世界盃

32 強 90 分鐘勝平負:8 個模型加一個校準融合,並排比較。

8 個模型 + 多校準融合,全程不看市場價 · 預測時間 2026-06-29

市場盲測——全程不讀取任何賠率或市場價。只預測每場 90 分鐘內的勝 / 平 / 負;打平也算一種結果,之後靠加時、點球決出誰晉級,不在預測範圍。8 個模型各自打分,一個多校準融合彙總成對外結論。已踢完的加拿大對南非不在其中;每場結束後我們都會公開打分。

第 1 場置信 ·
🇧🇷巴西日本🇯🇵

看好 巴西 90 分鐘內取勝,約 51%

巴西 51%平局 33%日本 16%
九個預測模型並排對比
Dixon-Coles (Bayesian-shrunk)444214
xG-corrected Elo393625
PRODEGY split attack/defence Elo473023
Physical-decay dynamic-K Elo403624
Tactical style-clash mElo453817
Line-breaks & offers efficiency (gradient-boosted)79210
Passing-network structure (random forest)69256
Stacked Ensemble433720
Multi-calibrated 8-in-1採用513316
巴西平局日本
查看推理與證據 →
第 2 場置信 ·
🇩🇪德國巴拉圭🇵🇾

看好 德國 90 分鐘內取勝,約 47%

德國 47%平局 36%巴拉圭 17%
九個預測模型並排對比
Dixon-Coles (Bayesian-shrunk)483814
xG-corrected Elo433621
PRODEGY split attack/defence Elo453223
Physical-decay dynamic-K Elo403624
Tactical style-clash mElo344026
Line-breaks & offers efficiency (gradient-boosted)62344
Passing-network structure (random forest)60355
Stacked Ensemble423721
Multi-calibrated 8-in-1採用473617
德國平局巴拉圭
查看推理與證據 →
第 3 場置信 ·
🇳🇱荷蘭摩洛哥🇲🇦

看好 荷蘭 90 分鐘內取勝,約 46%

荷蘭 46%平局 32%摩洛哥 22%
九個預測模型並排對比
Dixon-Coles (Bayesian-shrunk)343927
xG-corrected Elo443323
PRODEGY split attack/defence Elo382933
Physical-decay dynamic-K Elo443422
Tactical style-clash mElo293833
Line-breaks & offers efficiency (gradient-boosted)80191
Passing-network structure (random forest)573112
Stacked Ensemble443521
Multi-calibrated 8-in-1採用463222
荷蘭平局摩洛哥
查看推理與證據 →
第 4 場置信 ·
🇨🇮科特迪瓦挪威🇳🇴

看好 挪威 90 分鐘內取勝,約 41%

科特迪瓦 29%平局 30%挪威 41%
九個預測模型並排對比
Dixon-Coles (Bayesian-shrunk)323434
xG-corrected Elo162856
PRODEGY split attack/defence Elo292843
Physical-decay dynamic-K Elo172855
Tactical style-clash mElo173251
Line-breaks & offers efficiency (gradient-boosted)324127
Passing-network structure (random forest)501337
Stacked Ensemble433324
Multi-calibrated 8-in-1採用293041
科特迪瓦平局挪威
查看推理與證據 →
第 5 場置信 ·
🇫🇷法國瑞典🇸🇪

看好 法國 90 分鐘內取勝,約 68%

法國 68%平局 21%瑞典 11%
九個預測模型並排對比
Dixon-Coles (Bayesian-shrunk)69247
xG-corrected Elo67249
PRODEGY split attack/defence Elo382537
Physical-decay dynamic-K Elo68239
Tactical style-clash mElo80182
Line-breaks & offers efficiency (gradient-boosted)88111
Passing-network structure (random forest)82135
Stacked Ensemble493021
Multi-calibrated 8-in-1採用682111
法國平局瑞典
查看推理與證據 →
第 6 場置信 ·
🇲🇽墨西哥厄瓜多爾🇪🇨

看好 墨西哥 90 分鐘內取勝,約 38%

墨西哥 38%平局 34%厄瓜多爾 28%
九個預測模型並排對比
Dixon-Coles (Bayesian-shrunk)284725
xG-corrected Elo263737
PRODEGY split attack/defence Elo283240
Physical-decay dynamic-K Elo293734
Tactical style-clash mElo65305
Line-breaks & offers efficiency (gradient-boosted)542422
Passing-network structure (random forest)332839
Stacked Ensemble413821
Multi-calibrated 8-in-1採用383428
墨西哥平局厄瓜多爾
查看推理與證據 →
第 7 場置信 ·
🇧🇪比利時塞內加爾🇸🇳

看好 比利時 90 分鐘內取勝,約 44%

比利時 44%平局 37%塞內加爾 19%
九個預測模型並排對比
Dixon-Coles (Bayesian-shrunk)463816
xG-corrected Elo343828
PRODEGY split attack/defence Elo343333
Physical-decay dynamic-K Elo343828
Tactical style-clash mElo453916
Line-breaks & offers efficiency (gradient-boosted)63361
Passing-network structure (random forest)583111
Stacked Ensemble413920
Multi-calibrated 8-in-1採用443719
比利時平局塞內加爾
查看推理與證據 →
第 8 場置信 ·
🏴󠁧󠁢󠁥󠁮󠁧󠁿英格蘭剛果民主🇨🇩

