DUNCAN, R J.
In this criminal case, defendant appeals a judgment convicting him of two counts of first-degree theft, ORS 164.055, and two counts of second-degree theft, ORS 164.045, for checking out 50 books from libraries in Clackamas County and failing to return them. Defendant raises two assignments of error. In his first assignment of error, defendant contends that the trial court erred by admitting evidence regarding a previous conviction for theft for taking books from a book sale run by the Friends of the Tigard Library. We agree that the trial court erred in admitting the evidence regarding defendant’s book-sale theft, but we conclude that the error was harmless. In his second assignment of error, defendant contends that the trial court plainly erred in imposing $510 in court-appointed attorney fees in the absence of evidence that defendant was or might be able to pay them. We agree that the trial court plainly erred, and we exercise our discretion to correct the error. Accordingly, we reverse the portion of the judgment requiring defendant to pay attorney fees and otherwise affirm.
Defendant was charged with four counts of theft for checking out a total of 50 books — the maximum number that one patron may check out from the Clackamas County library system — from four Clackamas County libraries and failing to return them.1 Police recovered one of the books from Powell’s Books; Powell’s records showed that defendant had sold that book to Powell’s three days after he checked it out from the library. The book had been altered to remove its bar code and to obscure markings that identified it as a library book. Defendant admitted that he had sold perhaps nine more of the library books as part of a group of 500 books that he had arranged to sell through Craigslist. In his statements to the police, defendant contended that he had sold those books — the one sold to Powell’s and the others sold through Craigslist — accidentally, when they became intermingled with other books that he intended to sell. Two of the 50 books were eventually found with their bar codes removed, on the shelf in one of the libraries. The police [225]*225searched defendant’s house approximately three months after the books became overdue, but they did not find any of the missing library books there.
Before trial, the state moved to admit evidence of two prior thefts by defendant, including the fact that defendant had been convicted of those thefts after pleading guilty. First, the state sought to introduce evidence regarding defendant’s theft of 13 books from the Multnomah County Library, a theft accomplished by checking out the books from various library branches and failing to return them. Defendant sold three of those books to McKenzie Books on the same day that he checked them out from the library. Those books were altered similarly to the book recovered from Powell’s.
When questioned by the police about the Multnomah County Library books, defendant asserted that, after he checked the books out, he had loaned them to another person who was staying in the same house as he was, and when the books were returned to defendant, they had been altered. Defendant said that the person who had borrowed the books offered to sell them for defendant and then did so. Defendant pleaded guilty to the theft from the Multnomah County Library and was convicted.
Second, the state sought to introduce evidence regarding defendant’s theft of books from the Friends of the Tigard Library book sale. In that case, defendant did not check out books from the library. Instead, he went to the book sale, took books, and left without paying for them. When questioned by the police, defendant said that he had taken some books to the book sale and decided not to leave all of them there. In the process of taking some of the books back out, he may have accidentally taken a book that did not belong to him. Defendant also told police that he buys and sells books and that he uses an application on his smart phone to scan the ISBN on books to determine their value and learn which stores will buy the books. Defendant pleaded guilty to the Friends of the Tigard Library theft and was convicted.
After the state presented its evidence about the charged Clackamas County thefts to the jury, the court held [226]*226a hearing outside the presence of the jury to decide whether to admit evidence of the prior thefts. The state reasserted the arguments raised in its pretrial memorandum, namely, that “[s]o called prior bad act evidence is admissible in any case when it is offered to prove anything other than criminal propensity,” and that, in this case, evidence of the Multnomah County Library and Tigard book-sale thefts was “relevant to preparation, knowledge, lack of mistake, and intent,” all of which are nonpropensity purposes listed in OEC 404(3), which is set out below. 279 Or App at 227. As to both prior thefts, the state asserted that the evidence “would explain to the trier of fact how [defendant’s] actions were not a mistake or accident and his reason for doing this.” The state also argued that the probative value of the evidence outweighed any unfairly prejudicial effect.
