investors.

20-12206 Opinion of the Court 5

B. Van Vliet’s Representation of Riolo In January 2009, the Federal Bureau of Investigation (“FBI”)

served Riolo at his house with a subpoena for corporate records of Sterling and LaSalle. While the FBI was on his property, Riolo contacted his civil lawyer, Bart Houston. Houston, in turn, reached out to his law partner, Theresa Van Vliet, a highly experienced criminal attorney who had been in private practice for 18 or 19 years and was a former Assistant United States Attorney. Houston put Van Vliet on the phone with Riolo, who then put Van Vliet on the phone with the FBI. Van Vliet explained to the FBI agent that her law firm had the records responsive to the subpoena.

Riolo met Van Vliet in person shortly after the FBI’s visit to his house. He told her that he had been running Sterling and LaSalle for approximately 10 years and that he had not made any trades in foreign currency during that time. He explained that he had taken in approximately $44 million from investors. He admitted that he had mailed investors fraudulent statements reflecting that their money had been invested when it had not been. Van Vliet advised that he could face several counts of mail and wire fraud.

Van Vliet continued to communicate regularly with Riolo over the following months. She received a proposed bill of information from the government and engaged in plea negotiations on his behalf. She told him that, given the overwhelming evidence 6 Opinion of the Court 20-12206

against him, she could not mount a successful defense of his case. Riolo told her that the most important consideration for him was the length of any prison sentence he might receive. She explained that his sentence would depend in part on his guideline range under the Sentencing Guidelines. If he pled guilty, she advised, he would likely have the benefit of a three-point reduction under the “acceptance of responsibility” guideline. Civ. Doc. 74-5 at 3 ¶ 5 (internal quotation marks omitted). 2 She told him that a further reduction could be possible if he cooperated with the authorities. If he wanted the lowest sentence possible, she advised, his best course of action was to plead guilty and offer his cooperation to the government. In May 2009, the government formally charged Riolo by information with five counts of mail fraud.

1. Van Vliet’s May 7, 2009 Meeting with Riolo On May 7, 2009, Van Vliet met with Riolo to provide an estimate of his guideline range. Houston and Riolo’s then-wife, Lori Ann Gary, also attended the meeting. Van Vliet told Riolo that based on her calculations, aided by the work of an associate at her law firm, the district court likely would find an offense level of 30, a criminal history category of I, and a resulting guideline range of 97–121 months’ imprisonment. She flagged for him, however, that a higher guideline range was possible if the district court imposed

2 “Civ. Doc.” refers to docket entries in the district court in case number 9:11- cv-81028-KAM. “Crim. Doc.” refers to docket entries in the district court in case number 9:09-cr-80058-KAM-1.

20-12206 Opinion of the Court 7

a four-point enhancement for violating commodities laws while being a commodities trading advisor. But it was Van Vliet’s opinion that the commodities law enhancement would not apply. She reassured Riolo by telling him that the government had independently reached the same guideline-range calculation—that he would have an offense level of 30 and a criminal history category of I, with a resulting guideline range of 97–121 months’ imprisonment.

The parties offered conflicting evidence regarding the May 7, 2009 meeting at the evidentiary hearing in Riolo’s § 2255 proceedings . Riolo, Gary, and Houston all testified that Van Vliet told those in attendance she had reached a binding “deal” or “agreement ” with the government that Riolo’s offense level would be 30, with a resulting guideline range of 97–121 months’ imprisonment. Gary testified that she specifically asked Van Vliet whether he could possibly “get more than 10 years” if he pled guilty. Civ. Doc. 83 at 8. Van Vliet reassured her by saying “no, . . . this was the plea agreement that had been agreed upon with the prosecution and her.” Id. at 7. Houston, a civil lawyer unfamiliar with criminal sentencing , came away with the impression that Van Vliet and the government had something akin to “a stipulation in [civil] practice” about Riolo’s guideline range. Id. at 64. And according to Riolo, Van Vliet suggested that a guideline range of 97 to 121 months’ imprisonment “was locked in.” Id. at 108.

Van Vliet testified adamantly to the contrary. She said that “[n]ever . . . in [her] life” had she communicated to a client that she could reach a deal with the government about a defendant’s 8 Opinion of the Court 20-12206

guideline range. Id. at 237. She broadly denied ever telling Riolo that there was any sort of deal or agreement between her and the government about his sentence. She testified that she told Riolo, on multiple occasions, about “all of the nuances of the sentencing guidelines and his possible exposure.” Civ. Doc. 74-5 at 6 ¶ 18. For one thing, she told him that the probation office would make an independent guideline calculation to aid the sentencing court. For another, she informed him that the sentencing court would not be bound by either of the parties’ estimates regarding his guideline range or ultimate sentence.

2. Van Vliet’s May 11, 2009 Meeting with Riolo Riolo and Van Vliet met again a few days later, on May 11, 2009. This time they met for lunch at a restaurant. Also attending the meeting was Michael McManus, a former DEA agent turned private investigator who worked with Van Vliet. Van Vliet hired McManus to aid in Riolo’s representation. His role was to help Riolo understand “how he could cooperate” with law enforcement to obtain a potential guideline reduction. Civ. Doc. 83 at 188.

At the evidentiary hearing in Riolo’s § 2255 action, the parties offered conflicting testimony about the lunch meeting. According to McManus’s testimony, Van Vliet took the opportunity at the meeting to reexplain to Riolo how federal sentencing worked. McManus testified that Van Vliet told Riolo that the probation office would prepare a presentence investigation report (“PSR”), and the district judge would ultimately determine his guideline range and could “depart upwards” or “downwards.” Id. at 192. McManus 20-12206 Opinion of the Court 9

“kinda got the feeling that this may have been discussed before , . . . because she was saying, now, understand again, or something to that effect . . . you’ve got to have a presentence report.” Id. Van Vliet also remembered telling Riolo at this lunch meeting about “the Sentencing Guidelines . . . , how a plea works, how the Court goes into it, how, after he does a plea there would be a PSR.” Civ. Doc. 84 at 41. In contrast, Riolo denied that Van Vliet was present at his May 11, 2009 lunch with McManus and denied that she went over this information with him.

