Passengers who suffered injuries in a train wreck appeal the summary judgment that terminated their action. The first question is whether the trial court erred in deciding that defendant National Passenger Railroad Corporation (commonly known as “Amtrak”) was not negligent, as a matter of law, because the criminal act of sabotage in creating a defect in the rail was an unforeseeable intervening, superseding act, and the sole cause of the train wreck. The second question is whether the court abused its discretion in excluding the affidavits of plaintiffs’ experts, without which the plaintiffs had no proof of negligence on the part of Amtrak. We affirm.

FACTUAL AND PROCEDURAL HISTORY

At approximately 1:23 a.m. MST on October, 9, 1995, an Amtrak passenger train derailed in the Arizona desert, near Milepost 846.86, on a track then owned and maintained by Southern Pacific Transportation Company. 1

It is undisputed that the derailment was the result of a purposeful act of sabotage by one or more unknown persons. Physical evidence at the scene indicated that unknown individuals had deliberately removed the bolts and spikes holding the south rail in place. Two angle bars that hold the rails together had been removed. They were later found near the point of derailment (“POD”). The loosened rail was driven ahead by derailed wheels and was found lying 20 feet ahead of the POD. No evidence of the amount of lateral displacement of the rail before impact survived the force of the derailing. Likewise, no physical evidence revealed what signs of a defective roadbed, if any, would have been visible to the train crew prior to impact, assuming that the crew had been keeping a proper lookout as the locomotive approached the POD. The crew depositions contain the only evidence in the record about the quality of the lookout.

The saboteurs had taken deliberate pains to conceal their efforts. The track on which the derailment occurred was equipped with an electric warning system designed to illuminate a red light if the current flowing through the rail was interrupted at any point between “block” sig *828 nals. Because angle bars holding the ends of the rails together do not conduct the electric current adequately, a bond wire is connected to the ends of the rails and runs between them under the connecting angle bars to ensure that the warning circuit is complete. The saboteurs had circumvented the system and rewired it so that the block signal would continue to show a green light after the angle bars had been removed. The last green light that the train crew observed before the train derailed was located approximately 1.5 miles before the POD. The train was moving at approximately 50 miles per hour, a safe speed at the time and place of the event.

Two experienced Amtrak passenger locomotive engineers were in the cab of the lead locomotive. Gean Haffey (“Haffey”) was the engineer and Gary Lawrence (“Lawrence”) was the assistant engineer. In his deposition, Haffey said that he was “rolling the train”-inspecting the train as it rounded the corner by looking in the mirror or looking back through the window to make sure that the train was in normal condition. The Lawrence deposition contains two, possibly conflicting, statements concerning his actions immediately preceding the derailment. At one point, he testified that he was looking ahead through the windscreen. At another point, he testified that he was completing’ calculations on a projected arrival time when the train derailed. The quality of lookout maintained by the crew in this case is obviously relevant on the issue of negligence, but its materiality depends upon the existence of evidence that some visible sign of a defective track could have been seen if a perfect lookout had been kept at all times.

The train consisted of two engine units and twelve cars. Both engines and eight of the cars left the track. The POD was remote and inaccessible by paved roads. After the wreck, government agents, railroad investigators, and other experts visited the scene, inspected the mutilation of the roadbed, and prepared reports. The evidence showed that, pursuant to Federal Railroad Administration regulations, a qualified track inspector had twice inspected the track near Milepost 846.86 in the week before the derailment, and a tie plate at milepost 846.82 was replaced on October 5, 1995. There was no evidence of any defect in the track or roadbed prior to the separation of the rails by the sabotage. A report prepared by the former Southern Pacific Regional Engineer, Maintenance of Way, David Wickersham, was filed as an exhibit. The report described the scene at the POD as of 6:30 a.m. following the wreck. It also included photographs showing the debris left on the roadbed after the train left the track.

The complaint claimed that the defendant railroads failed to: (a) properly inspect, maintain, and repair the tracks; (b) modernize the tracks with “continuously welded rail;” (c) establish a proper speed limit; and (d) monitor the rails at the point of derailment. They claimed that Amtrak failed to: (a) slow to an appropriate speed; (b) keep a proper lookout; (c) properly train its engine crew; (d) properly inspect the rails; and (e) provide crashworthy cars.

Defendants moved for summary judgment on all claims. The district court granted the motion, holding: (a) plaintiffs had abandoned all claims except the negligence claim arising out of the operation of the train in the moments before derailment; (b) the negligence claim failed as a matter of law because the saboteurs’ criminal act was an unforeseeable intervening, superseding cause; and (c) the plaintiffs’ proffered expert testimony in support of their negligence claim was inadmissible as speculative and not sufficiently reliable. Plaintiffs have limited their appeal to the *829 issue of negligence on the part of Amtrak and the district court’s rejection of their proffered expert testimony.

