POSNER, Circuit Judge.

The district judge denied the defendant's application for bail pending appeal on the strength of the judge's opinion in United States v. Chiattello, 599 F. Supp. 970, 973 (N.D.Ind. 1985), which held that under the Bail Reform Act of 1984 "a release of a defendant on appeal may be granted only if the chance for reversal [of his conviction] is substantial. . . ." We must decide whether this is the correct standard under the new Act, a question made easier for us by the recent decision of another panel of this circuit in United States v. Bilanzich, 771 F.2d 292, 297-98 (7th Cir. 1985).

Under the new Act, bail may not be granted pending appeal unless the appeal "raises a substantial question of law or fact likely to result in reversal or an order for a new trial." 18 U.S.C. § 3143(b)(2). In United States v. Miller, 753 F.2d 19, 23-24 (3d Cir. 1985), the Third Circuit held that the quoted language entitles the defendant to bail pending appeal (provided he is not a danger to anyone or likely to flee, 18 U.S.C. § 3143(b)(1)) if his appeal raises a substantial question that is likely to result in a reversal if the court of appeals answers the question in the way the defendant asks it to do. This of course is not what the language says; but if it were read literally the district judge could not grant bail pending appeal unless he thought the conviction was going to be reversed — and if he thought that, why would he not have set aside the conviction himself, without putting the defendant to the bother of appealing and us to the bother of reversing? The literal reading amounts to saying that district judges shall not grant bail pending appeal, leaving the courts of appeals with sole authority to do that. See Fed.R.App.P. 9(b). As explained in United States v. Molt, 758 F.2d 1198, 1200 (7th Cir. 1985), we can apply the literal statutory standard (whether the conviction is likely to be reversed) without any of the awkwardness involved in a judge's predicting that his own decision is likely to be reversed.

If Congress meant to abolish the district courts' power to grant bail pending appeal, it chose an awfully roundabout way of expressing its desire. Roundabout or not, we must give effect to its desire if we can discern it. But there is no indication, beyond what can be found in the words "likely to result in reversal," that Congress really wanted to eliminate the district courts' authority to grant bail pending appeal; and against the inference that it did, there is the statement in the Senate report that section 3143(b)(2) "requires an affirmative finding that the chance for reversal is substantial. This gives recognition to the basic principle that a conviction is presumed to be correct." S.Rep. No. 225, 98th Cong., 1st Sess. 27 (1983), U.S. Code Cong. Admin.News pp. 3182, 3210. There is a difference between finding that "the chance for reversal is substantial" and finding that the conviction is "likely to be reversed." The former implies that the judge is not sure he is right, the latter that he thinks he probably is wrong.

All the courts of appeals that have discussed the standard under the new Act have agreed — more or less — with Miller. The most recent cases, which discuss the earlier ones, are, besides our own decision in Bilanzich, United States v. Affleck, 765 F.2d 944, 952-53 (10th Cir. 1985) (en banc); United States v. Handy, 761 F.2d 1279, 1280-83 (9th Cir. 1985); United States v. Powell, 761 F.2d 1227 (8th Cir. 1985) (en banc); United States v. Randell, 761 F.2d 122, 124-25 (2d Cir. 1985), and United States v. Valera-Elizondo, 761 F.2d 1020 (5th Cir. 1985). These cases identify an issue, which may be semantic but may be more, regarding the meaning of "substantial" in the statutory phrase, "substantial question of law or fact likely to result in reversal or an order for a new trial." Does it mean nonfrivolous, or does it mean, as first held in United States v. Giancola, 754 F.2d 898, 901 (11th Cir. 1985) (per curiam), "a `close' question or one that very well could be decided the other way"? All the recent cases — including Bilanzich — with the possible exception of Handy, compare 761 F.2d at 1280 (majority opinion), with id. at 1284-85 (dissent), adopt the formulation in Giancola. To define substantial as merely nonfrivolous would be to read out of the statute the words "likely to result in reversal." Yet those words were intended to have some force, if not necessarily the force of their literal meanings. Relatively few federal criminal appeals are frivolous, though most fail. To define substantial as nonfrivolous would allow bail pending appeal to be granted as a matter of course, which Congress obviously did not want. So in its particular statutory context "substantial" must, as we held in Bilanzich, see 298-99, mean close; it must mean that the appeal could readily go either way, that it is a toss-up or nearly so. So stated the test is congruent with the statement in the Senate report that the judge must find "that the chance for reversal is substantial." He just does not have to find that he is more likely to be reversed than affirmed.

The district judge in this case found that there was no substantial chance of reversal. He applied a correct test and applied it correctly, so the denial of bail pending appeal is

AFFIRMED.