3. Van Vliet’s Review of the Plea Agreement and Factual Proffer with Riolo

Less than a week after the lunch meeting with McManus, Van Vliet received a copy of the plea agreement from the government and sent it to Riolo by email. In the email Van Vliet said she “talked to [the prosecutor] and his calculations match ours. That is a level 30 with a range of 97 to 121 months. So any break we can possibly get on cooperation will be critical.” Civ. Doc. 76-2 at 2. The attached plea agreement—which Riolo admitted he reviewed —contained a thorough exposition of the federal sentencing process, describing the Sentencing Guidelines, the probation office ’s role in sentencing, and the fact that the court would not be bound by the advisory guideline range.

In addition to the plea agreement, Van Vliet also received a factual proffer from the government, which contained the factual basis for Riolo’s guilty plea. Upon reviewing the proffer with Van Vliet, Riolo raised a concern about a statement that Sterling 10 Opinion of the Court 20-12206

purportedly traded in “futures contracts.” Crim. Doc. 14 at 6 ¶ 2. Van Vliet raised his concern to the government, but it refused to remove the futures contract language. She advised Riolo that the language would not affect his guideline range. He signed the factual proffer. C. Change-of-Plea Hearing, Sentencing, and Direct Appeal In July 2009, the district court held a change-of-plea hearing.

At the hearing, after putting Riolo under oath, the district court asked Riolo whether he had reviewed the plea agreement “fully and completely” with Van Vliet and whether he understood it. Crim. Doc. 16 at 6. He answered yes to both questions. The court then asked whether Riolo had reviewed the factual proffer “fully and completely” with Van Vliet and whether he understood it. Id. at 18. Again, Riolo answered yes to both questions.

The district court also established that Riolo understood how the sentencing process would work following his guilty plea. The court sought to confirm his understanding that the probation office would prepare a PSR to help the court determine his guideline range. Riolo confirmed that he understood. Then, after detailing the procedure for filing objections to the PSR, the court asked whether he knew that the district court—and no one else—would determine his guideline range and sentence. Riolo answered “[y]es.” Id. at 11. Lastly—with three separate questions—the district court asked Riolo whether anyone had made “any promises or representations” to him about his sentence. Id. at 12–13. Three times, Riolo swore that no one had made any promises or 20-12206 Opinion of the Court 11

representations to him about his sentence. After this thorough colloquy , the district court found him competent to enter an informed plea and accepted his plea of guilty.

1. The Presentence Investigation Report Following the change-of-plea hearing, the probation office prepared the PSR. The PSR calculated Riolo’s offense level as 38, not 30 as he and Van Vliet had discussed, and his criminal history category as I, which yielded a guideline range of 235–293 months’ imprisonment.

The PSR included two four-point sentence enhancements that Van Vliet had not included in her estimate. In calculating Riolo ’s guideline range, first, the PSR applied the four-point enhancement for violation of a commodities law by a commodities trading advisor—an enhancement that Van Vliet and Riolo had discussed, but she had not included in her estimate. See U.S. Sent’g Guidelines Manual § 2B1.1(16)(B) (U.S. Sent’g Comm’n 2008). Second, the PSR applied another four-point enhancement for jeopardizing the safety and soundness of a financial institution. See id. § 2B1.1(b)(14)(B). Neither Van Vliet nor the government had anticipated the second enhancement. Van Vliet filed written objections to both enhancements.

2. Riolo’s Reaction to the PSR Calculation Riolo was “shocked” when he read the PSR. Civ. Doc. 83 at 124. He interpreted the PSR’s calculation to mean that “the AUSA 12 Opinion of the Court 20-12206

had gone back on [his] deal” to assign Riolo an offense level of 30, with a guideline range of 97 to 121 months’ imprisonment.

The day before his sentencing, Riolo drafted a letter to the district court. In the letter, he sought to withdraw his plea or, in the alternative, to obtain a continuance of his sentencing hearing, writing that he thought there was a “verbal agreement” between the government and Van Vliet that he “would be released [from prison] around the time [his] oldest child would enter high school.” Civ. Doc. 76-5 at 3–4. He sent the draft letter to Houston, his civil lawyer. Houston told Van Vliet about Riolo’s draft letter, and she responded to Riolo in an email.

Van Vliet’s email encouraged Riolo not to attempt to withdraw his guilty plea. She urged that withdrawal of his plea would prove “a very, very difficult hurdle” and that it may have adverse consequences, including the removal of any reduction he might receive for acceptance of responsibility or for cooperating with the government. Civ. Doc. 74-4 at 1. But, she added, the decision to send the letter was his alone. She encouraged him to seek a second opinion on her advice.

Riolo called another lawyer, Richard Rosenbaum. Houston had known Rosenbaum for years, and Riolo had earlier considered hiring Rosenbaum to represent him instead of Van Vliet. During their telephone conversation, Rosenbaum expressed reservations about helping Riolo withdraw his plea with so little time left before sentencing. Van Vliet was added to the conference call between Riolo and Rosenbaum, and the three discussed Riolo’s case. Riolo 20-12206 Opinion of the Court 13

left the phone call under the impression that Rosenbaum “was going to come and represent [him] at the [sentencing] hearing” the next day. Civ. Doc. 83 at 128.

Later that evening, Rosenbaum called Van Vliet at her home. She summarized for him the plea agreement, what occurred at the change-of-plea hearing, and other information relevant to Riolo’s case. Rosenbaum agreed with her opinion that Riolo should not attempt to withdraw his plea. Rosenbaum was reluctant to tell Riolo, however, because of Rosenbaum’s personal friendship with Riolo and/or Houston. Van Vliet and Rosenbaum agreed to tell Riolo that Rosenbaum could not represent him for another reason —because Rosenbaum did not have enough time to process the plea withdrawal before the sentencing the next morning. Rosenbaum sent Van Vliet an email communicating that he could not represent Riolo because he could not “clear conflict check and file a Notice of Appearance” before sentencing and asking her to forward the email to Riolo. Civ. Doc. 6 at 9.

