Peterson, Justice.
**534Albert and Ashley Debelbot appeal their malice murder convictions for the death of their infant daughter, McKenzy Debelbot.1 The Debelbots each argue that the evidence was insufficient to sustain **535their convictions, that their respective trial attorneys rendered constitutionally ineffective assistance, and that their convictions cannot stand for other reasons. We conclude that the evidence was sufficient to support the convictions of both Albert and Ashley. But while we are deeply troubled by at least two of the claims of ineffective assistance of counsel, the nature of the order below prevents meaningful review of all of those claims. Therefore, we vacate and remand for further proceedings consistent with this opinion.
Viewing the evidence in the light most favorable to the jury verdict, the trial evidence shows that Ashley gave birth to McKenzy on May 29, 2008, at Martin Army Community Hospital in Columbus,2 and they were discharged from the hospital on May 31. There were no signs that the baby was unhealthy or in distress. The Debelbots took McKenzy home and provided the sole care for the infant for the next 13 hours, during which time - according to the Debelbots - they fed and played with McKenzy, changed her diaper, and gave her a bath. In the early morning hours of June 1, the Debelbots took McKenzy to the hospital after noticing a *132bump on her head. McKenzy died a few hours later.
When police arrived, Albert appeared to be very distraught and cried several times, while Ashley did not appear to be nearly as upset as Albert and was never seen crying. During police interviews, however, Ashley got upset in explaining McKenzy's injuries and claimed not to know what happened to McKenzy. The police conducted a search of the Debelbots' apartment, but there is no indication that anything of evidentiary value was recovered.
A GBI medical examiner, Dr. Lora Darrisaw, performed an autopsy on McKenzy on June 2, 2008. Dr. Darrisaw testified that McKenzy had a fracture on the left side of her head, extensive fractures on the right side of her head, and bleeding in the brain. Dr. Darrisaw noted that McKenzy's birth occurred without any apparent complications and that the infant had food in her stomach at the time of the autopsy. Dr. Darrisaw also examined 19 microscopic slides of McKenzy's brain, found that no inflammatory cells had formed in response to the trauma, and concluded that the injuries preceded McKenzy's death by no more than 12 hours, possibly as little as one **536or two hours. Dr. Darrisaw opined that McKenzy's birth was not the cause of the trauma, because the extent of the injuries would not have allowed the infant to eat or do anything else and would have caused a very rapid death. Instead, Dr. Darrisaw concluded that McKenzy's death was a homicide and the cause of death was blunt force trauma, either by a series of blows to the head or by a "crushing type of injury."
Based on Dr. Darrisaw's conclusion, the police arrested the Debelbots, who both denied harming their child. Albert testified at trial, reiterating that neither he nor Ashley ever harmed McKenzy. In rebuttal, the State called Melvin Tarver, a felon with multiple convictions who shared a holding cell with Albert, who testified that Albert had confided in him on the first morning of the Debelbots' trial. According to Tarver, Albert said that, on the night McKenzy was brought home, he left the house to buy drugs and, when he returned, Ashley told him that she had spanked the infant and put her to bed.
Ashley also testified at trial and denied that she or Albert had hurt McKenzy. She admitted that McKenzy appeared to be healthy after birth, and both Albert and Ashley admitted that McKenzy appeared to be fine prior to the time they found a bump on her head and took her to the emergency room. Neither Albert nor Ashley called any medical experts to testify in their defense.
During its closing argument, the State made the following argument about reasonable doubt:
The Judge will charge you on reasonable doubt. Just keep in mind, and he will charge you, reasonable doubt does not mean beyond all doubt. It does not mean to a mathematical certainty. Which means we don't have to prove that ninety percent. You don't have to be ninety percent sure. You don't have to be eighty percent sure. You don't have to be fifty-one percent sure. It does not mean to a mathematical certainty.
And it does not mean beyond a shadow of a doubt. That's just something the TV made up. It's actually beyond a reasonable doubt. And that would be a doubt to which you can attach a reason. And I submit to you there is no reasonable doubt in this case.
(Emphasis added.) Neither of the Debelbots' trial counsel objected to this argument. The trial court later charged the jury on the burden of proof by stating:
The defendants are presumed to be innocent until proven guilty. Each defendant enters upon the trial of the case with a presumption of innocence in his or her favor. This **537presumption remains with the defendant until it is overcome by the State with evidence that is sufficient to convince you beyond a reasonable doubt that the defendant is guilty of the offense charged.
