UNITED STATES of America, Plaintiff-Appellee, v. Eric A. BLOOM, Defendant-Appellant, 846 F.3d 243

HAMILTON, Circuit Judge. In August 2007 Sentinel Management Group collapsed. Sentinel managed short-term cash investments for futures commission merchants, individuals, hedge funds, and other entities. Its bankruptcy left customers and creditors in the lurch: over $600 million was lost. In the wake of the collapse, Sentinel president and CEO Eric Bloom was convicted of wire fraud and investment adviser fraud. The government’s case rested on three theories.

First, the government presented evidence that Bloom, despite assuring customers otherwise, put their funds at significant risk by using them as collateral for Sentinel’s risky proprietary trading. Second, the government contended that Bloom fraudulently manipulated the rates of return paid to clients on their investments. Third, the government claimed that Bloom continued to accept new customer funds even after he knew that Sentinel was about to collapse. The jury found Bloom guilty on all counts, eighteen of wire fraud and one of investment adviser fraud.

On appeal, Bloom offers five arguments, which we address in turn. First, Bloom challenges the sufficiency of the evidence supporting his convictions. Second, he argues that his convictions were tainted by prosecutorial misconduct. Third, Bloom argues that the court erred by refusing to instruct the jury properly on the meaning of a federal regulation governing futures commission merchants. Fourth, he challenges several of the district court’s evi-dentiary rulings. Fifth, he argues that in sentencing the district court used too high a loss amount to calculate the sentencing guideline range.

We find no reversible error. I. Factual Background We first provide an overview of Sentinel’s business and its representations to customers regarding how it used their funds. Then we summarize Sentinel’s collapse in 2007 and Bloom’s later indictment, conviction, and sentencing. A. Sentinel’s Business Model Sentinel was founded in 1979 by Philip Bloom, the father of defendant Eric Bloom. The company had a single office in Northbrook, Illinois, and had about twenty-one employees. Sentinel managed investments for various clients such as hedge funds, financial institutions, and individuals.

Its primary business was handling short-term investments for futures commission merchants, also known as FCMs. FCMs represent investors who trade in the futures and options markets, and they are regulated by the Commodity Futures Trading Commission (CFTC). Defendant Eric Bloom joined the company in 1988. He worked in several different positions during his career at Sentinel, sometimes occupying multiple positions at once. Bloom served as head trader from 1988 until 2003 and chief compliance officer until 2006. He also served as president and CEO from 1991 until August 2007 when the company filed for bankruptcy.

Sentinel’s business model was unusual and perhaps unique. It was registered with the CFTC as an FCM, but it did not trade in futures or options. Instead, Sentinel invested funds for other FCMs and, like a mutual fund, paid a return based on profits and losses. Sentinel was the only company that the CFTC permitted to operate in this manner. There was a proviso, however. The CFTC required Sentinel to follow the CFTC regulations for FCMs. In particular, CFTC Rule 1.25 limited the types of securities Sentinel could purchase with customer funds. 17 C.F.R. § 1.25.

To minimize risk of loss and to assure that cash was returnable on demand, Rule 1.25 permitted investment only in highly liquid, highly rated securities such as U.S. Treasury bills. It also required Sentinel to keep the funds of customers segregated from each other and segregated from Sentinel’s own funds. Sentinel signed “seg letters” and Investment Management Agreements to this effect.

These letters and agreements represented to the CFTC and clients that Sentinel segregated customer funds from its own “house” funds, that Sentinel would not have any interest in the customer funds, and that Sentinel would comply with CFTC rules. Sentinel was also registered with the Securities and Exchange Commission (SEC) as an investment adviser. As an adviser, Sentinel owed its clients a fiduciary duty of good faith. SEC v.

Capital Gains Research Bureau, Inc., 375 U.S. 180, 194, 84 S.Ct. 275, 11 L.Ed.2d 237 (1963) (investment adviser has “an affirmative duty of utmost good faith, and full and fair disclosure of all material facts”) (internal quotation marks and citation omitted). Defendant Eric Bloom was named on Sentinel’s investment adviser registration as the person authorized to give investment advice on behalf of the company. The SEC Custody Rule also required segregation of customer funds. 17 C.F.R. § 275.206(4)-2. Sentinel provided two investment options for its clients: the 125 Portfolio and the Prime Portfolio.

The 125 Portfolio was for FCMs, and it allowed them to invest their customers’ funds. This portfolio was intended to provide safe, short-term investments with same-day liquidity. It was subject to CFTC regulations, including Rule 1.25. At the time, Rule 1.25 permitted investment only in securities with a rating of AA or better. After 2007, however, the rule was revised to limit investment to securities that are fully guaranteed by the federal government.

