L. HAND, Circuit Judge.

The Valldemosa Steamship Company, Ltd. appeals from a decree, entered upon two libels consolidated into one suit, arising out of the collision of its steamship, “Valldemosa,” and the “Woodrow Wilson,” owned by the United States, in which the “Valldemosa” was held solely at fault. The collision occurred' at about 2:54 A. M. on May 11, 1944-, in a fog so dense that it was described as leaving a visibility of only 200 or 300 feet. Two convoys, aggregating 110 vessels, had gathered on the 10th some distance to the southeast of New York Harbor; the “Valldemosa” was in one, and the “Wilson” in the other. To enter the Harbor all the vessels were arranged in a single group of three columns half a mile apart, and they approached Ambrose Lightship through the “Swept Channel” (a channel kept clear of mines) which ran on a course 342°, true, for a distance of about fifteen miles from the “Gong Buoy,” (designated in the record as “Buoy E”) to “Buoy I,” which was two miles about west-southwest of the Lightship. The fog set in at about 1:35 A. M. while the vessels were upon this course and before they had taken on the pilots, who were waiting on a pilot boat that had come down and was lying off Ambrose. The-Vall-demosa” with the fifth ship in the easterly column, and the “Wilson” the fourteenth ship in the westerly column; both carried regulation navigation lights. The “Vallde-mosa” made out the pilot boat a short distance west of Ambrose and took a pilot on board. The pilots had agreed that because of a fog all the vessels should make for an “emergency anchorage,” assigned for that purpose by the Navy, north of Ambrose Channel; and the collision occurred while the “Valldemosa” was seeking this refuge. The “Wilson” in the west column was awaiting a pilot, but had got none before the collision; her story was that she was slowly drifting on her original course when the “Valldemosa” overtook her and struck her. She was giving fog signals of two blasts every two minutes, which the International Rules prescribe for a vessel not under way, but which were improper under the Inland Rules. About a minute before the collision she made out the “Valldemosa” approaching about one point abaft her beam and four hundred feet away. Believing that the only chance of escape was more speed, her master ordered the engines full speed ahead, and put his wheel hard right to swing his stern away from the “Vallde-mosa.” The vessels met at an acute angle, the port bow of the “Valldemosa” striking the “Wilson” amidship and scraping along her starboard side. The judge found that the “Wilson” had kept her course — 342°, true — and that the “Valldemosa” had directed her course northerly until the Am-brose signal bore southeast, at which time she altered to 303°, true, and the signal bore astern. He found the “Valldemosa” at fault by having cut too soon across the projected line of the convoy, and for too great speed. ITe exonerated the “Wilson” because she was not going too fast, and had not crossed the projected line of the convoy. He thought it unnecessary to find the place where the collision occurred.

The “Valldemosa’s” position, when she took on board the pilot we can fix with reasonable accuracy; she must have been not far west of Ambrose Light; her master said not over 1800 feet, her pilot, not over half a mile. Presumably she had not yet left her position in the eastern of the three columns, and we shall assume that she was about 2000 feet off the Lightship. The “Valldemosa’s” witnesses swore that the first leg of her course was “northerly,”— the second mate even said that he brought her “to north by standard compass” — and that this leg she kept “until Ambrose bore southeast by the compass,” a matter of “about ten minutes.” The pilot, being satisfied that he “had cleared the area from Gedney to Ambrose,” then laid “a northwest course,” — 303°—so that the “Am-brose Light vessel bore approximately astern.” He had been on this course about ten minutes when the collision happened. As we have said, the judge accepted this testimony and found accordingly; but he *761also found that the “Wilson” had held her course — 342°—until the collision. If so, the collision happened where the projected course of the westerly column cut the second leg — 303°, true — of the “Valldeniosa’s” course. It makes no difference whether we take the testimony of the “Valldemosa’s” pilot literally, that she was on bo1h legs for the same length of time. If so, the collision was well north of the entrance to Ambrose Channel, which nobody asserts and which is absurd. On the other hand, if the “northerly” leg was shorter than the other, as the “Valldemosa” plots it on a chart, the collision nevertheless took place much further north than it can possibly have done, if the “Wilson” had not left her course, 342°, true. In short the two findings are inconsistent, and we are forced to choose between them on the probabilities, as is so often the case in collisions in the fog. It is most unlikely that the “Wilson” should have deliberately changed her course to the east; for to do so would have brought her across the path of the two columns to starboard in which, since she was only the fourteenth ship in her own column, new ships were constantly arriving, to say nothing of those clustered about which had already arrived. We must disregard the tide for all the vessels were equally subject to it; and it is the merest speculation to say that she lost steerageway and drifted to the east Incidentally, this putative drift quite contradicts the “Valldemosa’s” other claim that she was going at too high speed.

Coining then to the “Valldemosa,” it is true that the first leg of her course was designed to clear her of the oncoming columns; but it is highly significant that the length of that leg was measured by the bearing of the Ambrose fog signal. (We do not forget the estimate of ten minutes, but that is shown to be utterly unreliable— as such estimates always are — by the fact, already noticed, that if the two legs were of equal length, the result is absurd.) So far as the turn was made in reliance upon the bearing of the fog signal, it is enough entirely to discredit it that admiralty courts have over and over again refused to accept estimates based upon the direction of sound in a fog, after the classic discussion of Judge Brown in The Lepanto, D.C., 21 F. 651, 655-658.1 The most likely inference out of the inevitable confusion and the conflict of findings is that the “Valldemosa” turned left much earlier than she now admits, and cut across the columns very shortly after she took the pilot on board.