No person shall be convicted of any crime unless and until each element of the crime as charged is proven beyond a reasonable doubt.
The burden of proof rests upon the State to prove every material allegation of the *133indictment and every essential element of the crime charged beyond a reasonable doubt. There is no burden of proof upon the defendant whatsoever, and the burden never shifts to the defendant to produce evidence or to prove innocence.
However, the State is not required to prove the guilt of the accused beyond all doubt or to a mathematical certainty. A reasonable doubt means just what it says. A reasonable doubt is a doubt of a fair-minded, impartial juror honestly seeking the truth. A reasonable doubt is a doubt based upon common sense and reason. It does not mean a vague or arbitrary doubt but is a doubt for which a reason can be given, arising from a consideration of the evidence, a lack of evidence, or a conflict in the evidence.
The jury deliberated and found the Debelbots guilty of malice murder. The Debelbots filed motions for new trial in 2009, raising general grounds and asserting that their respective trial counsel was ineffective for, among other things, failing to call an expert to introduce medical testimony to counter the State's case and that Albert's counsel was ineffective for failing to object to the State's argument about reasonable doubt. Following numerous and lengthy hearings from 2014 to 2017, the motion-for-new-trial court - which had qualified the Debelbots' expert witnesses as experts - issued a short order denying relief and concluding that the Debelbots' proffered expert and non-expert witnesses who testified in support of the motion for new trial were not credible, and that all of the medical evidence presented by the Debelbots' expert witnesses was inadmissible under Harper v. State, 249 Ga. 519, 292 S.E.2d 389 (1982).
1. The Debelbots argue that the evidence was insufficient to convict them because the State failed to present direct evidence that either one of them inflicted or helped the other inflict the alleged injuries to McKenzy. We disagree.
Under both former OCGA § 24-4-6, in effect at the time of [the Debelbots'] trial, and present OCGA § 24-14-6, in **538order to convict [the Debelbots] of the crimes based solely upon circumstantial evidence, the proven facts had to be consistent with the hypothesis of [their] guilt and exclude every reasonable hypothesis save that of [their] guilt. Not every hypothesis is reasonable, and the evidence does not have to exclude every conceivable inference or hypothesis; it need rule out only those that are reasonable. The reasonableness of an alternative hypothesis raised by a defendant is a question principally for the jury, and when the jury is authorized to find that the evidence, though circumstantial, is sufficient to exclude every reasonable hypothesis save that of the accused's guilt, this Court will not disturb that finding unless it is insupportable as a matter of law.
Akhimie v. State, 297 Ga. 801, 804 (1), 777 S.E.2d 683 (2015) (citations omitted). Even if a person is not directly responsible for the crime, he or she may be convicted as a party to the crime if he or she "intentionally aids or abets the commission of the crime, or intentionally advises, encourages, hires, counsels, or procures another to commit the crime[.]" Smith v. State, 277 Ga. 95, 95, 586 S.E.2d 629 (2003) (citing OCGA § 16-2-20 ). "Whether a person is a party to a crime may be inferred from that person's presence, companionship, and conduct before, during, and after the crime," and "where the crimes involve relatives with close relationships, slight circumstances can support the inference that the parties colluded." Virger v. State, Case Nos. S18A1538, 2019 WL 654189 at *6, 2019 Ga. LEXIS 111 at *15-16 (Feb. 18, 2019) (punctuation and citations omitted).
The evidence in this case, although entirely circumstantial as to who committed the crimes, was legally sufficient to support the malice murder convictions. The evidence shows that McKenzy was healthy when she left the hospital following her birth. The State's expert testified that the injuries to McKenzy were non-accidental, were caused by blunt force trauma, and could not have occurred during birth or the infant's initial stay at the hospital following her birth. This unequivocal expert testimony that McKenzy had been murdered went essentially unrebutted; the only meaningful point brought out on cross-examination was that the expert did *134not know who committed the homicide. And between the initial hospital stay and her return to the hospital in distress, McKenzy was in the sole care of the Debelbots - a married couple who insisted that neither of them did anything to harm McKenzy. Although a close question, we conclude that this evidence was sufficient to allow the jury to find both Albert and Ashley guilty of malice murder under the standard of Jackson v. Virginia, 443 U.S. 307, 99 S.Ct. 2781, 61 L.Ed.2d 560 (1979). See Gomez v. State, 301 Ga. 445, 452-453 (3), 801 S.E.2d 847 (2017) **539(rejecting couple's argument that evidence was insufficient to convict for felony murder of their child, where they were the only caregivers during the time of the injury and the State's medical experts testified that the injuries were inconsistent with an accident).