At Sentinel, the pooled accounts of customer funds from the 125 Portfolio were called “Seg 1.” The second option for Sentinel customers—the Prime Portfolio—was for non-FCM clients such as hedge funds, financial institutions, and individuals. FCMs could also invest their own “house” funds (not customer funds) in the Prime Portfolio. The Prime Portfolio was slightly riskier. It was intended to provide a higher rate of return than the 125 Portfolio by investing in securities with longer maturity dates and slightly lower ratings.

Nonetheless, Sentinel and Bloom promised that this portfolio would not stoop below high-quality “investment grade” securities. These funds were called “Seg 3.” In addition to the two public portfolios, Sentinel had a “house account” for its own proprietary trading. This account was not constrained by the grade of securities. It could purchase securities of any rating or no rating. Defendant Bloom’s father, Philip Bloom, owned the majority of the funds in the house account. B.

Sentinel’s Representations to Customers About Its Use of Their Funds Sentinel told its customers their funds would be safe, and it backed this assurance with specific claims about its business model and investment practices. The pitch was that customers could earn higher-than-average interest, receive same-day cash redemptions, and keep their funds effectively bankruptcy-proof. Sentinel claimed this was possible because it pooled cash from multiple clients, which afforded it greater investment flexibility. It made these claims through marketing materials, sales presentations, and its website.

This was an attractive option to many Sentinel customers, particularly FCMs. Since FCMs were investing their customers’ funds, preservation of principal was paramount for them. Liquidity was also vital because FCMs receive margin calls that require them to provide cash to exchanges on behalf of their clients on short notice. Sentinel told customers how it could assure the safety of their customers’ principal.

Its marketing materials said that it purchased “only the highest quality and most liquid securities” and that its “objective [was] to achieve the highest yield consistent with preservation of principal and daily liquidity, not simply ‘the highest yield.’ ” Sentinel told customers of the 125 Portfolio that their supporting securities were highly rated and complied with Rule 1.25. Sentinel also promised that client funds would be segregated and thus thoroughly protected from bankruptcy.

For both portfolios, Sentinel said that it would pool client assets by portfolio and place them in segregated, bankruptcy-proof custodial accounts in the client’s name in the Bank of New York. The accounts would be bankruptcy-proof because, even if Sentinel went bankrupt, the securities remained in segregated accounts in the client’s name and would thus not be considered Sentinel’s assets. And if the Bank of New York failed, the assets would be transferred to another custodial institution. Customers were also told that they would know which securities generated their yields.

Marketing materials stated: “Sentinel sends daily emails ... to each client reporting the total amount invested, the interest earned, and supporting securities.” In other words, customers would know not only how much interest they were earning but also how they were earning it. In a letter to the CFTC, Sentinel said that customers’ profits and losses were based on their ownership shares of an account’s investments; there was no allocation of profits and losses across the different “Seg” accounts.

Sentinel also told customers that it used two means to assure same-day liquidity: repurchase agreements and a loan from the Bank of New York. A repurchase agreement is a transaction where one party (the borrower) sells a security to a counterparty (the lender) with an agreement to repurchase the security later with interest.

Sentinel was involved in both ends of these transactions: sometimes it used its customers’ cash to purchase securities, earning interest when it resold the security (lending); other times Sentinel sold securities to obtain cash quickly, repurchasing the security and paying interest afterwards (borrowing). Sentinel’s website said that it used “overnight” repurchase agreements to facilitate its liquidity needs. The second means to assure same-day liquidity was a loan with the Bank of New York.

Sentinel and Bloom told clients that this loan was specifically for the cash re-demptions of clients in the Prime Portfolio, not the 125 Portfolio. Bloom specified that it was a one-day transaction that was paid off the next day. It was purportedly only for liquidity purposes. Bloom did not tell clients that the loan might be used to fund leveraged trading of their portfolios or to fund other parties outside their portfolios. In Bloom’s trial the government offered extensive evidence that Sentinel’s actual use of customer funds diverged from these representations.

Sentinel’s claims regarding the safety of principal, the use of segregated accounts, and the sources of customers’ interest were often untrue. C. Collapse, Conviction, and Sentencing Sentinel collapsed in August 2007. The company was faring poorly in the spring of that year. By April it had more debts than assets. It had over $2.4 billion in repurchase agreement debt, and its Bank of New York loan balance was over $280 million. In 2007, Sentinel was also highly leveraged. It was using securities in its portfolios as collateral to borrow cash and then using that money to purchase additional securities.

