ALVIN B. RUBIN, Circuit Judge:

This appeal raises two issues: whether, in this products liability suit, the exclusion of evidence of post-accident design changes was prejudicial error and, if not, whether the evidence as a whole was sufficient to support both the jury verdict for defendants and the trial judge’s denial of post-judgment relief. Finding that the district judge erred neither in excluding the proffered testimony nor in finding the evidence sufficient, we affirm.

*885 I.

On June 2, 1977, a fire, followed by an explosion, occurred in a plant owned by Grenada Steel Industries, Inc. (Grenada Steel). In this diversity-based suit, Grenada contends that the fire was caused by a leak of acetylene gas through a valve on the cylinder that contained the gas. Alabama Oxygen Company, Inc., (Alabama Oxygen) sold acetylene gas to Grenada Steel. The gas was delivered in a metal cylinder, which was to be returned when the gas was consumed. The cylinder was equipped with a valve manufactured by Sherwood-Selpac Corporation (SSC) in December, 1972, utilizing a rubber o-ring seal around the valve plug to prevent leakage of acetylene gas. The valve’s design was based on a patent issued in 1965. The cylinder, with its SSC valve, was supplied to Alabama Oxygen in 1973. One year later, in 1974, SSC halted production of this type of valve. It was no longer being marketed when the fire occurred in 1977.

Liberty Mutual Insurance Company insured Grenada Steel. It paid $608,990.38 to Grenada Steel for losses .resulting from the fire. It then filed this suit, as subrogee, to recover from Alabama Oxygen and SSC the amount paid to Grenada Steel. Grenada Steel later joined in the suit to assert its own claim for damages not covered by its insurance. Since their interests are in most respects identical, we refer to both claimants jointly as Grenada Steel.

In the six-day jury trial, Grenada Steel proffered evidence that, following the fire, Regó, a competitive valve manufacturer, manufactured and designed an acetylene valve based on an alternative design and that SSC had itself manufactured a differently designed model. The district court excluded this evidence. Grenada Steel also introduced the testimony of an expert witness that the valve was defective. SSC countered with testimony that the valve’s design had a good reputation in the industry, its valves had been approved by Underwriters Laboratories, and the same design had been used in the industry for a number of years. Each side also presented contradictory factual and expert evidence concerning where the fire originated.

The district court, with counsel’s assistance, formulated both special interrogatories and a general verdict. In response to the special interrogatories, the jury found that the valve was not defective or unreasonably dangerous, had undergone substantial changes in condition after it was sold by the manufacturer and before the fire, and was not unfit for ordinary use for the purpose intended. It rendered a general verdict for the defendants. Grenada Steel does not assert that the answers to the special interrogatories were either inconsistent with each other or with the general verdict but only that they are “incorrect and do not square with the facts.” (Emphasis in Grenada Steel’s brief.)

II.

We start with an elementary proposition: in diversity cases, state substantive law applies. Erie Railroad v. Tompkins, 304 U.S. 64, 58 S.Ct. 817, 82 L.Ed. 1188 (1938). Therefore, in determining the sufficiency of the evidence we follow Mississippi courts, which have adopted the American Law Institute’s Restatement (Second) of Torts § 402A as the appropriate standard of strict liability for product manufacturers. Page v. Barko Hydraulics, 673 F.2d 134, 136 n. 1 (5th Cir.1982); State Stove Mfg. Co. v. Hodges, 189 So.2d 113, 118 (Miss.1966), cert. denied, 386 U.S. 912, 87 S.Ct. 860, 17 L.Ed.2d 784 (1967). In matters of procedure, however, such as the admissibility of evidence, federal rules apply. Fed.R.Evid. 1101(b); Rabon v. Automatic Fasteners, Inc., 672 F.2d 1231, 1238 n. 14 (5th Cir. 1982); Johnson v. C. Ellis & Sons Iron Works, Inc., 604 F.2d 950, 957 (5th Cir.1979).

The district court relied on Rule 407 of the Federal Rules of Evidence to exclude evidence that, after the accident, SSC and Rego marketed valves that were designed differently from the one patented in 1965. That rule provides:

When, after an event, measures are taken which, if taken previously, would have made the event less likely to occur, evi *886 dence of the subsequent measures is not admissible to prove negligence or culpable' conduct in connection with the event. This rule does not require the exclusion of evidence of subsequent measures when offered for another purpose, such as proving ... feasibility of precautionary measures, if controverted ....

