Forum Moderators: open
You're designing a metasearch engine and you have access to any search engine on the web, Google, Fast, AV, INK, etc.
Now look at ixQuick.com. The algo is simple. Look at the top ten results in each engine, give it one point for a top ten, and then rank it. ( I don't know how it ranks ties ).
Now I'm not trying to write a metasearch engine. But I think it would be an interesting discussion for ideas.
My algo, simple, and probably used before, is:
1. Grab top ten results like ixQuick.
2. Give each search engine that you query a weighted factor.
3. For each result from 1 to 10 (change to 10 to 1) and then multiply by the weight that you've given in step 2.
4. Sort results and return.
Question:
1. Which search engines would you use if you had access to all?
2. Given your list, what weight factor would you give them? Weight factor, for comparison purposes, would be given in percentages....
I jotted these numbers down:
Google: 40%
Fast: 25%
AOL: 12%
MSN: 10%
N.Light: 5%
NBCi: 3%
Excite: 2%
HotBot 2%
AV: 1%
So if I got a 4 on Google and a 2 on Fast, that's rated (11-4) * 0.4 + (11 - 2)* 0.25 = 5.05, but another site with 4 on Fast and 2 on Google = 5.35. Simple enough.
(Did that in my head so correct me if I'm wrong, anyway, the second site should rank higher).
My question is: What does everyone else think? This is a subjective question, ranking search engines on relevance.
Who would you rely on? What % weight factor would you give?
And who would you not bother to even query? Do you skip the last 5 in that list? Did I leave anyone out?
(will get a diverse set of relatively relevant sites)
Google: 15%
AV: 15%
Fast:15%
(further boost sites found popular by click popularity)
Direct Hit:10%
Yahoo:10%
(boost sites reviewed by editors)
ODP: 8%
Looksmart: 8%
(secondary directory sites giving importance to onpage criteria)
AOL: 5%
MSN: 5%
(other less relevant databases)
NBCi: 3%
Excite: 3%
HotBot(inktomi results): 3%
