How to write abstract for conference when you have no results yet?
Don't write results you don't have. Neither in the present, past or future tense. Just don't do it. Yet, I agree with you that there are circumstances where you do need to write an abstract on on-going work. For example, many big conferences in my field now ask for abstracts to be submitted up to 10 months in advance of the conference itself! If you are a post-doc staying on a 12-month project, you want to present something but you might not yet know how things will turn out. So, two techniques I propose:
Just write about the methodology, and present your goals in a general way, without “predicting” particular results but insisting on the importance of the topic. That is, emphasize strongly your points #1 and #2, and then describe point #3 as you would your “results”. Things like:
In this particular study, we compare the efficiency of methods A and B on given subsets of a reference database. We use a large number of different criteria for measuring efficiency, including …, … and … We also discuss in detail the implementation of subprocess X in method B, because its has not been specifically optimized in the existing literature.
I know it sounds vague, but that's the best you can achieve honestly, without pretending to know what you expect to find.
Bait and switch: if you have existing results in a closely related study, you can incorporate them as part of your results. Mix this approach with above, so that you have at least a few specific results to list in your point #4. Then, when you will make your presentation, just present your new results alongside the old (some people would remove completely the old results, but that makes it too much of a “bait and switch” for my taste). It is, after all, quite common for people to include newer results in orals/posters that they obtained after the original submission. It is not frowned upon, as long as you keep a decent agreement between the original abstract and the final content.
The challenge with ever-earlier deadlines for conferences (sometimes six months or more in advance of the actual date!) makes planning for a conference a very difficult prospect.
You're left with only a handful of options, none of them particularly appealing:
Submit an abstract on incomplete research, and hope that the work is completed in time for the conference. In this case, you say something like "we will present our work on X, Y, and Z." You make no claims about the findings related to your work in those areas, though. You also try to edit the abstract, as appropriate and if possible, to better reflect the subject material that you will actually present at the conference.
Submit an abstract on already completed work. The advantage is you know you will have the results and you can put together a good presentation. The downside of this is that it means you will be presenting last year's results at this year's conference. If you are in a "hot" field, this can mean ceding significant ground to your competitors if they get just a little bit luckier than you, and they have findings just before a deadline and you don't.
Ultimately, there's no right answer to which option to take. You have to decide this based on what is expected of you in your field, and what impact this will have on you and your career (if you can opt for the safer track, or if you have to go for the higher-risk option). The only thing that you should never do, as I said above, and as other posters have mentioned, is make claims that you have not obtained.
You cannot predict the future. You may obtain the results you hope for1. But things can also both go horribly wrong (your laboratory burns down, your samples mysteriously evaporate,...) or extremely interesting - you may happen to measure something beyond your dreams. Let your abstract only tell truths - what your (vague) setup is, what you want to measure and what you expect to happen. But don't pre-claim results when you cannot even foretell their existence for sure. Just be honest - say that you will present the results obtained by them, whatever they may be. I don't like cliffhangers, but they tend to work...
1 But make sure you don't "accidentally" measure only what you expect to be measured!