Moreover, even if tills be not true, and if she laid her course, 303°, true, about when she claims she did, she was still at fault for not giving the oncoming vessels a wider berth. The position of the “emergency anchorage” upon which the pilots had agreed is left somewhat vague, but there is nothing to suggest that it could not have been reached equally well if the “Vallde-mosa” had extended her supposed “northerly” leg much longer than she even claims she did, and had reserved her turn to the northwest — parallel with the Ambrose Channel — until she had in fact got well clear. In such a fog and with so many vessels arriving one after another, crowding on one another and all in search of pilots, it was by no means unlikely that some should drift substantially north of Ambrose, just as the “Wilson” did. How far they might go, no one could be sure, and the safe course— indeed the only safe course for a ship laden with benzol — was to give them as wide a berth as possible. The point where the projected course of the eastern column cuts the 303° course of the “Valldemosa” as she plots it in the chart, is only about a mile from Ambrose. To cut across without more margin than that, under the conditions as they existed oil that night, was in our opinion to cake undue risks and was a grave fault.

We think that Bondy, J., was also right about tile “Valldemosa’s” speed. It is true *762that the doctrine of The Nacoochee,2 does not require a ship to lose steerageway in a fog, and that the pilot says he did no more than keep it; but we think that this was plainly untrue. Although the “Valldemosa” on her own story had three minutes within which to stop after she made out the “Wilson,” she failed to do so; yet at “slow” her master swore that she could have stopped in one minute. The angle of the collision was probably not more than forty to forty-five degrees, the courses being 303° and 342° respectively; and the mutual approach was much slower than if the ships had been meeting at a broader angle. It is possible that the pilot was over-anxious to get to a safe anchorage as soon as he could with his perilous cargo; but whatever the reason, he was certainly moving too fast. We think that the “Valldemosa” was guilty of both the faults found by Bondy, J.

We have already discussed the position of the “Wilson” and have said in substance that the “Valldemosa” has failed to prove that she had encroached upon the waters of the other columns. Since the “Valldemosa” had the burden of proof as to the “Wilson’s” faults that is enough to dispose of this one. Her other possible faults were her speed and her fog signals. If her own witnesses are to be believed her speed was as low as practicable; and there was no conceivable reason for more speed and every reason against it. The argument from the “counter” record of engine revolutions between midnight and the collision, does not impress us in the least. By no legerdemain can this be made an index of her speed at the time of collision; at most it would show that for about three hours she had been moving faster on the average than she admitted. We do not know where she was at 1:35 A.M. when it first began to get “foggy,” and we cannot impute improper speed to her after it shut in. Moreover, the computation is unreliable anyway; for it presupposes that her revolutions at full speed were only 55, contrary to the testimony of her engineer who said that they were about 75. If so, the discrepancy is not 2045 revolutions, but only about 600, if we allow for possible margins in logging. Such a difference will support no inference of any sort.

Last as to the “Wilson’s” fog signals. We are assuming that the collision took place north of the line between Ambrose Lightship and the Navesink Light— the limit for the Inland Rules — and it follows that the Inland Rules applied, and that the “Wilson” was blowing the wrong signals. The “Valldemosa” argues that Bondy, J., was mistaken in finding that beyond a reasonable doubt this could not have contributed to the collision. However, even though we were to concede as much for argument, the “Wilson’s” master was not chargeable with fault. He had nothing to go by but the Ambrose signals, whose bearing was totally unreliable, as we have already said. He could not possibly have known when he had crossed the line; and indeed his error was at most not more than five points, and Judge Brown in The Lepante, supra,3 excused an error as great as that The chance of being misled was moreover greatly increased by the fact that all the vessels, as they came up, seriatim, began to blow fog signals. The ensuing “chorus” must have prodigiously added to the confusion; amid such a bedlam, to charge a master with fault for failing to know when he passed the line between two lights both obscured, would be the last measure of injustice.

Decree affirmed.

The Umbria. 166 U.S. 404, 408, 17 S.Ct. 610, 41 L.Ed. 1053; Lie v. San Francisco & Portland S. S. Co.. 243 U. S. 291, 296, 37 S.Ct. 270, 61 L.Ed. 720; The Oregon, D.C., 27 F. 751, 757; The F. & P. M. No. 2, C.O.Wis., 36 F. 264, 269; The Niagara. 2 Cir., 84 F. 902, 803; The Went Brooklyn, 2 Cir., 106 F. 751, 752; The Beaver, 9 Cir., 219 F. 131, 138; aff'd in Lie v. San Francisco & Portland. S. S. Co., supra; Steffens v. United States, 2 Cir., 32 F.2d 208, 208; The Eastern Knight, D.C., 56 F.2d 428; The Silver Palm, 9 Cir., 94 F.2d 754, 781.

137 U.S. 330, 11 S.Ct. 122, 34 L.Ed. 687.

21 F. 651.