The goal was to earn enough income from the additional investments to beat the cost of borrowing. This came with risks. For instance, if the value of the newly acquired security fell, losses were incurred from both the asset depreciation and the cost of borrowing. In April 2007, Sentinel’s overall leverage ratio was approximately 2 to 1. Seg 3 was leveraged at a ratio of 3 to 1, and the house account was 20 to 1. The extent of the leverage was concerning to Sentinel employees because “if the market turned in the wrong direction,” then Sentinel would suffer a “huge hit.” Sure enough, the market turned down.

By the summer of 2007, Sentinel no longer slouched toward bankruptcy; it careened. National credit markets began to contract, in part because of the crisis in the subprime mortgage industry. Sentinel’s repurchase-agreement counterparties reacted to these market problems by pushing back their securities to Sentinel, many of which were illiquid collateralized debt obligations (CDOs). On June 26, one large counterparty returned approximately $166 million in securities. To compensate, Sentinel borrowed more from the Bank of New York, going well beyond the company’s credit limit.

The loan balance reached a high point on June 28, 2007 of $546 million. Additional counterparties returned securities in July, and Sentinel’s customers demanded redemptions. Sentinel was buckling. On August 4, Sentinel’s outside lawyer advised defendant Bloom to “hold off taking any additional deposits ... even from existing customers.” Bloom and Sentinel refused. From August 6 to August 10, Sentinel accepted approximately $300 million from customers. Though Sentinel continued accepting new funds, customers were seeking redemptions at an even higher rate. The company’s deterioration accelerated.

Friday, August 10, 2007 was Sentinel’s last day in operation. On Monday, August 13, Sentinel emailed its customers to notify them that it was stopping redemptions. On August 16, the Bank of New York demanded that Sentinel repay its loan. The balance was $313 million. On August 17, the bank informed Sentinel that due to Sentinel’s failure to repay the loan, the bank planned to sell Sentinel’s securities that served as collateral for the loan. (Those securities were ultimately returned to Sentinel’s bankruptcy trustee.) August 17 was also the day Sentinel filed for bankruptcy.

This court has addressed Sentinel’s bankruptcy in several appeals. See Grede v. Bank of N.Y. Mellon Corp. (In re Sentinel Management Group, Inc.), 809 F.3d 958 (7th Cir. 2016); Grede v. FCStone, LLC, 746 F.3d 244 (7th Cir. 2014); In re Sentinel Management Group, Inc., 728 F.3d 660 (7th Cir. 2013). In 2012 a grand jury indicted Bloom and Sentinel’s chief trader, Charles Mosley, on eighteen counts of wire fraud and one count of investment adviser fraud.

The indictment alleged that they used customer funds as collateral for Sentinel’s proprietary trading, manipulated customer yield rates, and continued to accept customer funds even after it was clear Sentinel was about to fold. Bloom pled not guilty. Mosley pled guilty to investment adviser fraud and entered into a cooperation agreement with the government; however, he was not called as a trial witness. Bloom’s four-week trial included testimony from numerous former Sentinel employees, financial records, customer accounts, marketing materials, emails, audio recordings, and more.

The jury found Bloom guilty on all counts, and the court denied his motions for judgment of acquittal and a new trial. At sentencing, the court concluded that the Sentencing Guidelines recommended life in prison. After considering the sentencing factors in 18 U.S.C. § 3553(a), however, the court sentenced Bloom to 168 months in prison. The court said that the punishment was based primarily on. the “duty to protect the public by punishing past criminal conduct and deterring future criminal conduct[ ],” and that the sentence would be 168 months regardless of the guideline calculation. II.

Analysis Bloom challenges his convictions and sentence on five grounds. He argues that (

a) the evidence was insufficient to support his convictions; (

b) government misconduct tainted the convictions; (

c) the district court erred in its jury instruction regarding Rule 1.25; (

d) the court erred in several evidentiary rulings; and (

e) his sentence was improper because the court erred in its guideline loss calculation. A. Sufficiency of the Evidence Bloom argues that the evidence presented at trial was insufficient to support the guilty verdicts. He claims that Sentinel’s clients did not sustain losses because he defrauded them; instead, the losses were due to the bad business practices of Sentinel’s chief trader Mosley and the more general decline in securities markets in 2007.