2. The Debelbots, who were represented by different trial counsel, both allege that their respective trial attorneys were constitutionally ineffective in several respects; Ashley also asserts direct error by the trial court. While we are deeply troubled by at least two of the claims of ineffective assistance of counsel, we conclude that the nature of the order below prevents us from reviewing meaningfully the most significant alleged deficiency by trial counsel for both Ashley and Albert - their failure to offer an alternative explanation for McKenzy's injury. Accordingly, we vacate the court's order denying the Debelbots' motion for new trial and remand for further proceedings.
To prevail on their claim, the Debelbots must show both deficient performance by counsel and resulting prejudice. See Strickland v. Washington, 466 U.S. 668, 687, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984). To show deficient performance, each appellant must prove that his or her counsel acted or failed to act in an objectively unreasonable way, considering all the circumstances and in the light of prevailing professional norms. See id. at 687-690, 104 S.Ct. 2052. "This is no easy showing, as the law recognizes a 'strong presumption' that counsel performed reasonably," and to overcome this presumption, an appellant "must show that no reasonable lawyer would have done what his lawyer did, or would have failed to do what his lawyer did not." Davis v. State, 299 Ga. 180, 183 (2), 787 S.E.2d 221 (2016). To establish prejudice, each appellant must demonstrate that there is "a reasonable probability that, but for counsel's unprofessional errors, the result of the proceeding would have been different. A reasonable probability is a probability sufficient to undermine confidence in the outcome." Strickland, 466 U.S. at 694, 104 S.Ct. 2052.
As discussed above, the case against the Debelbots was entirely circumstantial, and the only evidence that a crime had been committed at all was the expert medical testimony of Dr. Darrisaw, the State's medical expert who performed the autopsy. Neither Albert nor Ashley offered a rebuttal expert; Ashley's trial counsel had spoken to Dr. John Plunkett, but a scheduling conflict prevented that expert from testifying at trial.
At the motion for new trial hearings, the Debelbots called four doctors to testify about the McKenzy's injuries, and the motion-for-new-trial court qualified these witnesses as experts in their fields: Dr. Plunkett (general and forensic pathology), Dr. Peter Dehnel (pediatrics), Dr. Julie Mack (general and pediatric radiology), and Dr. Daniel Sahlein (radiology, neuroradiology, neurology, and **540interventional neuroradiology). The four experts reviewed, among other things, CT-scans of McKenzy's brain, McKenzy's medical records, the autopsy report, photographs, Dr. Mack's 3D reconstruction of the CT-scans of McKenzy's skull and brain, and relevant medical literature. Collectively, the experts testified that McKenzy suffered from a vascular event in utero that caused the bleeding in her brain and this event was the cause of her death. The experts also testified that the right side of McKenzy's skull was malformed, including that a piece of skull was missing, and concluded that a portion of McKenzy's brain matter was missing. The experts further opined that the trauma of the birthing process caused the additional and more acute fracturing to the left side of McKenzy's skull, and denied that post-birth trauma caused McKenzy's injuries. The State offered Dr. Darrisaw and two other experts in response, who disagreed. We note that the motion-for-new-trial court conducted the *135quite lengthy and complex proceedings with care - the hearing transcript alone runs over 2000 pages - and afforded both the Debelbots and the State ample opportunity to make their respective cases.
Despite having qualified all four of the Debelbots' expert witnesses as experts, however, the motion-for-new-trial court dismissed all their testimony in two separate ways. First, the court concluded in one sentence that all the Debelbots' witnesses, expert and otherwise, were not credible. And second, the court concluded that all of the Debelbots' medical evidence was inadmissible under our decision in Harper. As we explain further below, the sweeping nature of these conclusions precludes our meaningful review at this time, requiring that we vacate and remand for more precision.
