Talk:PollyVote

Critical discussion
The PollyVote system seems interesting to me and I tend to think it is notable and worth covering in Wikipedia. The article lacks critical discussion of the PollyVote model, however, and was tagged by another editor as "reading like an advertisement". I think part of the problem is that the article was originally written by editors who were simply too credulous about the academic articles about the pollyVote system's performance. Some of the articles are argumentative, and, although their publication in peer-reviewed journals indicates some review and some acceptance by the political science and/or forecasting fields, their assertions should not be accepted as "true" for the encyclopedia.

For example, it is currently stated in the article that "It has been shown that damped polls provide more reliable forecasts," with footnote reference to J. E. Campbell's 1996 article in the academic journal American Politics Quarterly. The statement would probably be a fine summary of previous reseaarch for use in a new academic article in many journals. However, it is an overstatement for an encyclopedia. For one thing, the statement is too broad: it cannot always be true that a damped poll model is always better than a non-damped model. Campbell's study presumably provides evidence that damped models perform better in some settings. It probably provides evidence in support of statistical arguments that reject a null hypothesis of no benefit of damping, in favor of an alternative. However, statistics do not prove anything, they just provide evidence. The results of one study are rarely sufficient to achieve any kind of consensus among academics, and, it is not asserted here that there is consensus among academics. For wikipedia articles, the standard of "truth" has to be much higher than for individual academic articles; wikipedia statements in articles (as opposed to talk pages) should be at the level of scientific-consensus-accepted truth.

I wonder if others have comments which could be used to put a more critical perspective into this article. To be clear, I would prefer balance but I do think that this model is a useful benchmark, very similar to the BCS model averaging college football team rankings by various models, that is worth describing in wikipedia. doncram (talk) 18:50, 9 October 2008 (UTC)