In excluding the testimony, the district judge held that the rule applies not only to claims of negligence but also to those based on strict liability. He also found that the feasibility of a different design had not been challenged by SSC, and, therefore, ruled that evidence of the new design was inadmissible on that issue.

The initial question is whether Rule 407 applies to product liability cases. This is a question on which we have not yet passed. See Foster v. Ford Motor Co., 616 F.2d 1304, 1309 n. 11 (5th Cir.1980). 1 We examine first its applicability to the evidence concerning changes by SSC because the rule on its face, as discussed more fully below, does not deal with alternative designs or products introduced by third parties.

The Eighth Circuit has held repeatedly that Rule 407 is simply inapplicable to products liability cases. Unterburger v. Snow Co., 630 F.2d 599, 603 (8th Cir.1980); Farner v. Paccar, Inc., 562 F.2d 518, 528 n. 20 (8th Cir.1977); Robbins v. Farmer’s Union Grain Terminal Ass’n, 552 F.2d 788, 793 (8th Cir.1977); Abel v. J.C. Penney Co., 488 F.Supp. 891 (D.Minn.1980), aff’d, 660 F.2d 720 (8th Cir.1981). This view rests on two arguments. First, the rule is limited by its terms to efforts to prove negligence or culpable conduct. Yet in strict liability cases, the Eighth Circuit reasons, the focus is not on negligence or culpable conduct. Instead, liability stems from the unreasonably dangerous nature of the product. Second, its decisions rely heavily upon the policy considerations set forth in Ault v. International Harvester Co., 13 Cal.3d 113, 117 Cal.Rptr. 812, 528 P.2d 1148 (1974).

In Ault, decided before the adoption of the Federal Rules of Evidence, the court concluded that the blanket exclusionary rule was inapplicable to strict liability eases for economic reasons. The court asserted that it is “manifestly unrealistic to suggest that ... a producer will forego making improvements in its product, and risk innumerable additional lawsuits and the attendant adverse effect upon its public image, simply because evidence of the adoption may be admitted in an action founded on strict liability.” Id. at 121, 117 Cal.Rptr. at 815, 528 P.2d a.t 1151. 2

*887 Other circuits have taken a different path and held that Rule 407 applies with equal force in strict liability cases. Hall v. American Steamship Co., 688 F.2d 1062 (6th Cir. 1982); Josephs v. Harris Corp., 677 F.2d 985, 990-91 (3d Cir.1982); Cann v. Ford Motor Co., 658 F.2d 54 (2d Cir.1981), cert. denied, 456 U.S. 960, 102 S.Ct. 2036, 72 L.Ed.2d 484 (1982); Werner v. Upjohn Co., 628 F.2d 848 (4th Cir.1980); cert. denied, 449 U.S. 1080, 101 S.Ct. 862, 66 L.Ed.2d 804 (1981). In Bauman v. Volkswagenwerk Aktiengesellschaft, 621 F.2d 230 (6th Cir.1980), and Roy v. Star Chopper Co., 584 F.2d 1124, 1134 (1st Cir.1978), the courts applied Rule 407 in product liability cases without discussing the issue. In Oberst v. International Harvester Co., 640 F.2d 863 (7th Cir. 1980), the court held evidence of subsequent repairs inadmissible. It is not clear, however, whether this decision was based on Rule 407 or on Illinois law. 3

In Werner the Fourth Circuit noted that one of the basic concerns underlying Rule 407 was the encouragement of voluntary repairs by manufacturers. The court questioned “why this policy should apply any differently where the complaint is based on strict liability as well as negligence.” 628 F.2d at 848. The Werner opinion further notes that the Ault approach would admit subsequent repairs as evidence that the product was defective even in cases where the manufacturer was simply making further improvements to an already safe product. Id. See also Hall v. American Steamship Co., 688 F.2d at 1067 (following Werner).

Voluntary change to improve a product and reduce the possible hazard to a user should be encouraged. While there is no evidence concerning whether admission of evidence of change would deter such action by manufacturers, the assumption in the rule that it might have a deterrent effect is not demonstrably inapplicable to manufacturers upon whom strict liability is imposed. But our decision does not rest only on theses about the influence of possible tort liability on human conduct. It rests more firmly on the proposition that evidence of subsequent repair or change has little relevance to whether the product in question was defective at some previous time.