3. Riolo’s 293-Month Sentence and Direct Appeal The following morning, the district court held the sentencing hearing. There, Van Vliet reiterated her objections to the two four-point enhancements. She conceded that the four-point commodities law enhancement could properly be applied to Riolo’s sentence under the law, but she maintained her objection that the enhancement should not apply because the prerequisites for the enhancement existed only for a three-month period over the course of the ten-year offense conduct. She also maintained her 14 Opinion of the Court 20-12206

objection to the four-point enhancement for jeopardizing the soundness of a financial institution. As to this enhancement, Van Vliet argued that Sterling and LaSalle were not legitimate financial institutions and so the enhancement for endangering the safety and soundness of a financial institution should not apply. The district court overruled the objections and adopted the findings in the PSR as the findings of the court. After hearing live testimony from multiple victims, the district court imposed a sentence of 293 months’ imprisonment. 3 Riolo appealed, arguing in part that “the government promised him an offense level of 30,” rather than 38. United States v. Riolo (Riolo I), 398 F. App’x 568, 570 (11th Cir. 2010) (unpublished). The panel observed that “the record d[id] not contain any evidence of such an agreement” and also noted that Riolo represented under oath at the change-of-plea hearing “that no one had made any promises or representations with respect to his sentence.” Id. The panel affirmed. See id. at 571. D. Riolo’s 28 U.S.C. § 2255 Motion and Previous Collateral Appeals

3 Riolo’s sentence consisted of 240 months’ imprisonment on each of counts one through four, all to run concurrently, and 53 months on count five, to run consecutively to the sentence on the other counts, for a total of 293 months.

20-12206 Opinion of the Court 15

Riolo next moved pro se to vacate his sentence pursuant to 28 U.S.C. § 2255. 4 His motion raised three distinct grounds for relief on a theory of ineffective assistance of counsel. Under each ground he listed several claims. The magistrate judge prepared a report and recommendation, consolidating what appeared to be duplicative and overlapping grounds and claims into three broad theories for relief. After analyzing each one, the magistrate judge recommended that Riolo’s § 2255 motion be denied. Following Riolo ’s objections, the district court adopted the magistrate judge’s recommendation and entered a judgment. Riolo appealed.

On appeal, a panel of this Court construed Riolo’s pro se motion as containing 13 ineffective assistance of counsel claims. See Riolo v. United States (Riolo II), 567 F. App’x 684, 685–87 (11th Cir. 2014) (unpublished). The panel ruled that the district court had overlooked a few distinct ineffective assistance claims in denying Riolo’s motion. See id. at 688. Citing Clisby v. Jones, 960 F.2d 925 (11th Cir. 1992) (en banc), the panel vacated the district court’s judgment and remanded the case for the district court to consider the unaddressed claims. Riolo II, 567 F. App’x at 688.

On remand, the district court referred the case back to the magistrate judge for further review. The magistrate judge entered another report and recommendation, this time addressing the claims the district court had been directed to consider on remand.

4 Although Riolo initially proceeded pro se, he later retained counsel to represent him.

16 Opinion of the Court 20-12206

The magistrate judge found no merit in the remaining claims and again recommended that Riolo’s motion be denied. Riolo filed objections . The district court adopted the magistrate judge’s recommendation over the objections and entered a final judgment. Riolo appealed again.

On his second collateral appeal, Riolo argued that the district court failed to hold an evidentiary hearing to which he was statutorily entitled. See Riolo v. United States (Riolo III), 783 F. App’x 917, 918 (11th Cir. 2019) (unpublished). The panel observed that Riolo had made allegations that, if true, could amount to ineffective assistance of counsel. For example, Riolo alleged in his motion that Van Vliet advised him “that the offense level had been predetermined ” and that “there was no possibility” he would receive a higher sentence than what was represented by her initial guideline estimation. Id. at 919, 922 (internal quotation marks omitted). He also alleged that Van Vliet “never informed him that the probation office would conduct its own independent calculation of the offense level for the court.” Id. at 922 (alteration adopted) (internal quotation marks omitted). The panel vacated the judgment and remanded the case to the district court for an evidentiary hearing to determine whether Riolo’s allegations were true. See id. at 923. E. The District Court’s Order and Judgment On remand the district court held a two-day evidentiary hearing. Riolo testified at the hearing. He also presented testimony from other witnesses, including Gary and Houston. The 20-12206 Opinion of the Court 17

government’s witnesses included Van Vliet, McManus, and Sheila Tierney, the probation officer who prepared Riolo’s PSR.5 After the hearing, the district court entered a written order and judgment denying Riolo’s § 2255 motion. The court credited Van Vliet’s testimony over the testimony from Riolo and his other witnesses. It also considered the colloquy at the change-of-plea hearing. Ultimately, the district court found that Van Vliet had never represented to Riolo that she had a deal with the government about his guideline range.

The district court made other findings of fact reflecting that Van Vliet had thoroughly informed Riolo of the federal sentencing process and the implications of his guilty plea. Specifically, the district court found that—before the change-of-plea hearing—Van Vliet reviewed each provision of the plea agreement with Riolo, went over the factual proffer with him, explained the probation office ’s role in sentencing, and discussed that the district court would not be bound by any parties’ proposed calculation under the Sentencing Guidelines. The court also found that Van Vliet did not thwart Riolo’s attempt to withdraw his plea the day before sentencing when she communicated with Rosenbaum about his case. After making these findings of fact, the district court determined that Van Vliet did not provide ineffective assistance of counsel to Riolo. The court denied Riolo’s § 2255 motion and granted a certificate of

5 Relevant testimony from the evidentiary hearing is summarized above in Parts I.A–C.

18 Opinion of the Court 20-12206

appealability on the question “of whether [his] guilty plea was knowingly entered.” Doc. 91 at 29. 6 This is Riolo’s appeal.