In response, defendant distinguished the Tigard book-sale theft from the Multnomah County theft. He argued, inter alia, that the act of theft involved in the Tigard book-sale case was dissimilar from the charged acts because defendant stole the Tigard books by taking them from the book sale without buying them, whereas in the present case — as in the Multnomah County case — he checked out the books in accordance with the terms of his agreement with the libraries and committed a crime only if he intended to deprive the libraries of them. (It was undisputed that he did, in fact, deprive the libraries of them by failing to return them.) As a result, defendant contended, the evidence of the Tigard book-sale theft was “not particularly probative” of any fact at issue in this case. The state replied that, because defendant admitted that he checked out the library books at issue in this case, the real question for the jury was whether defendant intended to deprive the library of the books and that the evidence of both prior thefts would help the jury answer that question.
The court admitted the evidence of both prior thefts. It reasoned that the evidence went to “mistake or accident,” which it characterized as “really the only [way for defendant] to avoid culpability” in the case. The court also stated that the evidence was admissible to show defendant’s plan. Then it noted that
[227]*227“there is a lot of similarity [between the two prior thefts and the present charges], and the only dissimilarity * * * is that the incident in Tigard was an incident where the books * * * were taken out of the library without permission.
“In the case before the Court, the books were taken out of the library with permission, as they were in Multnomah County. But there are things that I think are sufficiently similar about the instances that make this evidence very probative as to whether there was a scheme or plan to accomplish this result. As with the case that’s before the Court, in the prior instances, there was an identification of the higher value of the books — an effort to identify the books with a higher value, at least arguably.
“Multiple locations were involved in the prior instances. The tags were removed, of course, the ink labels were blacked out. By the tags, I mean the bar codes, the ink labels were blacked out. [A Multnomah County Library employee] said the [inventory] tags were stripped from the books so the alarms wouldn’t go off. The books were introduced into a distribution channel. McKenzie Books in the one instance; Powell’s Books in the other. These are well-known distribution channels for used books — a lot of books. Cash was obtained for the books.”
The court balanced the probative value of the evidence against the risk of unfair prejudice and concluded that it should be admitted.
As noted above, on appeal, defendant contends that the trial court erred in admitting the evidence of the theft from the Tigard book sale; he does not challenge the court’s admission of evidence of the theft from the Multnomah County Library. In his opening brief, defendant asserts that the admission of the challenged evidence is controlled by OEC 404(3) and State v. Johns, 301 Or 535, 725 P2d 312 (1986). OEC 404(3) provides:
“Evidence of other crimes, wrongs or acts is not admissible to prove the character of a person in order to show that the person acted in conformity therewith. It may, however, be admissible for other purposes, such as proof of motive, opportunity, intent, preparation, plan, knowledge, identity, or absence of mistake or accident.”
[228]*228In Johns, the Supreme Court explained that evidence of prior bad acts may be admissible to show absence of mistake or accident if it is relevant under “‘the doctrine of chances — the instinctive recognition of that logical process which eliminates the element of innocent intent by multiplying instances of the same result until it is perceived that this element cannot explain them all.’” 301 Or at 553 (quoting 2 John Henry Wigmore, Evidence § 302,246 (Chadbourne Rev 1979)). Under the doctrine, “the proponent uses uncharged misconduct evidence inductively and probabilistically.” Id. at 552. The doctrine is based on the view that
‘“recurrence or repetition of the act increases the likelihood of a mens rea or mind at fault. In isolation, it might be plausible that the defendant acted accidentally or innocently; a single act could easily be explained on that basis. However, in the context of other misdeeds, the defendant’s act takes on an entirely different light. The fortuitous coincidence becomes too abnormal, bizarre, implausible, unusual or objectively improbable to be believed. The coincidence becomes telling evidence of mens rea!"
Id. at 552-53 (quoting Edward Imwinkelried, Uncharged Misconduct Evidence § 5:05, 10-11 (1984)). In other words, “multiple instances of similar conduct are unlikely to occur accidentally.” State v. Leistiko, 352 Or 172, 182, 282 P3d 857, adh’d to as modified, 352 Or 622, 292 P3d 522 (2012). To decide whether a prior instance of conduct is similar enough to the charged conduct to be admissible to show an absence of mistake or accident,
“the trial judge should make these determinations:
“(1) Does the present charged act require proof of intent?