It has been suggested that evidence of a change in the manufacturing process or the design of a product has probative value because “a business is not likely to change a product unless the change promotes safety and is feasible.” R.' Lempert & S. Saltzburg, A Modern Approach to Evidence 189 (1977). But this argument is based on little direct evidence of why manufacturers make product changes. Similarly, most of the arguments made against admissibility rely on undocumented assumptions about how evidence of such changes might affect litigation. A priori judgments concerning why manufacturers do or do not alter their products, made by such dubious experts as judges, lawyers, and law professors, suffer from excessive reliance on logical deduction *888 and surmise without the benefit of evidence of industry practice or economic factors. It seems to us, with no greater expertise than like-trained lawyers and judges, that changes in design or in manufacturing process might be made after an accident for a number of different reasons: simply to avoid another injury, as a sort of admission of error, because a better way has been discovered, or to implement an idea or plan conceived before the accident. The cost of making the change, the acceptance of the altered product in the market place, and whether the change is one of several that can be made contemporaneously might also be considerations.

We cannot really know why changes are made by industry generally or why a change was made in a particular product in the absence of evidence on the question. Instead, we ought to consider the probative value of such evidence on the point at issue. The real question is whether the product or its' design was defective at the time the product was sold. See S. Saltzburg & K. Redden, Federal Rules of Evidence Manual 181 (3d ed.1982). The jury’s attention should be directed to whether the product was reasonably safe at the time it was manufactured. In this case, for example, there was ample expert testimony concerning that very point. The introduction of evidence about subsequent changes in the product or its design threatens to confuse the jury by diverting its attention from whether the product was defective at the relevant time to what was done later. Id. at 182 (addressing only product design). Interpreted to require the evidence to focus on the time when the product was sold, Rule 407 would conform to the policy expressed in Rule 403, the exclusion of relevant information if its probative value is substantially outweighed by the danger of confusion. See 23 C. Wright & K. Graham, supra, § 5288, at 144.

For these reasons, we follow the path taken by the First, Second, Third, Fourth and Sixth Circuits and hold Rule 407 applicable to strict liability cases.

Grenada Steel argues that, even if Rule 407 applies, the evidence was admissible under the exception to that rule for evidence offered to show feasibility of precautionary measures. But this exception pertains if the issue of feasibility is contested. Grenada Steel argues that, in design defect cases, feasibility is “inherently” an issue and an integral element of its proof. This, however, assumes that their claim that another design was feasible makes a contest. It takes two to tango, in court as well as on the ballroom floor, and the mere assertion that the manufacturer did not make a change does not controvert the feasibility of change. The manufacturer raised no issue that the alternative design suggested by Grenada Steel’s experts was not feasible. Its defense was only that the design it used was not defective.

Feasibility may “almost always” be in question in design defect cases, as suggested by Professors S. Saltzburg and K. Redden, in their excellent Federal Rules of Evidence Manual 180 (3d ed.1982). But “almost always” like “hardly ever” admits of the unusual case that prompts the qualifying adverb. Sometimes, as here, the manufacturer does not suggest that another design was impractical but only that it adopted an acceptable one. It is not obliged to prove perfection but only that the product “meets the reasonable expectations of the ordinary consumer as to its safety.” Page v. Barko Hydraulics, 673 F.2d at 138.

The Fourth Circuit has ruled that “it is clear from the face of ... rule [407] that an affirmative concession, is not required. Rather, feasibility is not in issue unless controverted by the defendant.” Werner v. Upjohn Co., 628 F.2d at 855. Two commentators suggest a uniform rule:

The administration of Rule 407 would be greatly simplified if the appellate courts were to hold that in all of these [negligence and product liability] situations, feasibility of precautionary measures will be deemed “controverted” unless the defendant is prepared to make an unequivocal admission of feasibility.

*889 23 C. Wright & K. Graham, Federal Practice & Procedure § 5288, at 144 (1980). The defendant manufacturer was not put to that alternative. We need not, therefore, determine whether denial of a request for an admission of feasibility would per se create a contest. In this case, however, the defendant clearly stated that feasibility was not controverted. No request for an admission to that effect was made. Therefore, the evidence was properly excluded under rule 407.

