If you read my Cliff Top Algo thread, (mostly me posting notes as I analyse this algo) I believe the current algorithm is the same as before but with a Bayesian filter applied. The pages are analysed by the filter, if you get a high score on that filter the page will be removed from the serps as a spam duplicate, a middle score and it will be pushed down the ranking, only low score will let it rise to the top.
This means you should avoid single themed sites or single theme pages to appear high in the Google rankings.
Use keywords sparsely.
Mix themes on the site and on the pages.
Avoid 'specialist' sites.
The more specialist the site, the more inbound linkage is needed to counter the effect of the penalty.
If your content is on one subject consider sharing sites with other people in different topics to dilute your keywords.
Normally to avoid duplication you should use a good Thesaurus, but beware of choosing obscure words that will likely be higher scoring than more common words.
For example I enter a single keyword.
1. Specialist directory site PR7
2. Specialist directory site PR6
3. On topic niche PR6 + fresh boost
4. Same as 3
5. On topic niche site PR7
6. Diverse site,
7. Off topic site
8. On topic niche site with fresh boost
9. Diverse site
10. Diverse site
11. Diverse site
12. Diverse site (free hosting!)
13. Dual theme site that mentions the word in passing.
14. Diverse theme site
Note how quickly the results tend to diverse topic & vague sites. Compare that to yahoo.
1. On topic niche site
2. On topic niche site
3. On topic niche site
4. On topic niche site
5. On topic niche site
6. On topic niche site
7. Diverse site
9. On topic niche site
10. On topic niche site
11. On topic niche site
12. On topic niche site
13. On topic niche site
14. Diverse site
15. On topic niche site
16. Diverse edu site
17. On topic niche site
This is the effect of the filter, those niche sites score too well for the subject keyword, so they sink in the ratings. The more ontopic the lower they go.
8. In the yahoo listing I would consider to be the definitive English language site on this subject. (Not in top 100 on Google).