“(2) Did the prior act require intent?
“(3) Was the victim in the prior act the same victim or in the same class as the victim in the present case?
“(4) Was the type of prior act the same or similar to the acts involved in the charged crime?
“(5) Were the physical elements of the prior act and the present act similar?
[229]*229“(6) If these criteria are met, is the probative value of the prior act evidence substantially outweighed by the danger of unfair prejudice, confusion of issues or misleading the jury, undue delay or presentation of cumulative evidence?”
Johns, 301 Or at 555-56.
Defendant contends that the Tigard book-sale theft did not meet the fifth Johns requirement. He points out that the Supreme Court has explained that that inquiry requires the court to consider both the similarities and the differences between the physical elements of the two crimes:
“The circumstances of each crime as a whole must be compared. First, the trial judge must find that there are significant similarities in the physical elements of the two crimes. If that test is met, then the trial judge must consider the differences between the physical elements of the two crimes. The differences may be minimal — for example, the offender may have used different words to indicate his intent. On the other hand, the differences may be so great that they overwhelm the similarities. The point is: The dissimilarities must be as fully considered as the similarities in answering this question.
“Determining what constitutes a significant similarity is a matter to be decided on a case-by-case basis. Some similarities are so common as to be trivial (for example, the offender spoke English during both crimes) while others may be so unusual as to be significant even standing alone (for example, the offender spoke a foreign language when he intended to rape, but spoke English otherwise). Most often the significance of the similarities will arise out of their combination.”
State v. Pratt, 309 Or 205, 214, 785 P2d 350 (1990). And, as defendant notes, we have explained that “the acts must bear something close to a point-by-point correspondence.” State v. Deloretto, 221 Or App 309, 317, 189 P3d 1243 (2008), rev den, 346 Or 66 (2009).
To the extent that the trial court also admitted the Tigard book-sale theft evidence as evidence of a plan, defendant points out that the Supreme Court has explained that “something more than the similarity required for other [230]*230crimes evidence to be admissible to prove intent is necessary for it to be admissible to prove a plan,” Leistiko, 352 Or at 189, and, accordingly, asserts that the court also erred in admitting the evidence on that basis. In response, the state argues that, although the question of whether the evidence was similar enough to satisfy the fifth Johns requirement “may be a close one,” ultimately, it was similar enough, and, regardless, any error was harmless.2
After this case was submitted, the Supreme Court decided State v. Williams, 357 Or 1, 15, 346 P3d 455 (2015), in which it held that OEC 404(4)3 supersedes OEC 404(3) in criminal cases, “except, of course, as otherwise provided by the state or federal constitutions.” The court decided that evidence that a defendant has a propensity to sexually abuse children is logically relevant under OEC 4014 and ultimately concluded that, “in child sexual abuse prosecutions where the state offered prior bad acts evidence to prove that the defendant had a propensity to sexually abuse children, due process ‘at least requires that, on request, trial courts determine whether the probative value of the evidence is outweighed by the risk of unfair prejudice.’” State v. Turnidge (S059155), 359 Or 364, 431, 374 P3d 853 (2016) CTurnidge) (quoting Williams, 357 Or at 19).
In a memorandum of additional authorities, defendant asserts that, in light of Williams, the issue in this case is now “whether the trial court properly exercised its [231]*231discretion under OEC 403.”5 Defendant contends that it did not because we have held that “[a] trial court abuses its discretion in admitting prior bad acts evidence under OEC 403 if the court erroneously determines that the evidence meets the Johns test.” (Citing State v. Baughman, 276 Or App 754, 769-71, 369 P3d 423 (2016) (considering Johns test in evaluating whether trial court properly conducted OEC 403 balancing in a child sexual abuse case).) The state has not responded to defendant’s arguments in his memorandum of additional authorities.