II. LEGAL STANDARD In § 2255 proceedings, we review legal conclusions de novo and factual findings for clear error. See Martin v. United States, 949 F.3d 662, 667 (11th Cir. 2020). “A claim of ineffective assistance of counsel is a mixed question of law and fact reviewed de novo.” Id.

III. DISCUSSION

The question of whether an attorney provided ineffective assistance of counsel in the context of a guilty plea is subject to the familiar two-part inquiry first spelled out in Strickland v. Washington , 466 U.S. 668 (1984). The movant “must show (1) his counsel’s performance was deficient and (2) the deficient performance prejudiced his defense.” Martin, 949 F.3d at 667 (citing Strickland, 466 U.S. at 687). Although the prejudice inquiry differs somewhat in

6 We note that the certificate of appealability is ambiguous in identifying the question on appeal. On the one hand, the question could be understood as asking whether the district court conducted a sufficient colloquy at the change-of-plea hearing to render Riolo’s plea voluntary. On the other hand, the question could be understood as asking whether Van Vliet provided ineffective assistance of counsel in preparing Riolo to plead guilty such that his plea was not voluntary. The parties have consistently treated the district court’s certificate of appealability as covering the latter question, and we do the same.

20-12206 Opinion of the Court 19

the guilty plea context, our review of the adequacy of an attorney’s performance is much the same. To show deficient performance, the movant must establish that his attorney’s representation “fell below an objective standard of reasonableness.” Id. (internal quotation marks omitted). The “petitioner bears the heavy burden of showing that no competent counsel would have taken the action that his counsel did take.” Gissendaner v. Seaboldt, 735 F.3d 1311, 1323 (11th Cir. 2013) (internal quotation marks omitted).

In this case, Riolo challenges the district court’s findings of fact that underlaid the court’s conclusion that Van Vliet’s performance in representing him was not deficient. Given our standard of review, he can prevail in this appeal only if he can show that the findings of fact were clearly erroneous. We therefore begin by reviewing the district court’s findings of fact for clear error. Seeing no clear error in the district court’s findings, we next consider whether Van Vliet provided ineffective assistance of counsel under the facts found by the district court. A. The District Court Committed No Clear Error in Making Its Challenged Factual Findings. Riolo contends that five of the district court’s factual findings were clearly erroneous. Specifically, he asserts that the district court clearly erred in finding that, before the change-of-plea hearing , Van Vliet: (1) reviewed each provision of the plea agreement with him; (2) reviewed the factual proffer with him days before the hearing; (3) explained the probation office’s role in sentencing to him; and (4) never communicated to him that there was a binding 20 Opinion of the Court 20-12206

agreement with the government that his offense level would be 30 and that he would have a guideline range of 97–121 months’ imprisonment . Riolo further contends that the district court clearly erred in finding that Van Vliet (5) did not thwart his attempt to withdraw his guilty plea after the change-of-plea hearing.

A finding of fact is clearly erroneous only when we are “left with a definite and firm conviction that a mistake has been committed .” United States v. Almedina, 686 F.3d 1312, 1315 (11th Cir. 2012) (internal quotation marks omitted). When there are two reasonable constructions of the evidence, “the factfinder’s choice between them cannot be clearly erroneous.” Id. (internal quotation marks omitted).

The district court made its findings of fact after crediting Van Vliet’s testimony over Riolo’s and his other witnesses’ testimony. “[W]e allot substantial deference to the factfinder . . . in reaching credibility determinations with respect to witness testimony.” Rivers v. United States, 777 F.3d 1306, 1316 (11th Cir. 2015) (alteration adopted) (internal quotation marks omitted). “Generally, we refuse to disturb a credibility determination unless it is so inconsistent or improbable on its face that no reasonable factfinder could accept it.” Id. at 1317 (internal quotation marks omitted). Keeping in mind that we review the challenged findings of fact for clear error and that we extend substantial deference to the district court’s credibility determinations, we proceed to the findings of fact in question.

20-12206 Opinion of the Court 21

1. The District Court Did Not Clearly Err in Finding that Van Vliet Reviewed the Plea Agreement with Riolo Before the Change-of-Plea Hearing.

Riolo asserts that the district court clearly erred in finding that Van Vliet reviewed each provision of the plea agreement with him before the change-of-plea hearing. In support of his argument, Riolo points to an inconsistency in Van Vliet’s testimony at the evidentiary hearing. She testified that she reviewed the plea agreement with Riolo at the May 7, 2009 meeting—the meeting that Gary and Houston also attended. As the government concedes, in fact Van Vliet received the plea agreement on May 12, 2009, at the earliest. Because she testified that she reviewed the plea agreement with Riolo before she actually received the document, he argues, the district court clearly erred in finding that she reviewed each provision of the plea agreement with him before the change-of-plea hearing.

Although Riolo is correct that a discrepancy exists, he attempts to make a mountain out of this molehill. As an initial matter , the district court did not find that that Van Vliet reviewed the plea agreement with Riolo on May 7, 2009. Instead, the court found that she “went through each provision of the plea agreement” with him at some point before the change-of-plea hearing. Civ. Doc. 91 at 23. Riolo offers no argument, nor any record citation, to undermine the finding that Van Vliet reviewed the plea agreement with him before the change-of-plea hearing.