On challenges to the sufficiency of the evidence, we review the evidence in the light most favorable to the guilty verdict and draw “all reasonable inferences in its favor.” United States v. Pust, 798 F.3d 597, 600 (7th Cir. 2015). We will affirm a conviction if a rational trier of fact could have found the defendant committed the essential elements of the crime. Id. This hurdle is high but not insurmountable. See, e.g., United States v. Weimert, 819 F.3d 351, 354 (7th Cir. 2016) (reversing conviction for insufficient evidence). To establish that Bloom committed wire fraud, the government had to prove that:

(1) Bloom participated in a scheme to defraud; (2) with intent to defraud; and (3) interstate wires were used in furtherance of the fraud. 18 U.S.C. § 1343; United States v. Sheneman, 682 F.3d 623, 628 (7th Cir. 2012). Bloom was also convicted of investment adviser fraud, which required the government to prove that Bloom was an investment adviser who willfully used a means of interstate commerce to employ a device or scheme to defraud Sentinel clients or that he willfully aided another in doing so. 15 U.S.C. §§ 80b-6(

l) and (2) and 80b—17; 18 U.S.C. § 2. Bloom claims that the evidence was insufficient to prove that he had the intent to defraud or participated in a scheme to defraud. We find sufficient evidence for the jury to have found Bloom guilty under each of the government’s three theories.

The government provided ample evidence that Bloom was intentionally involved in (1) a scheme to use customer funds as collateral for Sentinel’s mounting loan with the Bank of New York, which was in turn used to finance Sentinel’s own risky trading; (2) a scheme to manipulate client yield rates by reallocating interest from the House and Prime Portfolios to the 125 Portfolio in an effort to make the 125 Portfolio seem more lucrative than it was; and (3) a scheme to solicit and accept new funds during Sentinel’s final days without revealing that the company was on the brink of bankruptcy. 1.

Customer Funds as Collateral First, the government provided sufficient evidence that Bloom intentionally defrauded Sentinel customers by using their funds as collateral for the house’s risky trading. Several witnesses testified to this effect, including Sentinel’s chief financial officer and other employees who helped to arrange collateral for the Bank of New York loan in the form of customer funds. The testimony of the witnesses was corroborated by documents, emails, and recorded phone calls. The evidence showed that Sentinel adopted a high-risk, high-leverage trading strategy.

The strategy involved using securities in Sentinel’s portfolios as collateral to borrow cash through repurchase agreements and from the Bank of New York. The cash was then used to purchase additional securities. This strategy sometimes created leverage in customer accounts, especially the Prime Portfolio, but it created the most leverage in the house account. The house account was highly leveraged for quite some time before Sentinel collapsed.

Sentinel’s chief financial officer testified that she discussed the matter with Bloom because she was concerned about leveraging customer accounts, such as Seg 1, for the benefit- of the house loan that financed the house’s trading. She also emailed him about her concerns on October 19, 2006 and testified about this email during trial. In April 2007 she started sending Bloom monthly leverage reports that discussed the mounting repurchase agreements and the Bank of New York loan. The leverage in the house account continued climbing in the summer of 2007, peaking at a ratio of 52 to 1 in July.

During this time, Sentinel purchased a significant amount of collateralized debt obligations (CDOs), which had low or no ratings and long maturity dates. These securities were much riskier than the type of investments Sentinel told clients it would make with funds invested in its 125 Portfolio or Prime Portfolio. Some of the funding for Sentinel’s trading came from its credit line with the Bank of New York. When funds in the house account were not sufficient collateral for that loan, Sentinel employees testified, they turned to the customer accounts.

One Sentinel employee testified that, in accordance with her training from Bloom, whenever the house account was insufficient collateral for the bank loan, she turned to Seg 1 and Seg 3 for the biggest face-value securities. She then transferred those securities into the lienable accounts that provided collateral for the bank loan. In short, customer funds were used as collateral for Sentinel’s own bank loan, which in turn funded Sentinel’s risky trading. Bloom talked about this scheme with his employees, even discussing it over the phone during his July 2007 honeymoon on an African safari.

Sentinel’s customers were unaware that their funds were being used in this manner. Sentinel sent its customers daily email statements listing the securities in which the customers had pro rata shares. These statements indicated that client funds were segregated; they did not indicate that the funds had been moved into lienable accounts to serve as collateral for a loan that funded trading for Sentinel’s house account. Taken together, this evidence was more than sufficient for the jury to conclude that Bloom intentionally defrauded his customers. 2.

Manipulation of Customer Yield Rates The government also offered sufficient evidence that Bloom intentionally defrauded customers by manipulating yield rates. Several witnesses testified about Bloom’s involvement with this practice. Them testimony was corroborated by documents and recorded phone calls. Sentinel used the yields from the house account and the Prime Portfolio to inflate artificially the returns from the 125 Portfolio. Sentinel’s high-risk trading in the house account generated higher returns than the more conservative 125 Portfolio.