In Turnidge, 359 Or 364, the Supreme Court recently clarified the scope of its holding in Williams, and that clarification causes us to frame the issue somewhat differently than defendant does in his memorandum. In Turnidge, the defendant was convicted of aggravated murder and other crimes, including bank robbery, for planting a bomb outside a bank and calling in a bomb threat to a nearby bank in 2008. The bomb exploded, killing two people and injuring others. 359 Or at 366. The defendant objected to, among other things, the state’s evidence that, in 1995, he had called in a bomb threat to another bank in the same vicinity, and then watched the police respond to the call from a restaurant near the bank. Id. at 426.
Before the trial court, the state argued that, under OEC 404(3) and Johns, the evidence of the 1995 bomb threat was admissible to prove motive, ability, planning, preparation, intent, and knowledge. Id. at 427-28. On review, the state renewed those arguments and also contended, in light of Williams, that “OEC 404(4) preempts the limitations that OEC 404(3) otherwise places on the admission of evidence of ‘other crimes, wrongs, or acts,’ and that such evidence is always admissible under OEC 404(4) if it is relevant — even for a propensity purpose — as long as its admission does not violate due process.” Id. at 428-29.
To evaluate those arguments, the court began by explaining its holding in Williams. In doing so, it noted that, [232]*232in Williams, it had “reserved” the question of “the extent to which prior bad acts evidence can be admitted solely for propensity purposes in criminal cases other than ones involving child sexual abuse.” Turnidge, 359 Or at 432. Then it noted that it still did not need to resolve that reserved question because, in Turnidge, the evidence at issue was not offered to prove that the defendant had a propensity to engage in certain behavior:
“The state’s theory of admissibility [at trial] was not that the 1995 bomb threat evidence showed that defendant had the character trait of being a bank robber or bomber, and that the jury therefore should infer from his propensity to rob or bomb banks that he bombed West Coast Bank in 2008. *** Succinctly stated, the state’s theory of relevance included that the 1995 incident was part of defendant’s planning process for committing several of the charged crimes. That theory falls squarely within the nonexclusive list of nonpropensity purposes for which prior bad acts evidence historically has been admissible, which are largely codified in OEC 404(3).”
Id. at 432-33 (emphasis in original; footnote omitted). Accordingly, the court declined to address the reserved question of whether “prior bad acts evidence can be admitted solely for propensity purposes in criminal cases other than ones involving child sexual abuse.” Id. at 432. Instead, the court concluded that, “under settled cases interpreting OEC 404(3),” the 1995 bomb-threat evidence was admissible as evidence of plan; accordingly, the court did not need to address “the potential application of OEC 404(4) here.” Id. at 433 n 37.
In this case, as in Turnidge, the state’s theory of admissibility of the Tigard book-sale theft evidence at trial was not based on a propensity theory — that is, the state did not argue, for example, that defendant had the character trait of being a book thief and that the jury should therefore infer from his propensity to steal books that his failure to return his checked out library books was theft. Instead, the state argued — and the trial court agreed — that the Tigard book-sale theft evidence showed that defendant’s failure to return the Clackamas County library books was [233]*233not accidental, contrary to defendant’s assertion to the police. Moreover, here, unlike in Turnidge, the state does not raise any propensity theory on appeal. Given the lack of any propensity argument before the trial court or on appeal, we confine our analysis to the questions presented in defendant’s opening brief: whether the court erred under OEC 404(3) in admitting the Tigard book-sale theft evidence and whether any error harmed defendant.6 See State v. Hudman, 279 Or App 180, 187, 379 P3d 659 (2016) (in a case decided after Turnidge, noting that, where “the state has not asked us to affirm defendant’s conviction on the alternative basis that, even if the other-act evidence was improperly admitted under OEC 404(3), the evidence was nonetheless admissible under OEC 404(4),” “our task on appeal is limited to determining whether the trial court erred in admitting [the disputed evidence] under OEC 404(3)”). As noted above, we conclude that the court erred, but that the error was harmless.