22 Opinion of the Court 20-12206

Indeed, the record amply supports the district court’s finding . In response to the court’s question at the change-of-plea hearing regarding whether he had read the plea agreement and discussed it “fully and completely” with Van Vliet, Riolo answered, under oath, “Yes, I did.” Crim. Doc. 16 at 6. “There is a strong presumption that the statements made during the [plea] colloquy are true.” United States v. Medlock, 12 F.3d 185, 187 (11th Cir. 1994). Riolo “bears a heavy burden” to show that his own statements, made under oath at the time, were false. United States v. Rogers, 848 F.2d 166, 168 (11th Cir. 1988). At the evidentiary hearing, Riolo attempted to explain away his testimony during the plea colloquy by saying that he was “very emotional” and not as “focused as [he] perhaps should have been on the questions being asked,” which led him to make admissions that were not truthful. Civ. Doc. 83 at 119. But the district court, having observed Riolo’s demeanor at the evidentiary hearing, did not find this explanation persuasive. Riolo has not come close to carrying his heavy burden to show that his own statements made during the plea colloquy were false. See Rogers , 848 F.2d at 168.

Apart from Riolo’s own testimony at the change-of-plea hearing, other evidence in the record supports the district court’s finding that Van Vliet reviewed the plea agreement with Riolo. She testified that she went through the “specific terms of [the] plea agreement” with him before the change-of-plea hearing—testimony that the district court credited over Riolo’s. Civ. Doc. 84 at 55. And her billing records indicate that she and Riolo spoke on 20-12206 Opinion of the Court 23

May 19, 2009, which was before the change-of-plea hearing. According to her billing records for that day, she “follow[ed] up with him regarding . . . [the] plea agreement.” Civ. Doc. 76-4 at 7. Riolo fails to demonstrate that Van Vliet’s testimony was “so inconsistent or improbable on its face that no reasonable factfinder could accept it,” notwithstanding Van Vliet’s inability to remember precise dates more than 10 years after the fact. Rivers, 777 F.3d at 1317 (internal quotation marks omitted). We conclude that there was no clear error in the district court’s finding that Van Vliet reviewed each provision of the plea agreement with Riolo at some point before the change-of-plea hearing.

2. The District Court Did Not Clearly Err in Finding that Van Vliet Reviewed the Factual Proffer with Riolo Before the Change-of-Plea Hearing.

Riolo next argues that the district court clearly erred in finding that Van Vliet reviewed the factual proffer with him days before his change-of-plea hearing. The thrust of Riolo’s argument is that he received the factual proffer on the morning of the change- of-plea hearing and did not have time to digest and understand it before the hearing took place.

He rests his argument that he received the factual proffer the morning of the change-of-plea hearing on another chronological inconsistency in Van Vliet’s testimony. Van Vliet initially took the position that she sent the factual proffer, along with the plea agreement , to Riolo on May 18, 2009, in an email. Riolo counters that there was no factual proffer attached to her May 18, 2009 email. At 24 Opinion of the Court 20-12206

the evidentiary hearing, however, Van Vliet acknowledged that she had been mixed up on the dates, but she maintained that she sent the factual proffer two or three days before the change-of-plea hearing. The district court credited her live testimony.

The discrepancy, which Van Vliet explained to the district court’s satisfaction, does not establish that the court’s finding was clearly erroneous. Again, Riolo’s sworn testimony at the plea colloquy supports the district court’s finding. As with the plea agreement , the district court asked whether Riolo had read the factual proffer and discussed it “fully and completely” with Van Vliet. Crim. Doc. 16 at 18. Riolo answered, “Yes, I did.” Id. At the evidentiary hearing, he admitted to asking Van Vliet about language in the factual proffer about futures contracts. She testified that she raised Riolo’s concerns to the prosecutor and sought to have the futures language removed, but the prosecutor refused. Riolo’s and Van Vliet’s testimony at the evidentiary hearing that they discussed the factual proffer together and had time to raise concerns to the prosecutor, who in turn had time to deny their request, supports the district court’s finding that Van Vliet reviewed the factual proffer with Riolo days before the change-of-plea hearing. As if all this were not enough, Riolo admitted on cross-examination that he went over the factual proffer with Van Vliet and had enough time to review it. There is no clear error in the district court’s finding of fact.

20-12206 Opinion of the Court 25

3. The District Court Did Not Clearly Err in Finding that Van Vliet Reviewed the Probation Office’s Role in Sentencing with Riolo Before the Change-of-Plea Hearing.

Riolo next asserts that the district court clearly erred in finding that Van Vliet reviewed the probation office’s role in sentencing ahead of the change-of-plea hearing. For this factual finding, Riolo points to no chronological inconsistency in Van Vliet’s testimony . Instead, he merely asserts that she was untruthful in testifying that she told him about the probation office’s role in sentencing .

We are not persuaded. Once again, the plea colloquy supports the district court’s finding. During the colloquy, the district court asked if Riolo understood that the probation office would prepare a PSR and what the presentence investigation process would entail. Riolo responded that he understood. Provided with no reason to doubt this exchange at the plea colloquy, we again find that Riolo’s sworn answers supported the district court’s finding .

But that evidence was not all. Recall McManus testified that he attended a lunch meeting with Van Vliet and Riolo on May 11, 2009, where Van Vliet communicated the probation office’s role to Riolo. And Van Vliet also testified—credibly, according to the district court—at the evidentiary hearing that “on more than one occasion ” she explained to Riolo that the probation office would “make a guidelines calculation.” Civ. Doc. 83 at 238–39. All this 26 Opinion of the Court 20-12206

evidence supports the district court’s finding that Van Vliet informed Riolo of the probation office’s role in federal sentencing. He has failed to demonstrate clear error in this finding of fact.7 4. The District Court Did Not Clearly Err in Finding that Van Vliet Never Told Riolo that the Government Had Agreed to an Offense Level of 30.