We ordinarily afford great deference to credibility determinations by trial courts, including in the motion-for-new-trial context. See, e.g., Grant v. State, 295 Ga. 126, 130 (5), 757 S.E.2d 831 (2014) ("In reviewing a claim of ineffective assistance, we give deference to the trial court's factual findings and credibility determinations[.]"). But there are several different kinds of credibility at issue here, and the deference that we may afford depends on which kind the motion-for-new-trial court found lacking. First, there is the kind of credibility that goes to the truth or falsity of a fact that has been asserted. This is the kind of credibility determination we almost always afford strong deference. See, e.g., Cartwright v. State, 291 Ga. 498, 499-500 (2) (a), 731 S.E.2d 353 (2012) (deferring to trial court's crediting of trial counsel's testimony that trial counsel did not advise defendant, as defendant claimed, that his juvenile record could be used to impeach him); Murphy v. Balkcom, 245 Ga. 13, 13, 262 S.E.2d 784 (1980) ("[a]fter hearing all the evidence, the habeas judge, being the finder of facts, believed the state's witnesses and disbelieved **541petitioner and thus denied habeas relief" on the claim that attorney did not file an appeal as allegedly requested by petitioner). But the court could not have found that kind of credibility lacking from everything said by all of the Debelbots' witnesses during lengthy hearings spanning thousands of pages; in fact, many of the facts asserted by those witnesses were undisputed, were the same as facts asserted by the State's witnesses, or were otherwise amply supported by the record. To the extent that the court intended to resolve particular fact disputes on this basis, on remand it should specify on which such material points it found a lack of credibility.
Second, there is the kind of expert credibility that goes to whether the trier of fact believes the experts know what they are talking about based on whether they have sufficient experience and qualifications. Indeed, the State challenged the qualifications of the Debelbots' experts at considerable length. And during closing arguments, the motion-for-new-trial court summarized many of those challenges as an argument that one of the experts "wouldn't have survived a vigorous voir dire by the State," and that the judge who presided over the trial would not have accepted him as an expert. But - notwithstanding the vigorous voir dire the State actually conducted - the motion-for-new-trial court did accept the proffered experts as experts at the hearing. And the court did not explicitly reverse that determination in its order, so we are unsure of the extent to which the adverse credibility determination applied here.
Finally, the relative weight given to testimony of expert witnesses is also nearly always at issue in cases like these. But the question under Strickland is not whether the motion-for-new-trial court found the experts persuasive, but whether a reasonable juror could. See, e.g., Jones v. State, 302 Ga. 488, 493 (2) (b), 807 S.E.2d 344 (2017) (given the equivocal testimony of the proffered medical expert, there was no reasonable probability of a different outcome had trial counsel secured this expert witness for trial); Humphrey v. Nance, 293 Ga. 189, 223-224, 744 S.E.2d 706 (2013) (holding that because there was no reasonable probability the jury would have found expert credible, there was no reasonable probability that the outcome of the trial would have been different even if expert's testimony had been presented). For example, there may not be a reason to challenge an expert's qualifications, but a court could nevertheless find that a reasonable *136jury would not find the expert's opinion persuasive because it was too speculative, was too equivocal, or did not stand up to the strength of other evidence in the record. But we afford little deference to this kind of credibility determination. See Woodard v. State, 296 Ga. 803, 810 n. 5, 771 S.E.2d 362 (2015) (explaining that in examining whether a defendant has shown Strickland prejudice, we review the **542record de novo and weigh the evidence as we would expect reasonable jurors to have done). Because the order was not specific about what kinds of credibility the motion-for-new-trial court found lacking as to which pieces of critical testimony, we cannot determine what kind of deference we afford each of the credibility findings.
Similarly, the court's determination that all the Debelbots' medical evidence was inadmissible under Harper is insufficiently precise. Harper guides a trial court's determination of whether certain scientific evidence is admissible in a criminal case when it is based on a procedure, technique, or theory that "has reached a scientific stage of verifiable certainty[.]" Harper, 249 Ga. at 525-526 (1), 292 S.E.2d 389 ; see also, e.g., Caldwell v. State, 260 Ga. 278, 286-287 (1) (b), 393 S.E.2d 436 (1990) (concluding that DNA testing and analysis were based on sound scientific theory and, if proper procedures were followed, analysis could produce reliable results). Some of the medical evidence presented by the Debelbots' experts was simply the result of the application of commonplace medical expertise to undisputed facts. Some of the medical evidence was conclusions stemming from medical theories the State challenged as not generally accepted. And some of the medical evidence came from the use of imaging software - for the 3D reconstruction of CT-scans of McKenzy's skull and brain - that the State challenged as not FDA-approved. But although the State framed its Harper challenges differently over time and ultimately very broadly in closing arguments, many of its challenges appear to be more properly characterized as challenges to the qualification of the witnesses as experts or to the persuasiveness of the experts' testimony in the light of conflicting testimony from the State's experts. It seems plain that Harper- to the extent it applies here - does not apply to literally every word of medically related testimony from the Debelbots' experts. It also seems likely that at least portions of that testimony are appropriately analyzed under Harper.