The district court also excluded evidence that Rego, another manufacturer, later offered an alternatively designed valve. Grenada Steel argues that the admission of evidence showing that someone other than the defendant made a design change cannot discourage voluntary repairs. The party making the repair is not penalized by the admission of the evidence. Therefore, neither the text of rule 407 nor the policy underlying it excludes evidence of subsequent repairs made by someone other than the defendant. See Louisville & Nashville Railroad Co. v. Williams, 370 F.2d 839, 843-44 (5th Cir.1966); Farrier v. Paccar, Inc., 562 F.2d 518, 520 n. 20 (8th Cir.1977); Lolie v. Ohio Brass Co., 502 F.2d 741 (7th Cir. 1974) (per curiam); Wallner v. Kitchens of Sara Lee, Inc., 419 F.2d 1028, 1032 (7th Cir.1970); Steele v. Wiedemann Co., 280 F.2d 380 (3d Cir.1960); 10 J. Moore & H. Bendix, Moore’s Federal Practice § 407.05 (1982); 23 C. Wright & K. Graham, Federal Practice & Procedure § 5284, at 113 (1980); McCormick on Evidence § 275, at 667 (E. Cleary 2d ed.1972).

Nevertheless, we think the district court’s exclusion of this evidence was proper because it lacked sufficient probative value and injected the dangers of confusion and misleading the jury. In Ward v. Hobart Manufacturing Co., 450 F.2d 1176 (5th Cir. 1971), a Mississippi diversity-based negligent design case, we held that it was error for a district court to consider design changes developed after the manufacture of the product in question. Such repairs were simply irrelevant to the reasonableness of the design at the time of manufacture. Alternative designs may indicate that the product was unreasonably dangerous, see Wade, On the Nature of Strict Tort Liability For Products, 44 Miss.L.J. 825, 837 (1973), but only if they were available at the time of manufacture. We fail to see how an alternative design, developed by another person years after the product in question was manufactured, is relevant to whether the product was reasonably safe at the time it was made. Therefore, the district court properly excluded evidence of Rego’s alternative design in the absence of an issue on which that evidence would have had sufficient probative value to outweigh its unfair prejudicial effect.

III.

As Grenada Steel observes, the trial was fought in a classic battle of the experts. The defendants’ experts testified, based largely on their interpretation of post-fire photographs, that the fire started in another part of the plant than the room where the acetylene cylinder was used. Grenada Steel argues that this testimony “was general, self-serving, and based on secondhand data and information, whereas [Grenada Steel’s] proof of the defective and unreasonably dangerous condition of the subject valve was firmly supported by firsthand observations, tests, and examinations of the subject valve and its interior as well as the fire-damaged premises shortly after the fire.” These orotund arguments were addressed to the jury and that was the proper audience. It was for the jury to decide which of the experts was more credible, which used the more reliable data, and whose opinion — if any — -the jury would accept. The jury might credit one or more of the expert opinions or it might reject them all. Post-Boeing, the argument that, when the jury accepts one expert rather than another, we substitute our appraisal for theirs will not fly, for the standard is fastly placed: after reviewing all of the evidence “in the light and with all reasonable inferences most favorable to the party opposed to the motion,” we must be persuaded that “the facts and inferences point so strongly and overwhelmingly in favor of one party *890 that ... reasonable [jurors] could not arrive at a contrary verdict.” Boeing Co. v. Shipman, 411 F.2d 365, 374 (5th Cir.1969) (en banc). Neither the trial court nor this one decides “by which side has the better of the case, nor ... [on the basis of] a complete absence of probative facts to support a jury verdict.” Judges decide only whether there is “a conflict in substantial evidence” that creates a jury question. Id. at 375. The district court acted well within its discretion in denying Grenada Steel’s motion for judgment notwithstanding the verdict.

Our position with respect to the district judge’s denial of the motion for a new trial is similar. The initial decision to grant or deny such a motion rests with the district judge. This court reviews the denial of a motion for a new trial only for abuse of discretion. Shows v. Jamison Bedding, Inc., 671 F.2d 927, 930 (5th Cir.1982). So long as “a reasonable basis exists for the jury’s verdict,” we will not disturb the district court’s ruling. United States v. 329.73 Acres of Land, 666 F.2d 281, 284, rehearing en banc granted on other grounds, 681 F.2d 264 (5th Cir.1982). We see no need to address directly each of the evidentiary arguments presented by Grenada Steel. We simply note that there was sufficient evidence to support the jury’s verdict.

Finally, we note that the jury’s general verdict, coupled with its answers to the. special interrogatories, indicates that it believed the fire did not originate with the acetylene cylinder. Thus, even if our analysis of the evidentiary issues were different, the result would be the same. If the jury found that the valve did not cause the damage, the question of its defectiveness is moot.

In view of this determination, it is unnecessary for us to consider the separate arguments urged by Alabama Oxygen that it was entitled to exoneration on the basis that it was but a bailor of the cylinder and valve and had only the duty to make a reasonable inspection and its final contention that, being insured by Liberty Mutual, under the same policy issued to Grenada Steel, it was immune to the subrogation action.