DISCUSSION

We review de novo the evidence evaluated by the summary judgment court in the light most favorable to the nonmoving party. See Berry v. Valence Tech., Inc., 175 F.3d 699 , 703 (9th Cir.1999). Like the trial court, we must draw all reasonable inferences supported by the evidence in favor of the non-moving party and then decide whether any genuine issues of material fact exist, and whether the district court correctly applied the relevant substantive law. See id. A fact issue is genuine “if the evidence is such that a reasonable jury could return a verdict for the nonmoving party.” Anderson v. Liberty Lobby, Inc., 477 U.S. 242, 248 , 106 S.Ct. 2505 , 91 L.Ed.2d 202 (1986). “A grant of summary judgment may be sustained on any basis supported by the record.” Daubert v. Merrell Dow Pharmaceuticals, Inc., 43 F.3d 1811 , 1314 (9th Cir.1995).

Assuming, arguendo, that the saboteurs’ criminal act did not “supersede” the alleged negligence of Amtrak, the crucial question is whether, if both engineers had been keeping a proper lookout, under all the circumstances then existing, they would have seen anything to cause them to take remedial action.

To survive summary judgment the plaintiffs have the burden to produce some evidence, other than speculation or guesswork, that something was present at the POD to be seen by crewmen keeping a proper lookout. It is at this point that the plaintiffs’ case failed. Apart from the proffered expert testimony, the record contains no evidence that a better lookout would have discovered a visible defect in the track in time to avoid the derailment.

THE EXPERT AFFIDAVITS

The plaintiffs offered the affidavits of three experts. The court excluded the proposed testimony of all three. Plaintiffs contend that the district court mishandled its gatekeeping function for the admission of testimony offered under Federal Rule of Evidence 702 as set forth in Kumho Tire Co., Ltd. v. Carmichael, 526 U.S. 137 , 119 S.Ct. 1167 , 143 L.Ed.2d 238 (1999) and Daubert v. Merrell Dow Pharmaceuticals, Inc., 509 U.S. 579 , 113 S.Ct. 2786 , 125 L.Ed.2d 469 (1993).

In Daubert, the Court charged trial judges with the responsibility of acting as gatekeepers to “ ‘ensure that any and all scientific testimony ... is not only relevant, but reliable,’ ” and the Court in Kumho Tire clarified that this gatekeeper function applies to all expert testimony, not just testimony based in science. See Kumho Tire, 526 U.S. at 147 , 119 S.Ct. 1167 (quoting Daubert, 509 U.S. at 589 , 113 S.Ct. 2786 ). Rule 702, which governs the admissibility of expert opinion testimony, states: “If scientific, technical, or other specialized knowledge will assist the trier of fact to understand the evidence or to determine a fact in issue, a witness qualified as an expert by knowledge, skill, experience, training, or education, may testify thereto in the form of an opinion or otherwise.”

First, Rule 702 requires that the evidence “assist the trier of fact,” or in other words, that it be relevant. Daubert, 509 U.S. at 591 , 113 S.Ct. 2786 . Second, the rule demands that the evidence be reliable. See id. at 590 , 113 S.Ct. 2786 . Rule 702 requires that expert testimony relate to scientific, technical, or other specialized knowledge, which does not include unsupported speculation and subjective beliefs. See id.

*830 Plaintiffs argue, with some support in the record, that the court rejected all the proffered affidavits without any analysis of the experts’ qualifications, without examination of the basis of the experts’ proffered opinions, and without any attempt to determine whether their opinions met the Kumho Tire test of the rigors of the professional discipline. See Kumho Tire, 526 U.S. at 152 , 119 S.Ct. 1167 (noting that the objective of the gatekeeping requirement “is to make certain that an expert ... employs in the courtroom the same level of intellectual rigor that characterizes the practice of an expert in the relevant field”). Plaintiffs are partly right, but do not establish an abuse of discretion.

The trial court did consider the proffered evidence, but the consideration appears to have been abbreviated because the court had already determined that there was no liability. However, despite the difficulty created by the court having conflated its views of liability with its view of the reliability of the expert testimony, we have reviewed all of the expert affidavits, together with the other evidence of the physical facts that was before the experts. We conclude that the trial court correctly excluded the expert testimony for the reasons that follow.

1. Charles Culver

One expert, Charles Culver (“Culver”), a locomotive engineer with 27 years of experience and court-tested qualifications in other cases, was qualified professionally to testify, but his testimony concerning the duty to maintain a proper lookout in front of the train went no further than to establish a fact never really in dispute: locomotive engineers have a continuing duty to keep a lookout. He went on to offer an opinion that a track separa-, tion should have been visible on a moonlit night in the added light of the locomotive. The facts before him, however, provided no basis for an assumption that a visible gap in the rail, or other evidence of a defect, was present to be seen if the crewmen had been keeping a proper lookout.