Riolo asserts that the district court erred in finding that Van Vliet never told him that she had a deal with the government establishing that he would have an offense level of 30 and a guideline range of 97–121 months’ imprisonment. 8 No clear error has been

7 In addition, it is undisputed that Riolo independently reviewed the plea agreement before the change-of-plea hearing. The plea agreement also informed Riolo of the probation office’s role in sentencing, noting that “the Court w[ould] compute an advisory sentence under the Sentencing Guidelines and . . . the applicable guidelines will be determined by the Court relying in part on the results of a Pre-Sentence Investigation by the United States Probation Office . . . , which . . . will commence after the guilty plea has been entered .” Civ. Doc. 74-5 at 21 ¶ 2. 8 Riolo argues in his brief:

The district court improperly avoided testimony and evidence regarding the alleged offense level agreement, and then erred by misinterpreting Riolo’s claims of an agreement on the offense level instead as a sentence guarantee, and by judging Riolo ’s reasonable understanding of the agreement on defense counsel’s post hoc alleged understanding rather than Riolo’s understanding.

20-12206 Opinion of the Court 27

shown. The plea colloquy again supports the district court’s finding . The district court asked three times whether anyone had made “any promises or representations” to Riolo about his sentence. Crim. Doc. 16 at 12–13 (emphasis added). In response, Riolo swore three times that he never received any promises or representations about what his sentence might be. Riolo later testified at the evidentiary hearing that he thought the sentencing judge was referring to “unseemly type[s] of things” in this line of questioning. Civ. Doc. 83 at 122. We cannot say Riolo has carried his burden to show that his own statements, made under oath at the time, were false. Rogers, 848 F.2d at 168.

Other evidence in the record supports the district court’s finding that Van Vliet never told Riolo she had a deal with the government about his offense level or guideline range. McManus testified that at the May 11, 2009 lunch meeting, Van Vliet informed Riolo that “[i]t’s up to the judge to decide” his sentence and that “[the judge] can depart upwards [or] downwards.” Civ. Doc. 83 at 192. And Van Vliet testified that she “never said that [the government would] recommend 97 months or anything other than what was in the plea agreement.” Id. at 232. All this evidence supports the district court’s finding.

Appellant’s Br. at 17. In substance, we understand Riolo to be arguing that the district court clearly erred in finding that Van Vliet never told him that there was a binding agreement regarding his offense level and guideline range.

28 Opinion of the Court 20-12206

Riolo points to the email Van Vliet sent to him on May 18, 2009, in which she wrote: “I have talked to [the prosecutor] and his calculations match ours. That is a[n offense] level 30 with a range of 97 to 121 months.” Civ. Doc. 76-2 at 2. This email, Riolo argues, substantiates his assertion that Van Vliet had been telling him all along that there was a binding agreement between her and the government as to his sentencing range. But the email is not the smoking gun Riolo seems to think it is. The email says nothing about an agreement between Van Vliet and the government. Rather, it indicates that Van Vliet’s and the government’s independently-made guideline estimates “match[ed].” Id. Van Vliet appears to have sought to reassure herself and Riolo that her guideline calculation was correct by comparing it with the government’s. The email does not convince us that the district court clearly erred in finding that she never told Riolo that she had a deal with the government about his guideline range. We find no clear error here.

But even if we were to assume that Van Vliet did tell Riolo that there was an agreement between herself and the government with respect to his offense level or guideline range, the plea agreement and portions of the plea colloquy sufficiently dispelled any notion that such an agreement would be binding. The plea agreement , for example, contained a detailed explanation about how the court would determine his sentence. It explained that the probation office would prepare a PSR which would contain an estimate of his guideline range. It also explained to Riolo that, although the district court had to consider the advisory guideline range, it would 20-12206 Opinion of the Court 29

not be bound by the range in sentencing him; the court could impose an appropriate sentence above or below the guideline range.

At the change-of-plea hearing, the district court made sure Riolo understood that Van Vliet could not have made any binding agreement with the government about his guideline range or about his sentence. It told Riolo that “even though [Van Vliet] may have advised you based upon her understanding of the facts of the case and her understanding of the law and the Sentencing Guidelines what she thinks or believes, in her best professional opinion, the advisory guideline sentencing range will turn out to be, . . . my decisions may be different from what she has advised you.” Crim. Doc. 16 at 11. We are more than satisfied that Riolo was apprised that the district court would not be bound by any guideline estimate Van Vliet may have given him.

5. The District Court Did Not Clearly Err in Finding that Van Vliet Never Prevented Riolo from Withdrawing His Guilty Plea.

Lastly, Riolo asserts that the district court clearly erred in finding that Van Vliet never “thwarted” his attempt to withdraw his guilty plea. Appellant’s Br. at 37. To support this argument, Riolo points to the events that occurred the day before his sentencing . Riolo sent an email to Houston containing a draft letter addressed to the district court in which he requested a plea withdrawal or, in the alternative, a continuance. Van Vliet got word of Riolo’s letter and promptly wrote to advise him not to attempt to withdraw his plea. She told him that the decision was his alone, 30 Opinion of the Court 20-12206

however, and encouraged him to seek a second opinion. After he received her email, Riolo contacted Rosenbaum, another attorney. Van Vliet testified that later that evening Rosenbaum called her to discuss the case in greater detail. During that conversation, Rosenbaum reached the conclusion that “there was no viable basis for withdrawal.” Civ. Doc. 74-5 at 5 ¶ 16. Rosenbaum then asked Van Vliet to communicate to Riolo that he could not represent him at sentencing. Taken together, Riolo contends that these circumstances establish that Van Vliet thwarted his attempt to withdraw his guilty plea.

Riolo’s argument lacks merit. Van Vliet merely gave Riolo advice on the possible consequences of attempting to withdraw his guilty plea the day before sentencing. She did not stand in the way; she reminded him that the decision was his alone. She also encouraged him to get a second opinion. When he took her advice and sought a second opinion from Rosenbaum, she laid out the circumstances of his case for Rosenbaum, who independently concluded that Riolo should not attempt to withdraw his guilty plea. Riolo provides us with no basis to conclude that there is clear error in the district court’s finding on this issue. B. Van Vliet’s Representation of Riolo Was Not Deficient Under Strickland. Having reviewed all the factual findings Riolo challenges on appeal, we now turn to whether the district court erred in concluding that Van Vliet did not provide constitutionally ineffective assistance of counsel to Riolo. Our answer is no. First, we explain why 20-12206 Opinion of the Court 31

our rejection of his argument that the district court’s factual findings must be overturned renders Betancourt v. Willis, 814 F.2d 1546 (11th Cir. 1987)—the case upon which he principally relies— inapposite. Second, we review his argument that Van Vliet’s miscalculation of his guideline range by itself amounts to ineffective assistance of counsel.