For these reasons, the judgment is AFFIRMED.

Notes

. The problem has given rise to much commentary. See 23 C. Wright & K. Graham, Federal Practice and Procedure § 5285 (1980); Costello & Weinberger, The Subsequent Repair Doctrine and Products Liability, 51 N.Y.S.B.J. 463 (1979); Davis, Evidence of Post-Accident Failures, Modifications and Design Changes in Products Liability Litigation, 6 St. Mary’s L.J. 792 (1975); Note, The Case For the Renovated Repair Rule: Admission of Evidence of Subsequent Repairs Against the Mass Producer in Strict Products Liability, 29 Am.U.L.Rev. 135 (1979); Note, Products Liability and Evidence of Subsequent Repairs, 1972 Duke L.J. 837; Note, Evidence of Subsequent Repairs: Yesterday, Today and Tomorrow, 9 U.Cal.D.L.Rev. 421 (1976); Note, Federal Rule of Evidence 407 and its State Variations: The Courts Perform Some “Subsequent Remedial Measures” Of Their Own in Products Liability Cases, 49 U.Mo.K.C.L.Rev. 338 (1981); Comment, Subsequently Remedying Strict Products Liability: Cann v. Ford, 14 Conn.L.Rev. 759 (1982). See also Department of Commerce, Model Uniform Product Liability Act § 107, 44 Fed.Reg. 62,714, 62,728 (October 31, 1979); Annot., 50 A.L.R. Fed. 935 (1980); Annot., 47 A.L.R.3d 1001 (1974).

. The Ault decision has been followed in other states both by decision and by legislative action or court rule. See Caterpillar Tractor Co. v. Beck, 624 P.2d 790, 793-94 (Alaska 1981); Good v. A.B. Chance Co., 39 Colo.App. 70, 78-80, 565 P.2d 217, 224 (1977); Caprara v. Chrysler Corp., 52 N.Y.2d 114, 124-25, 436 N.Y.S.2d 251, 256, 417 N.E.2d 545, 550 (1981); Ginnis v. Mapes Hotel Corp., 86 Nev. 408, 416-17, 470 P.2d 135, 140 (1970); Shaffer v. Honeywell, Inc., 249 N.W.2d 251 (S.D.1976); Chart v. General Motors Corp., 80 Wis.2d 91, 258 N.W.2d 680 (1977).

Maine admits evidence of subsequent repairs even if they are offered to show negligence. Me.R.Evid. 407(a). There is thus no need for a strict liability exception in that state. The text of Committee Comments to Wyo.R.Evid. 407 and Colo.R.Evid. 407 indicate that, while those states adopted the language of the federal rule, they also incorporated the Ault decision. Ha- *887 wan R.Evid. 407 is also identical in wording to the federal rule, except that it adds to the list of exceptions; “proving dangerous defect in products liability cases.” Alaska R.Evid. 407 contains a similar proviso.

. Several states have also continued to follow the traditional rule excluding evidence of subsequent repair even in strict liability cases. See Hallmark v. Allied Prods. Corp., 132 Ariz. 434, 440-41, 646 P.2d 319, 325-26 (1982); Moldovan v. Allis Chalmers Mfg. Co., 83 Mich.App. 373, 268 N.W.2d 656 (Mich.Ct.App.1978); Price v. Buckingham Mfg. Co., 110 N.J.Super. 462, 266 A.2d 140, 141 (1970) (per curiam); Lamonica v. Outboard Marine Corp., 48 Ohio App.2d 43, 355 N.E.2d 533, 535 (Ohio Ct.App.1976); Haysom v. Coleman Lantern Co., 89 Wash.2d 474, 573 P.2d 785 (1978) (en banc).

Ariz.Rev.Stat.Ann. § 12-686(a) expressly excludes evidence of subsequent repair to prove evidence of a defect in products liability actions. Neb.Rev.Stat. § 27-407 is identical to the federal rule, but adds at the end: “Negligence or culpable conduct, as used in this rule, shall include, but not be limited to, the manufacture and sale of a defective product.”

Other cases continue to apply the traditional rule in products liability actions based upon a failure to warn. In those cases, however, the basis for liability more closely resembles negligence than strict liability. See Ortho Pharmaceutical Corp. v. Chapman, 388 N.E.2d 541, 561-63 (Ind.App.1979); Smith v. E.R. Squibb & Sons, Inc., 405 Mich. 79, 273 N.W.2d 476 (Mich. 1979).