The Prime Portfolio, which was slightly riskier than the 125 Portfolio, likewise generated higher returns. Sentinel redistributed some of these returns from the house account and the Prime Portfolio to the 125 Portfolio, effectively using the riskier accounts to subsidize the more conservative account. With these inflated rates of return in the 125 Portfolio, Sentinel could attract new clients by outperforming its competition. Indeed, it advertised these rates in its marketing material and on its website.

Sentinel employees testified that they doctored the yield rates on a daily basis from 2004 until the company’s bankruptcy in 2007. Instead of paying customers the interest they actually earned, Sentinel employees divvied up interest payments according to the instruction of Bloom or Charles Mosley. Bloom in fact created a spreadsheet to help employees calculate how to redistribute funds, which was called the “Daily Yield/Rate Calculation.” The spreadsheet listed both the actual interest earned by customers’ securities and the rate set by Sentinel.

The rate setting often favored the customers in “Seg 1” (125 Portfolio). For example, on December 7, 2006, the interest actually earned by Seg 1 was $96,942.63. After the rate manipulation by Sentinel, that portfolio was allocated $103,832.46. On that same day, Seg 3 (Prime Portfolio) actually earned $148,005.09 in interest and the house account earned $17,949.85. After Sentinel’s rate adjustment, Seg 3 customers were paid just $112,657.32 and the house account received $49.11. Although co-defendant Mosley oversaw the daily rate setting, Bloom was aware of and involved in the practice.

He created the spreadsheet that employees used, and he received a daily email with the rate calculation spreadsheet for that day. Employees also sometimes consulted Bloom himself with specific rate-setting questions. For instance, on July 30, 2007, Bloom spoke with an employee who was setting the rates and agreed to inflate the 125 Portfolio to keep it competitive. At the end of that day, Seg 1 actually earned $63,477.53, but was paid $100,420.34 using the interest earned by customers in Seg 3 and by the house account.

Another employee testified that he raised the matter with Bloom before May 15, 2007, and Bloom acknowledged that Seg 3 and the house account supplemented Seg 1. Sentinel’s customers were unaware of this practice because Sentinel sent daily account statements that indicated a traditional form of rate calculation. These statements said they were “Customer Segregated Funds,” and each contained an “Asset List” of the securities in which the customer owned a partial share.

The statements then provided the clients’ “Net Equity,” “Net Interest” earned from the security, and the “Net Interest Rate.” The customers who received these statements naturally believed that they conveyed the amount of interest the customers had earned from them securities; they had no reason to believe that them interest was used to subsidize other customers in other portfolios. This evidence was sufficient for the jury to conclude that Bloom intentionally defrauded his clients by manipulating their yield rates. 3.

Soliciting and Accepting Neiv Funds When Sentinel’s Collapse Was Imminent Third, the government offered sufficient evidence for the jury to conclude that Bloom defrauded customers by soliciting and accepting funds even after he knew that Sentinel was about to collapse. Sentinel was faltering in the summer of 2007, and it was clearly headed for bankruptcy by the beginning of August. Sentinel’s loan balance with the Bank of New York was significantly over its $300 million limit, peaking at over $500 million at the end of June.

The bank even temporarily refused to send transfers to Sentinel customers because of the outsized loan. In addition, one of Sentinel’s major clients had $150 million of its assets frozen at the request of the CFTC, further crippling Sentinel. On top of that, Sentinel’s repurchase-agreement counterparties were pushing back'their securities to Sentinel while Sentinel clients were demanding redemptions at an unsustainable rate. On Friday, August 3, 2007, Sentinel consulted with its outside lawyer.

He reviewed Sentinel’s financial records and on Saturday, August 4 advised Bloom to “hold off taking additional deposits until we get some of these issues worked out, even from existing customers.” Bloom did not hold off. The following Monday, Sentinel accepted approximately $110 million in customer funds. On Wednesday, August 8, Sentinel’s outside lawyer said he would withdraw from representation unless Sentinel ceased operations and notified the CFTC. Again, Sentinel refused, and the attorney withdrew on August 9.

Sentinel continued accepting tens of millions of dollars from customers from Monday, August 6 through Friday, August 10 before halting all redemptions on Monday, August 13. The jury also heard evidence that Bloom and his family made unusual personal withdrawals from Sentinel in its last days. Bloom’s father Philip, who was the board chairman; suddenly withdrew all of his funds from the house account on July 18, 2007. This $11 million withdrawal constituted the majority of funds in the house account. (The record contains no explanation for this withdrawal.

On appeal Bloom indicates that it was because his father was diagnosed' with prostate cancer in June 2007.) Defendant Eric Bloom likewise made unusual withdrawals right before the company collapsed. In June Bloom paid himself a premature $50,000 bonus (bonuses were usually paid at the end of the year), and in late July he gave himself a $200,000 retroactive raise. Bloom argued that the money was intended as a down payment for a home he planned to build with his new wife.