Without more precise credibility and Harper findings, we are unable to conduct a Strickland analysis to determine whether either Albert's or Ashley's trial counsel was ineffective. Regarding an assessment of deficient performance, even if trial counsel could be said to have made certain tactical decisions, a defendant may still establish deficiency if he can show those tactical decisions were unreasonable in the light of the particular circumstances of a case. See Terry v. State, 284 Ga. 119, 120 (2), 663 S.E.2d 704 (2008) ("The reasonableness of counsel's conduct is examined from counsel's perspective at the time of trial and under the particular circumstances of the case."). Whether either trial counsel acted reasonably under the circumstances of this case depends largely, if not entirely, on whether the **543expert evidence presented at the hearings was admissible and credible. Compare Mosby v. State, 300 Ga. 450, 454 (2), 796 S.E.2d 277 (2017) ("Deficient performance of counsel is not shown by trial counsel's failure to present a witness whose testimony would have been inadmissible.") with Jowers v. State, 260 Ga. 459, 462, 396 S.E.2d 891 (1990) (trial counsel's failure to present evidence that was available and could have supported an alternative theory was not sound trial strategy).
Similarly, when we consider whether a defendant was prejudiced by the alleged deficiency of trial counsel, we measure the evidence that should have been - but was not - presented to the jury against the totality of the evidence that was presented. See Strickland, 466 U.S. at 695, 104 S.Ct. 2052 ("In making [the prejudice] determination, a court hearing an ineffectiveness claim must consider the totality of the evidence before the judge or jury."); see Humphrey v. Morrow, 289 Ga. 864, 870, 717 S.E.2d 168 (2011) (in determining prejudice from trial counsel's deficient performance in failing to present certain evidence, we must assess "how a jury might have reacted to the additional evidence"). This prejudice determination *137is necessarily affected by the quantity and quality of the evidence that was presented to the jury and that which should have been, and "a verdict or conclusion only weakly supported by the record is more likely to have been affected by errors than one with overwhelming record support." Strickland, 466 U.S. at 696, 104 S.Ct. 2052. Because the trial court's limited findings in the light of the voluminous testimony do not make it possible to review what, if any, evidence was properly ruled to be inadmissible under Harper, we cannot engage in a meaningful prejudice analysis here. Therefore we vacate the motion-for-new-trial court's order and remand for further findings, where the court should specify which material medical evidence is properly subject to such a Harper analysis and perform that analysis.
We also note our serious concern regarding the State's closing argument during trial that flatly stated that proof beyond a reasonable doubt in this murder case does not require the jury to be even 51% sure- in other words, requires less than even the preponderance of the evidence required to meet the burden of proof in a civil case. That is obviously wrong. A case like this one, where there was no direct evidence to prove that Albert, Ashley, both of them, or neither of them killed McKenzy, could turn on reasonable doubt, and the verdict could be affected by an argument that 50-50 proof is good enough. And the trial court's jury instruction on reasonable doubt - which in many cases may cure previous misstatements on the subject - did not cure the State's obviously wrong argument here. The State's point was to define reasonable doubt as not requiring the **544State to prove its case to "a mathematical certainty" - a phrase the State repeated twice. Of course, that is a phrase that occurs in the pattern instruction as well, and so when the trial court gave that instruction, it may well have been understood by the jury not as correcting the State's error, but as reinforcing it. We cannot conceive of any good reason that a competent criminal defense attorney could have to fail to object to such an egregious misstatement of the law. Whether that deficiency supports reversal, however, depends on weighing prejudice from all deficiencies, see Daughtry v. State, 296 Ga. 849, 853 (2), 770 S.E.2d 862 (2015) (the effect of prejudice resulting from trial counsel's deficient performance is viewed cumulatively), which will depend on the conclusions of the motion-for-new-trial court on remand.
Accordingly, we vacate the trial court's order denying the Debelbots' motion for new trial and remand for proceedings consistent with this opinion.3
Judgment vacated and case remanded.
All the Justices concur.