The court erred in admitting the evidence of the Tigard book-sale theft under OEC 404(3) because it was not relevant to show intent — that is, mistake or lack of accident — based on the doctrine of chances. As a general matter, the circumstances of this case did lend themselves to that type of evidence, because defendant admitted he had done the actus reus — he checked out the library books and did not return them — and, as the court noted, the real dispute went to whether defendant intended not to return the books. Defendant’s assertion to the police that, although he had sold some of the books (making it impossible to return them), he had done so accidentally, put defendant’s intent at issue. Given those circumstances, evidence that defendant had previously committed the same actus reus would make it more likely that defendant’s failure to return the books to the Clackamas County libraries was not a mistake.
[234]*234However, as defendant argues, the evidence of the Tigard book-sale theft was too dissimilar from the conduct at issue to be probative of defendant’s lack of mistake or accident in this case. The actus reus was different — in the Tigard theft, defendant took books from the book sale without paying for them, rather than taking them with permission and then failing to return them. The doctrine of chances rests on the principle that “multiple instances of similar conduct are unlikely to occur accidentally.” Leistiko, 352 Or at 182. However, the fact that bad conduct, as a general category, has occurred more than once does not allow any inference about the likelihood that the charged conduct happened by accident. See Deloretto, 221 Or App at 317 (noting that, for prior acts to be probative of intent in performing charged acts, “we have consistently held that the acts must bear something close to a point-by-point correspondence”; rejecting state’s argument that “a verbal incident is sufficiently similar to actual touching” to be probative of lack of mistake in the touching (emphasis in original)). Even where the prior act and the charged act involve similar kinds of bad conduct — here, both involved theft of books from library-related organizations — the similarities between the physical elements must outweigh the differences. See Pratt, 309 Or at 214 (where one victim was abducted, raped at a motel, and, apart from the rape, not seriously injured and the second was not abducted, was raped, if at all, in a truck or by the side of the road, and then killed, differences between the two rapes were so significant that, “[a]side from its tendency to show that defendant is the sort of man who commits rape, the [first] abduction and rape is not probative of defendant’s intent to rape [the second victim]”); see also, e.g., Hudman, 279 Or App at 188-89 (defendant’s joint marijuana-growing operation and plans to sell marijuana too dissimilar from defendant’s later theft of marijuana to sell it to meet Johns test).
In this case, the physical elements of the charged thefts were unusual in that defendant did not steal the books merely by physically removing them from the library; instead, the thefts would be committed when, having obtained the books, he had a concurrent intent not to return them. That may have taken place when he removed them [235]*235from the library with permission or when he later decided not to return them and then withheld them from the library.7 That defendant had stolen books in the past by taking them from a book sale without permission is not similar enough to support an inference that defendant did not accidentally fail to return the library books that he took with permission. And defendant correctly points out that the specific similarities on which the trial court relied — the fact that defendant targeted higher-value books, multiple locations were involved, the books were altered, and the books were sold— are all facts that relate only to the Multnomah County evidence, not to the Tigard book-sale evidence. Accordingly, the Tigard book-sale theft evidence was not admissible under OEC 404(3) to prove defendant’s lack of mistake in failing to return the books. It did not “bear something close to a point-by-point correspondence” to the charged conduct and did not meet the “stringent test for similarity.” Deloretto, 221 Or App at 317.
That conclusion also dictates that the trial court erred in admitting the Tigard book-sale theft evidence to show plan. As defendant notes, the Supreme Court has explained that, where other-act evidence is “not sufficiently similar * * * to be admissible to prove intent,” “[i]t necessarily follows that [it also is] not sufficiently similar * * * to be admissible to prove a plan” of this type. Leistiko, 352 Or at 189; see also Hudman, 279 Or App at 189-90 (same).8 Thus, the court erred under OEC 404(3) in admitting the Tigard book-sale theft evidence.