The weighing of such evidence—and deciding whether it showed Bloom’s knowledge of the company’s impending collapse—is the province of the jury, who were apparently not persuaded by Bloom’s explanations. This evidence was sufficient for the jury to conclude that Bloom intentionally defrauded customers by taking new funds when he knew Sentinel was about to fold. We emphasize that Sentinel was not merely experiencing financial distress during its last week in business (August 6-10).

The company had undergone financial difficulty throughout the summer of 2007, but Bloom was not charged with fraud under this theory for the funds that Sentinel received during that time. He was charged with fraud only for the funds Sentinel received during its last business week, after the company’s outside lawyer had advised Bloom to stop accepting funds, and the writing of Sentinel’s bankruptcy was on the wall. Reviewing this evidence in the light most favorable to the -government, the government provided sufficient evidence for the jury to find Bloom guilty under each of its theories. B.

Claims of Prosecutorial Misconduct Bloom argues next that the district court erred by refusing to grant a new trial based on prosecutorial misconduct. He claims that the government misled the jury about the evidence and unfairly attacked the defense for wasting time. We find no prosecutorial misconduct. We review for abuse of discretion a district court’s denial of a new trial based on alleged prosecutorial misconduct. United States v. Xiong, 262 F.3d 672, 674-75 (7th Cir. 2001). This requires two steps. First, we determine whether the prosecutor’s conduct was improper in itself. United States v.

Durham, 766 F.3d 672, 684 (7th Cir. 2014). If it was, we ask whether, in light of the whole record, the improper conduct denied the defendant the right to a fair trial. Id. As we have noted, this standard is “difficult to satisfy,” United States v. Bell, 624 F.3d 803, 811 (7th Cir. 2010), but it is not impossible. See, e.g., United States v.

Farinella, 558 F.3d 695, 701-02 (7th Cir. 2009) (reversing for insufficient evidence and directing acquittal, but noting that “had the government presented enough evidence to sustain a conviction, we would have reversed the judgment and ordered a new trial on the basis of the prosecutor’s misconduct”). Bloom first claims that the government “misled the jury” regarding the evidence. This argument duplicates Bloom’s earlier challenges to the probative value and sufficiency of the evidence. Bloom argues that the government misconstrued the innocent nature of his phone calls on July 6, 17, 30, and August 19, 2007.

He claims the government committed misconduct when it argued that the “Daily Yield/Rate Calculation” spreadsheet that Bloom helped to create was designed to cheat Sentinel customers. And he argues that the government misrepresented the evidence when it claimed that he approved of the use of customer funds as collateral for Sentinel’s loan with the Bank of New York. None of these accusations has merit. The government was permitted to argue reasonable inferences about the evidence, United States v. Klebig, 600 F.3d 700, 718 (7th Cir. 2009), and it did so here.

In addition, the district court reminded the jury at several points that the arguments made by the lawyers are not evidence and that jurors should rely on their own understanding of the facts. See United States v. Mallett, 496 F.3d 798, 802 (7th Cir. 2007) (court assumes that jury follows instructions absent evidence to the contrary). Bloom also argues that the government improperly attacked defense counsel during trial in two incidents.

First, the government objected that defense counsel’s cross-examination was repetitious and “practically bordering on a waste of the jury’s time.” The government later said that defense counsel was putting up a “smoke screen.” The court overruled the defense’s objection. We agree with the district court: the government did not act improperly. It criticized defense counsel’s tactics, not defense counsel personally. This is permissible. See United States v.

Washington, 417 F.3d 780, 787 (7th Cir. 2005) (finding no prosecutorial misconduct, in part, because the prosecutor’s arguments “focused on the lameness of the defense rather than defense counsel personally”). C. Jury Instruction on Rule 1.25 Bloom argues next that the district court erred when it rejected his proposed jury instruction that the CFTC’s Rule 1.25 “permits the use of leveraged trading by FCMs using funds covered by Rule 1.25” or, at minimum, does not prohibit the use of leveraged trading.

This understanding of the rule was based on the testimony of Paul Bjarnason, a former CFTC official who participated in drafting Rule 1.25 and who testified as a defense expert here. Bjarnason was also a chief accountant at Sentinel until 2001 and served as a consultant to Sentinel afterwards. The district court rejected Bloom’s proposed instruction and instead gave the jury the full text of the rule. Bloom claims that whether Rule 1.25 allows leverage was a “purely legal question” and that the court’s failure to instruct the jury properly requires a new trial.