看好 英格蘭 90 分鐘內取勝,約 58%

英格蘭 58%平局 30%剛果民主 12%
九個預測模型並排對比
Dixon-Coles (Bayesian-shrunk)63289
xG-corrected Elo68239
PRODEGY split attack/defence Elo502426
Physical-decay dynamic-K Elo67249
Tactical style-clash mElo543115
Line-breaks & offers efficiency (gradient-boosted)49465
Passing-network structure (random forest)67285
Stacked Ensemble463222
Multi-calibrated 8-in-1採用583012
英格蘭平局剛果民主
查看推理與證據 →
第 9 場置信 ·
🇺🇸美國波黑🇧🇦

看好 美國 90 分鐘內取勝,約 48%

美國 48%平局 33%波黑 19%
九個預測模型並排對比
Dixon-Coles (Bayesian-shrunk)513613
xG-corrected Elo493318
PRODEGY split attack/defence Elo354223
Physical-decay dynamic-K Elo453421
Tactical style-clash mElo513514
Line-breaks & offers efficiency (gradient-boosted)592417
Passing-network structure (random forest)502228
Stacked Ensemble433621
Multi-calibrated 8-in-1採用483319
美國平局波黑
查看推理與證據 →
第 10 場置信 ·
🇨🇭瑞士阿爾及利亞🇩🇿

看好 瑞士 90 分鐘內取勝,約 49%

瑞士 49%平局 33%阿爾及利亞 18%
九個預測模型並排對比
Dixon-Coles (Bayesian-shrunk)563113
xG-corrected Elo433324
PRODEGY split attack/defence Elo413029
Physical-decay dynamic-K Elo443323
Tactical style-clash mElo343729
Line-breaks & offers efficiency (gradient-boosted)63370
Passing-network structure (random forest)71245
Stacked Ensemble443521
Multi-calibrated 8-in-1採用493318
瑞士平局阿爾及利亞
查看推理與證據 →
第 11 場置信 ·
🇪🇸西班牙奧地利🇦🇹

看好 西班牙 90 分鐘內取勝,約 75%

西班牙 75%平局 20%奧地利 5%
九個預測模型並排對比
Dixon-Coles (Bayesian-shrunk)76204
xG-corrected Elo672310
PRODEGY split attack/defence Elo432235
Physical-decay dynamic-K Elo662311
Tactical style-clash mElo503119
Line-breaks & offers efficiency (gradient-boosted)65314
Passing-network structure (random forest)38584
Stacked Ensemble463222
Multi-calibrated 8-in-1採用75205
西班牙平局奧地利
查看推理與證據 →
第 12 場置信 ·
🇵🇹葡萄牙克羅地亞🇭🇷

看好 葡萄牙 90 分鐘內取勝,約 43%

葡萄牙 43%平局 38%克羅地亞 19%
九個預測模型並排對比
Dixon-Coles (Bayesian-shrunk)464113
xG-corrected Elo403624
PRODEGY split attack/defence Elo343333
Physical-decay dynamic-K Elo383725
Tactical style-clash mElo364024
Line-breaks & offers efficiency (gradient-boosted)54460
Passing-network structure (random forest)58357
Stacked Ensemble413821
Multi-calibrated 8-in-1採用433819
葡萄牙平局克羅地亞
查看推理與證據 →
第 13 場置信 ·
🇦🇷阿根廷佛得角🇨🇻

看好 阿根廷 90 分鐘內取勝,約 64%

阿根廷 64%平局 26%佛得角 10%
九個預測模型並排對比
Dixon-Coles (Bayesian-shrunk)70255
xG-corrected Elo77194
PRODEGY split attack/defence Elo472726
Physical-decay dynamic-K Elo77194
Tactical style-clash mElo77203
Line-breaks & offers efficiency (gradient-boosted)70264
Passing-network structure (random forest)523612
Stacked Ensemble463321
Multi-calibrated 8-in-1採用642610
阿根廷平局佛得角
查看推理與證據 →
第 14 場置信 ·
🇦🇺澳大利亞埃及🇪🇬

90 分鐘內最可能打平,約 38%

澳大利亞 26%平局 38%埃及 36%
九個預測模型並排對比
Dixon-Coles (Bayesian-shrunk)284725
xG-corrected Elo373924
PRODEGY split attack/defence Elo234136
Physical-decay dynamic-K Elo383824
Tactical style-clash mElo274231
Line-breaks & offers efficiency (gradient-boosted)02080
Passing-network structure (random forest)153748
Stacked Ensemble374122
Multi-calibrated 8-in-1採用263836
澳大利亞平局埃及
查看推理與證據 →
第 15 場置信 ·
🇨🇴哥倫比亞加納🇬🇭

看好 哥倫比亞 90 分鐘內取勝,約 60%

哥倫比亞 60%平局 29%加納 11%
九個預測模型並排對比
Dixon-Coles (Bayesian-shrunk)63307
xG-corrected Elo73225
PRODEGY split attack/defence Elo423028
Physical-decay dynamic-K Elo73225
Tactical style-clash mElo593011
Line-breaks & offers efficiency (gradient-boosted)67285
Passing-network structure (random forest)58375
Stacked Ensemble463321
Multi-calibrated 8-in-1採用602911
哥倫比亞平局加納
查看推理與證據 →

只給勝 / 平 / 負概率——不涉及金額、賠率或任何市場數據。對外公佈的結論來自多重校準融合模型;下方 8 行模型展示了各方法在哪裡一致、在哪裡分歧。比賽結算後再做打分。