The ruling that the APIs were not copyrightable was directly tied to the details of the specific APIs in this case. Among other things, this is because existing Ninth Circuit caselaw was unfavorable to an absolute bar on API copyrights, so Judge Alsup couldn't easily rule on the law independently of the case in front of him. To make an analogy, he could have found a phone number or address uncopyrightable as a general rule, but to decide if a copied portion of a book is copyrightable he has to look at what's in it (both the portion and the book as a whole).
Appeals courts aren't supposed to overturn factual findings unless they are "clearly erroneous". Judge Alsup was careful with his findings and he has experience with programming that the selected appeals-court judges are unlikely to have. While it is theoretically possible they'd second-guess his factual findings, it would be a very big stretch. And if you take his factual findings as-is, it becomes pretty hard to challenge his legal interpretation of those facts (especially since none of the coverage I've read suggests that his legal reasoning was unusual or controversial).
It isn't impossible for Oracle to win on appeal, but their changes are slim. Beyond that, if they win on appeal all they get is a do-over on the API copyright question (and it is unclear what the parameters of that do-over would be). Given that we know that the jury deadlocked at 9-3 in favor of Google on fair use (and was 11-1 in Google's favor earlier in the deliberations), even if Oracle got their do-over the odds are against them winning anything meaningful from it.
Also, it's important that the Judge instructed the jury to assume APIs were copyrightable. So even if the API copyright judgement is reversed the jury's decisions still stand.
The jury made no decision on fair use, and the judge declared a mistrial on that portion of the case. His subsequent ruling moots the whole issue (no copyright, no infringement) --- but if that ruling were reversed, there would be a new trial on at least fair use. And IIRC, Google has indicated that they'd want a new trial on all copyright matters at that point.
Sites like Groklaw (who understand legal language far better than me) were praising the judge for how well he constructed his ruling to withstand appeals. There are no obvious gaping holes in his logic, and so I am sure Oracle's legal team will have to write a pretty clever appeal brief in order to be granted one.
Is the ruling really 'pretty airtight' ? ie: does Oracle really have no chance at winning on an appeal?