We review de novo a district court’s refusal to provide a requested jury instruction “ *when the underlying assignment of error implicates a question of law,’ but ‘general attacks on the jury instructions are reviewed for an abuse of discretion.’ ” United States v. DiSantis, 565 F.3d 354, 359 (7th Cir. 2009), quoting United States v. Macedo, 406 F.3d 778, 787 (7th Cir. 2005).

A defendant is entitled to a jury instruction if “(1) the instruction provides a correct statement of the law; (2) the theory of defense is supported by the evidence; (3) the theory of the defense is not part of the government’s charge; and (4) the failure to include the instruction would deprive the defendant of a fair trial.” United States v. Kokenis, 662 F.3d 919, 929 (7th Cir. 2011), quoting United States v. Campos, 541 F.3d 735, 744 (7th Cir. 2008). The district court’s decision on the jury instruction for Rule 1.25 was not a reversible error, but it does give us pause.

Bloom is correct that whether Rule 1.25 permitted leverage is a question of law, and in general it is the duty of the trial judge to instruct the jury on the relevant law. See United States v. Sanchez, 604 F.3d 356, 360 (7th Cir. 2010) (“It is the basic premise of our legal system that juries are the triers of fact only; it is for the judge, not the jury, to interpret the law.”), quoting United States v. White, 472 F.3d 458, 463 (7th Cir. 2006). The government even objected to Bjarnason’s testimony at trial because the

interpretation of Rule 1.25 was “within the exclusive province” of the court. We have said that the “meaning of statutes, regulations, and contract terms is ‘a subject for the court, not for testimonial experts. The only legal expert in a federal courtroom is the judge.’ ” United States v. Lupton, 620 F.3d 790, 799-800 (7th Cir. 2010), quoting United States v. Caputo, 517 F.3d 935, 942 (7th Cir. 2008). Despite these concerns, we conclude that the district court’s ’ruling on the requested instruction was not a reversible error because Rule 1.25 was only a minor issue in the case against Bloom.

See, e.g., United States v. Choiniere, 517 F.3d 967, 972 (7th Cir. 2008) (finding no reversible error where trial court failed 'to give a requested jury instruction on a “minor issue” that was not central to the charges). Bloom was indicted and convicted not for violating Rule 1.25 but for wire fraud and investment adviser fraud. See id. at 972-73 (upholding denial of jury instruction because it “does not reflect a theory that has any evidentiary support because it is entirely non-responsive to the conduct that was alleged in 'the indictment”), quoting White, 472 F.3d at 462.

Bloom’s conviction did not depend on the intricacies of Rule 1.25. Instead, it required the government to prove beyond a reasonable doubt that hé defrauded his customers. The government had to prove that Bloom’s business practices materially deviated from the promises he made to his customers. The government made this showing in several ways, including for instance, that Bloom failed to segregate customer funds despite telling customers otherwise. There was ample evidence for the jury to convict Bloom without considering the nuances of Rule 1.25.

He was not deprived of a fair trial because the district court rejected his proposed instruction. Even assuming the correctness of Bjar-nason’s

interpretation of Rule 1.25, this still would not have justified Bloom’s proposed jury instruction. Bjarnason testified that, while Rule 1.25 permitted leverage of the customer accounts, the leverage must be for the benefit of those customers: “Leverage done, for example, in the SEG 1 account should be for the benefit of the SEG 1 customers, the investors.” The evidence showed, however, that Sentinel leveraged customer funds for the benefit of its own house account. Bloom’s proposed instruction would have been misleading even under his own expert’s account. D.

Evidentiary Rulings Bloom claims that the district court erred when it admitted certain out-of-court statements by Sentinel employees and Philip Bloom. Specifically, he argues that the court applied Federal Rule of Evidence 801(d)(2)(

D) incorrectly by admitting statements of Sentinel employees, which were then imputed to Bloom. Bloom claims that he did not have the requisite intent to defraud and that without these statements the government had no “direct evidence” of his involvement in any fraudulent activity. Bloom focuses on three groups of out-of-court statements, which include emails from the chief financial officer to Philip Bloom, emails from an office manager, and one email from an assistant trader. He also contends the government unfairly blamed him for his father’s sudden $11 million withdrawal from the house account several weeks before Sentinel collapsed. We review de novo a district court’s legal

interpretation of the Federal Rules of Evidence, but we review for abuse of discretion the court’s ultimate decision to admit or exclude evidence. United States v. Loughry, 660 F.3d 965, 969 (7th Cir. 2011), citing United States v. Rogers, 587 F.3d 816, 819 (7th Cir. 2009). Federal Rule of Evidence 801(d)(2)(

D) provides that a statement is not hearsay if it is “offered against an opposing party” and “was made by the party’s agent or employee on a matter within the scope of that relationship and while it existed.” We have acknowledged that the law regarding this rule is “somewhat muddled.” Stephens v. Erickson, 569 F.3d 779, 793 (7th Cir. 2009), quoting Aliotta v. National Railroad Passenger Corp., 315 F.3d 756, 761 (7th Cir. 2003). We need not chart these waters, however, because even if there were any errors admitting these statements under Rule 801(d)(2)(D), they were harmless.