[236]*236Nevertheless, we conclude that the court’s erroneous admission of evidence of the Tigard book-sale theft was harmless. See State v. Davis, 336 Or 19, 28, 77 P3d 1111 (2003) (court “must affirm a judgment, despite any error committed at trial, if, after considering all the matters submitted, the court is of the opinion that the judgment ‘was such as should have been rendered in the case’” (quoting Or Const, Art VII (Amended), § 3)). Defendant argues that the erroneously admitted evidence of the Tigard book-sale theft “served as strong evidence of defendant’s intent” and thus went “‘directly to the heart of defendant’s factual theory of the case.’” (Quoting Davis, 336 Or at 34.) He contends that the evidence was not cumulative because, unlike the evidence of the Multnomah County Library theft, the Tigard book-sale theft evidence gave the jury an instance of “literal theft” that would be more persuasive than the Multnomah County Library theft evidence, as to which the jury might believe that defendant had misplaced or otherwise lost the books, rather than stealing them.
We reject defendant’s argument. As the state points out, the Multnomah County Library evidence, which defendant did not challenge, was “powerful evidence rebutting defendant’s claim” that he accidentally sold the books at issue in this case. Given that evidence — which, as described, included evidence that defendant checked out 13 books from various library branches, sold three of the books on the same day he checked them out, and pleaded guilty to theft— there is little likelihood that the Tigard book-sale evidence affected the verdict. All of the inferences supported by the Tigard book-sale evidence were supported more strongly by the Multnomah County Library evidence. Defendant pleaded guilty to theft in the Multnomah County Library case; it was, and the jury would have understood it to have been, just as much of a theft as the theft in the Tigard book-sale case.9 Thus, the trial court’s erroneous admission of the Tigard book-sale theft evidence was harmless.
In a supplemental brief, defendant assigns error to the trial court’s imposition of $510 in court-appointed [237]*237attorney fees. He asserts that, in the absence of evidence regarding his ability to pay the fees, the trial court plainly erred in imposing them. The state concedes that the court erred, but argues that, in light of the “insubstantial” amount of fees imposed, we should not exercise our discretion to correct the error. See State v. Baco, 262 Or App 169, 171, 324 P3d 491, rev den, 355 Or 751 (2014) ($510 in court-appointed attorney fees “is not a substantial amount given that defendant’s probationary sentence does not prevent him from working and that defendant agreed to the state’s recommendation of attorney fees in the same amount for another charge sentenced at the same time”).
We accept the state’s concession that the trial court plainly erred in imposing fees on this record, which is silent on defendant’s ability to pay the fees. State v. Coverstone, 260 Or App 714, 716, 320 P3d 670 (2014); see also ORS 151.505 (court may not order defendant to pay court-appointed attorney fees unless defendant “is or may be able to pay” them); ORS 161.665 (same); Bacote v. Johnson, 333 Or 28, 33-34, 35 P3d 1019 (2001).
The court sentenced defendant to 26 months in prison and ordered him to pay restitution of $3,834.73. The $510 attorney fee award is not insubstantial considered in light of defendant’s 26-month prison sentence, the $3,834.73 restitution award, and the lack of evidence in the record indicating that defendant would be able to pay the fees. See, e.g., State v. Baker, 278 Or App 327, 328, 372 P3d 626 (2016) (exercising discretion to correct erroneously imposed $510 in attorney fees where defendant was sentenced to 24 months in prison); State v. Hunt, 271 Or App 347, 353, 350 P3d 521 (2015) (distinguishing Baco, where the defendant was sentenced to probation, on the ground that defendant was sentenced to 14 months in prison and “[t]he record contains no evidence that he has [a source of income other than working] or that he has or will have the capacity to pay the fees”; exercising discretion to correct erroneously imposed $510 in attorney fees); State v. Brown, 272 Or App 321, 325-26, 355 P3d 129 (2015) (exercising discretion to correct erroneously imposed $600 in attorney fees where defendant was sentenced to 33 months in prison; noting reasoning in Hunt that a prison sentence means “that, for a significant period [238]*238of time, the defendant would likely have no way of earning money to pay the fee”). Accordingly, for the reasons set out in Baker, we exercise our discretion under Ailes v. Portland Meadows, Inc., 312 Or 376, 382, 382 n 6, 823 P2d 956 (1991), to correct the error.
Portion of judgment requiring defendant to pay attorney fees reversed; otherwise affirmed.