1. Given the District Court’s Findings of Fact, Betancourt Is Inapposite.

Riolo relied primarily on Betancourt in his brief and at oral argument. In Betancourt, we affirmed the grant of a § 2254 petition in which the petitioner, Jairo Betancourt, argued that his attorneys provided ineffective assistance because of assurances they made about his sentence before he pled guilty. Id. at 1547. Betancourt argued that his attorneys told him that if he pled guilty, the state court would initially impose a twelve-year sentence but “promised to reduce his sentence later to equal the lowest sentence received by either of his co-defendants in federal court.” Id. Betancourt pled guilty based on his attorneys’ representations. Id. at 1548. He was sentenced to twelve years’ imprisonment and three years’ probation . Id.

After one of Betancourt’s co-defendants was sentenced to five years’ imprisonment, Betancourt’s lawyer filed a motion with the state court to reduce his sentence. Id. The state court judge, “not recalling any sentence reduction agreement and finding no evidence of it on the record, denied the motion.” Id. Betancourt then moved to withdraw his plea and to recuse the sentencing judge.

32 Opinion of the Court 20-12206

The sentencing judge voluntarily recused, and another state court judge held an evidentiary hearing on Betancourt’s motion to withdraw his plea. The state court ultimately denied his motion. Id.

Betancourt filed a petition for habeas corpus in federal court.

Id. The district court granted his habeas petition based on ineffective assistance of counsel. The court found that “the representation of [Betancourt’s] counsel fell below an objective standard of reasonableness by representing to [him] that the court had agreed to a later sentence reduction, by failing to memorialize the alleged plea agreement by letter, affidavit or other appropriate means, and by neglecting to enter it upon the record.” Id. The district court also found that Betancourt was prejudiced by the defective representation . Id. We affirmed on appeal, observing that the evidence was “uncontroverted that [Betancourt] was completely unaware of the ultimate consequences of his plea because his counsel misrepresented the existence of a sentence reduction agreement.” Id. at 1549.

The circumstances here are far different from those in Betancourt. As explained above, we are bound to accept the district court’s finding that Van Vliet never told Riolo there was any binding agreement between her and the government about his sentence . Therefore, Betancourt is entirely distinguishable. Riolo has failed to show that Van Vliet provided ineffective assistance of counsel of the kind we considered in that case.

20-12206 Opinion of the Court 33

2. Van Vliet’s Miscalculation of Riolo’s Guideline Range Did Not Amount to Deficient Performance.

Riolo next asserts an ineffective-assistance claim apart from his challenges to the district court’s findings of fact. He points out it is undisputed that Van Vliet underestimated his guideline range by more than 100 months. That alone, he contends, constitutes ineffective assistance of counsel.

We turn to Strickland’s two-pronged inquiry for assessing ineffective assistance claims. We must ask whether (1) counsel’s performance was deficient and if (2) the deficient performance prejudiced the movant. See Martin, 949 F.3d at 667 (citing Strickland, 466 U.S. at 687). “Surmounting Strickland’s high bar is never an easy task, and the strong societal interest in finality has special force with respect to convictions based on guilty pleas.” Id. (internal quotation marks omitted). Because a § 2255 movant must satisfy both prongs of Strickland, we need not consider one prong if the defendant fails to satisfy the other. See Holladay v. Haley, 209 F.3d 1243, 1248 (11th Cir. 2000).

Our analysis begins and ends with Strickland’s first prong:

Van Vliet’s performance was not deficient. To show deficient performance , Riolo must establish that Van Vliet’s representation “fell below an objective standard of reasonableness.” See Martin, 949 F.3d at 667 (internal quotation marks omitted). As an initial matter, we note that at the time Van Vliet represented Riolo, she was a criminal lawyer with a great deal of experience with the Sentencing Guidelines and advising clients about the federal sentencing 34 Opinion of the Court 20-12206

process. Uncontroverted portions of her testimony indicate she had been in private practice for 18 or 19 years. Before that, she worked in the Criminal Division of the Department of Justice and as an Assistant United States Attorney. Although not determinative , Van Vliet’s experience bolsters the presumption that the quality of her representation did not fall below the objective standard of reasonableness. See Provenzano v. Singletary, 148 F.3d 1327, 1332 (11th Cir. 1998) (observing the bolstered presumption of reasonableness where the attorney had been practicing for 20 years).

To be sure, experienced attorneys make mistakes. Van Vliet’s estimate of Riolo’s guideline range was far off the mark—by more than 100 months. Riolo points to the Fifth Circuit case of United States v. Herrera to argue that a petitioner may have a legitimate ineffective-assistance claim when an attorney gives the petitioner incorrect advice regarding his exposure under the Sentencing Guidelines. See 412 F.3d 577, 580 (5th Cir. 2005) (remanding for an evidentiary hearing on petitioner’s claim that his counsel underestimated his guideline range by 27 months). But we need not decide this issue today. To resolve this appeal, we can assume, without deciding, that a miscalculation of sufficient magnitude can constitute deficient performance and cause prejudice under Strickland.