The emails of the office manager and the chief financial officer were also admitted as co-conspirator statements under Rule 801(d)(2)(E), and Bloom does not challenge those rulings in his appeal. The statement by the assistant trader was a two-sentence email with portfolio documents attached. At trial, Bloom’s counsel expressly agreed to admission of the portfolio documents under the business record foundation. See Fed. R. Evid. 803(6)(B). The two sentences of the email were not a significant part of the four-week trial, and their admission did not affect Bloom’s substantial rights.

Any error would have been harmless. Fed. R. Crim. P. 52(a). Further, the government’s statements regarding Philip Bloom’s unusual withdrawal were appropriate to show Bloom’s intent, as discussed above in

Part II-A-3. E. Sentencing Finally, Bloom challenges his sentence. In computing a sentencing guideline range for Bloom, the district court used a total loss amount of $666 million under U.S.S.G. § 2B1.1. That enormous loss produced a guideline range of life in prison. The sentencing judge, however, imposed a 168-month prison sentence, saying that his decision was driven primarily by the sentencing factors under 18 U.S.C. § 3553(a).

Bloom claims that the district court erred in its loss calculation by failing to consider other causes for the losses, such as the collapse of the securities markets in 2007, the wrongdoing of Sentinel’s head trader Charles Mosley, and the CFTC’s former, less stringent rules for investing in securities. Bloom also claims that the loss calculation was too high because the court incorrectly assumed that Sentinel clients were fraudulently induced to invest in the company.

Rather, Bloom argues, his case did not involve fraudulent inducement because the government did not provide evidence that Sentinel clients would not have entrusted their money to him had they known about the fraud. We review de novo the district court’s

interpretation of the Sentencing Guidelines, and we review its factual findings for clear error. United States v. Walsh, 723 F.3d 802, 807 (7th Cir. 2013). The Supreme Court’s decision in United States v. Booker, 543 U.S. 220, 125 S.Ct. 738, 160 L.Ed.2d 621 (2005), held that the federal Sentencing Guidelines are advisory. Since Booker, sentencing judges have had broad discretion to impose non-guideline sentences by weighing the factors under § 3553(a). In United States v.

Lopez we explained, as we often have: In a case ... presenting a rather technical and arcane question in applying the Sentencing Guidelines, it is perhaps worth another reminder that the Guidelines are, after all, guidelines. They must be considered seriously and applied carefully. In the end, however, the defendant’s sentence is the responsibility of the district judge, after careful consideration of all the relevant factors under 18 U.S.C. § 3553(a)— A district court facing a tricky but technical issue under the Guidelines may exercise its discretion under

section 3553(

a) and may spell out on the record whether and to what extent the resolution of the guideline issue affected the court’s final decision on the sentence. 634 F.3d 948, 953-54 (7th Cir. 2011) (citations omitted); see also United States v. Harris, 718 F.3d 698, 703 n.2 (7th Cir. 2013) (“Sentencing Guidelines are advisory, not mandatory, and ... district judges are free to deal with such abstract and artificial [guideline] issues by telling the parties and reviewing courts that the decision on the final sentence did not depend on their resolution.”); United States v. Sanner, 565 F.3d 400, 406 (7th Cir. 2009) (“When a judge proceeds in this manner, she must make clear that the § 3553(

a) factors drive the sentence without regard as to how the prior conviction fits under a particular guideline. Doing so will make the often nit-picking review of issues like this under our now advisory guideline scheme unnecessary.”). Here, the sentencing judge faced a challenging technical issue under the Guidelines and followed this advice. He reviewed all of the sentencing materials and carefully considered the Guidelines.

The judge then acknowledged the controversy surrounding the loss calculation but said that he would impose the same sentence regardless of the guideline range: “this is a sentence the Court would impose regardless of what the ultimate proper sentencing guideline calculation is determined to be as the sentence is driven by the statutory sentencing factors, specifically the need to ... deter future conduct and to protect the public.” The court appropriately exercised its discretion under § 3553(a). The guideline calculation did not produce any reversible error. Bloom’s convictions and sentence are AFFIRMED. .

An additional charge of making false statements to an employee benefit plan in violation of 18 U.S.C. §§ 2 and 1027 was dropped during trial.