The district court abused its discretion by excluding Culver’s testimony regarding railroad operating procedure and standards of care of a locomotive engineer. That testimony was relevant on the subject of an engineer’s lookout and appropriate procedure. Moreover, based on Culver’s extensive experience as an engineer and as an instructor in rules, safety, train handling, field efficiency testing, and derailment investigation, at least part of his testimony was reliable. See Kumho Tire, 526 U.S. at 150 , 119 S.Ct. 1167 (“Engineering testimony rests upon scientific foundations, the reliability of which will be at issue in some cases.... In other cases, the relevant reliability concerns may focus upon personal knowledge or experience.”); United States v. Hankey, 203 F.3d 1160, 1168-69 (9th Cir.2000). However, the fatal problem with Culver’s opinion beyond the need for a proper lookout, was that Culver assumed that the saboteurs had left behind something visible to anyone keeping a proper lookout. That assumption finds no support in the physical facts as described by the reports and other evidence in the record.

Culver’s conclusion that the engineers on board the train should have observed the rail when it was illuminated by the engine’s lights at approximately 500 feet (that “the tops of rails frequently gleam, or shine, in the light thrown by the locomotive headlights,”) was good as far as it went. But there was no factual basis for the assumption that anything was visible on “the twin ribbons of rail [that] often can be seen shining for a considerable distance ahead of the locomotive.” Therefore, Cul-ver’s opinion was properly excluded be *831 cause it was not sufficiently founded on facts.

2. James Sobek

The district court apparently compared the testimony of Culver with the testimony of another of plaintiffs’ experts, James Sobek (“Sobek”), a technical lighting engineer (not an engine driver). Without any factual knowledge of how much displacement in centimeters or inches the saboteurs had achieved, Sobek opined that the displaced rail created a visible phenomenon that could be seen at 500 feet from the point of derailment.

However, contrary to Culver’s testimony, Sobek claimed that the rails themselves “would have appeared black,” but that the ballast and ties surrounding the rails “were diffuse reflectors and would have been seen as much brighter surfaces,” and that “[t]he net effect is that the rails would have been seen as two black curved lines against the grayish ballast and ties.” The district court found the opinions of the two experts on what could have been seen to be equally inadmissible, stating:

It is not necessary to decide whether the discrepancy between Culver’s opinion and that of Sobek would necessarily disqualify them from consideration had their opinions been solidly based on some proof that there was something to be seen at the POD. However, there was no evidence that there was anything the engineers should have seen. The circumstantial evidence that the saboteurs had carefully concealed their handiwork by circumventing the electric warning system suggests that they likewise took care not to leave visible evidence of track displacement for the engineers to see as they approached the disconnected rail at night, no matter how closely they watched the forward track. Therefore, the district court properly excluded the testimony of the lighting expert because it was not sufficiently fact based to be reliable on the question of the distance from which a proper lookout could have seen the sabotaged rail.

3. James Loumiet

Similarly, the proffered testimony of James Loumiet (“Loumiet”), the accident reconstruction expert, was inadmissible on the issue of negligence. The affidavit was unsupported by facts and made estimates of damage after assuming that the sabotage left gaps in the rail that would have been visible. No one but the saboteurs themselves knew what the rails looked like at the point of sabotage before the derailment. There was no factual support in the record for Loumiet’s opinion that the engineer should have applied the train’s emergency brakes 400 feet before the train reached the POD. That opinion would have been helpful to a jury evaluating evidence of negligence if it had been based on evidence that the defective rail could have been seen at 400 feet.

In United States v. Various Slot Machines on Guam, 658 F.2d 697, 700 (9th Cir.1981), we said that “in the context of a motion for summary judgment, an expert must back up his opinion with specific facts.” The factual basis for the expert’s opinion must be stated in the expert’s affidavit and although the underlying factual *832 details need not be disclosed in the affidavit, the underlying facts must exist. See Bulthuis v. Rexall Corp., 789 F.2d 1315, 1318 (9th Cir.1985). “If further facts are desired, the movant may request and the district court may require their disclosure.” Id. In the case at bar, our study of the record does not reveal the existence of underlying facts which could support the opinions of the lighting and locomotive engineers about what the defendant’s crew members should have seen that would have given a careful operator notice of a hazard in time to avoid it.

CONCLUSION

Because there was no proof that any sign of the sabotage was visible to properly alert crewmen in time for them to apply the brakes, there was no material question of fact to go to a jury on the issue of due care.

AFFIRMED.

. The named defendants, after a series of mergers, now operate under the name of the Union Pacific Railroad.