Ineffective-assistance claims are fact-bound, and here the well-developed factual record convinces us that Van Vliet’s miscalculation was not the product of deficient performance. First of all, even though Van Vliet estimated Riolo would receive an offense level of 30 with a range of 97–121 months’ imprisonment, at the 20-12206 Opinion of the Court 35

same time she warned him that the district court could impose an additional four-point enhancement for “a violation of commodities law” while he was “a commodities trading advisor.” U.S. Sent’g Guidelines Manual § 2B1.1(16)(B)(ii) (U.S. Sent’g Comm’n 2008). Riolo admitted that “[s]he raised the possibility” that the enhancement might apply. Civ. Doc. 83 at 143. Second, Van Vliet objected to the enhancement after the probation office produced the PSR. We cannot say that under these circumstances—where counsel anticipated the enhancement at issue, warned her client that it could apply, and then objected to its inclusion in the PSR—her performance was deficient.

The second four-point enhancement was imposed for “jeopardiz [ing] the safety and soundness of a financial institution.” Id. at § 2B1.1(b)(14)(B)(i). Neither Van Vliet nor the government anticipated that this enhancement would apply. No doubt this is in part because our Court had not interpreted it, as both the parties and the district court observed at sentencing. Van Vliet believed the enhancement would not apply because Riolo’s corporations—Sterling and LaSalle—were “fictional” organizations. Crim. Doc. 38 at 4. Because they were not legitimate financial institutions, she argued , he could not qualify for a sentence enhancement that pertained to conduct threatening the safety of legitimate financial institutions . We are loathe to say that an attorney provided deficient performance by failing to anticipate the application of a guideline when the attorney’s reasonable interpretation of that guideline was 36 Opinion of the Court 20-12206

that it had no bearing on the defendant’s situation and no authoritative decision offered guidance to the contrary. 9 To be sure, at the time of Van Vliet’s consultations with Riolo , other circuits had affirmed application of the financial institution enhancement in circumstances involving sham financial institutions . See, e.g., United States v. Hoffecker, 530 F.3d 137, 201 (3d Cir. 2008) (“[T]here is nothing in the guideline to suggest that the Commission intended to limit the enhancement only to apply to legitimate financial institutions; we will not read that limitation into the language of the guideline.”), superseded by regulation on other grounds as stated in, Rad v. Att’y Gen. United States, 983 F.3d 651, 668 n.13 (3d Cir. 2020); United States v. Collins, 361 F.3d 343, 348 (7th Cir. 2004) (“[W]hen it walks and talks like a financial institution , even if it’s a phony one, it is . . . covered by [the Guideline].” (quoting United States v Randy, 81 F.3d 65, 69 (7th Cir. 1996)(emphasis in original)); see also United States v. Dale, 374 F.3d 321, 328–330 (5th Cir. 2004) (affirming application of the financial institution guideline to a fraudulent entity the panel characterized as a “Ponzi scheme”), vacated on other grounds, 543 U.S. 1113 (2005). 10 But “[t]he test for ineffectiveness is not whether counsel

9 We express no view on the merits of Van Vliet’s argument about the proper application of the financial institution enhancement. 10 The decisions discussed in this paragraph considered the financial institution enhancement using the 1997 version of the Sentencing Guidelines. See Hoffecker, 530 F.3d at 196 n.6.; Collins, 361 F.3d at 345; Dale, 374 F.3d at 327– 20-12206 Opinion of the Court 37

could have done more; perfection is not required.” Waters v. Thomas, 46 F.3d 1506, 1518 (11th Cir. 1995) (en banc). Rather, the question is whether counsel’s representation fell “within the wide range of reasonable professional assistance.” Id. (internal quotation marks omitted). Van Vliet anticipated all but one of the sentencing enhancements the district court ultimately imposed. She double checked her own work, reaching out to the government to learn of its calculations. The reasonableness of Van Vliet’s representation is underscored by the fact that the government independently arrived at the same guideline calculation. And when the probation office applied the enhancements, Van Vliet promptly objected. Under these circumstances, Van Vliet’s performance was far from defective under Strickland’s standard.

Given our determination that Van Vliet’s conduct met the objective standard of reasonableness, we have no cause to consider Strickland’s second prong—prejudice. See Holladay, 209 F.3d at 1248. The district court properly denied Riolo’s § 2255 motion.

28. In the 1997 version, the financial institution enhancement appeared in Part F. See U.S. Sent’g Guidelines Manual § 2F1.1(b)(6) (U.S. Sent’g Comm’n 1997). The relevant provision was moved to Part B, but it remained materially unchanged in the 2008 manual, which was the most recent version available to Van Vliet at the time of Riolo’s sentencing. See U.S. Sent’g Guidelines Manual § 2B1.1(14)(B)(i) (U.S. Sent’g Comm’n 2008).

38 Opinion of the Court 20-12206

IV. CONCLUSION

For the above reasons, we affirm the district court’s dismissal of Riolo’s § 2255 motion.

AFFIRMED.

20-12206 JORDAN, J., Concurring 1

JORDAN, Circuit Judge, Concurring:

I join Judge Jill Pryor’s comprehensive opinion for the court in full. I write to point out that the majority of our sister circuits have held that significant errors in advice about sentencing exposure can, depending on the circumstances, constitute deficient performance . See, e.g., United States v. Mayhew, 995 F.3d 171, 178– 79 (4th Cir. 2021); United States v. Herrera, 412 F.3d 577, 581–82 (5th Cir. 2005); Magana v. Hofbauer, 263 F.3d 542, 550 (6th Cir. 2001); United States v. Gordon, 156 F.3d 376, 380 (2d Cir. 1998); United States v. Gaviria, 116 F.3d 1498, 1512 (D.C. Cir. 1997); United States v. Day, 969 F.2d 39, 42–44 (3d Cir. 1992); Iaea v. Sunn, 800 F.2d 861, 864–65 (9th Cir. 1986). Contra United States v. Gordon , 4 F.3d 1567, 1570–71 (10th Cir. 1993) (“A miscalculation or erroneous sentence estimation by defense counsel is not a constitutionally deficient performance rising to the level of ineffective assistance of counsel.”). We have avoided the issue in the past, see Diveroli v. United States, 803 F.3d 1258, 1263 (11th Cir. 2015), and do so again today, but we will have to confront it at some point.