As　an　emerging　word　puzzle　game　provided　by　New　York　Times　daily,　Wordle　attracts　everyone　worldwide　because　the　rules　are　simple,　and　the　game　is　specifically　designed　for　relaxing　in　fragmented　time.　This　study　aims　to　enhance　the　Wordle　game　experience　by　analyzing　historical　data,　predicting　future　gamer　numbers　using　double　ARIMA　that　improves　upon　traditional　ARIMA,　and　exploring　word　attributes.　We　investigated　the　impact　of　word　frequency　and　the　number　of　repeated　letters　on　puzzle　difficulty,　revealing　a　correlation　through　Pearson＇s　coefficient　test.　Utilizing　a　BP　neural　network,　we　predicted　word　difficulty　and　optimized　K-means　clustering　for　a　comprehensive　analysis.　Additionally,　we　discovered　a　relationship　between　gamer　numbers　and　special　days,　conducting　sensitivity　analysis　to　refine　our　model.　Our　study　contributes　by　pioneering　the　double　ARIMA　approach,　refining　clustering　methods,　inferring　word　difficulty　mathematically,　and　uncovering　hidden　insights　in　human　behavior　through　multidimensional　data　analysis.　©　2023　ACM.