本文以零智慧代理人,建構了一個簡單的政治預測模擬市場,代理人基模型根據台灣各縣市的真實政治傾向在不同群聚效果下,對模擬結果與未來事件交易所其預測2008年及2012年的「總統選舉各候選人得票率預測」合約作比較,觀察市場偏誤的原因。 經過模擬,證實本文的人工市場是有效率的,模型的設定能捕捉到Forsythe et al. (1999) 對預測市場的系統偏差 (systematic biases) 和失誤(mistakes),所提出心理學中的一廂情願效果 (wishful thinking effect) 和錯誤共識效果 (false consensus effect),此外,從不同群聚效果的模擬結果發現,未來事件交易所其預測2008年及2012年的「總統選舉各候選人得票率預測」合約,交易結果隱藏著台灣政治傾向分布。 Zero-Intelligence agents constructs a simple political simulation of market in this paper. Agent-based model based on real political tendencies under the effect of different segregations in the cities and counties of Taiwan. The simulation results compared with the contract of the election for the president of Taiwan, forecasted by XFuture, to observe the market bias. As the simulation, the efficiency of the artificial market is confirmed. The model can match the opinion on the prediction market systematic bias and mistakes approved by Forsythe, et al (1999).As the view of Forsythe, et al (1999), the market involves the wishful thinking effect and the false consensus effect. In addition, the simulation results, the political tendencies in Taiwan, matches the contract of the election for the president of Taiwan forecasted by XFuture.