先日から導入している相対タイムは単純に\(\frac{走破タイム}{平均タイム}\)でした。しかし、これではノイズが多過ぎるって話から\(\frac{走破タイム}{基準タイム}\)に変更する事にしました。既にコーディングは済ませ、CSV出力して昨夜Model Builderで24時間学習スタートさせてます。この修正前に同じくModel Builderで24時間学習させたモデルは\(R^{2}\)値 0.2614で
| 3,455R | 1点 | 芝(1,689R) | ダート(1,646R) | 障害(120R) | 8頭以下(219R) | 9～12頭(809R) | 13頭以上(2,427R) | 多点 |
| 単勝 | 22.17% (72.16%) |
21.43% (69.27%) |
23.03% (76.52%) |
20.83% (53.00%) |
33.33% (85.34%) |
25.34% (73.41%) |
20.11% (70.55%) |
51.66% (77.06%) |
| 複勝 | 53.14% (83.40%) |
53.94% (83.94%) |
52.13% (83.27%) |
55.83% (77.58%) |
69.86% (94.20%) |
58.96% (84.05%) |
49.69% (82.20%) |
87.38% (83.10%) |
| 枠連 | 12.30% (78.40%) |
11.92% (83.28%) |
12.39% (73.58%) |
16.16% (85.66%) |
-- (--) |
13.72% (66.50%) |
11.04% (77.39%) |
26.96% (80.55%) |
| 馬連 | 9.44% (76.89%) |
9.24% (80.61%) |
9.36% (73.86%) |
13.33% (66.08%) |
22.83% (94.75%) |
10.38% (52.90%) |
7.91% (83.27%) |
20.67% (69.73%) |
| ワイド | 22.72% (85.32%) |
23.74% (89.26%) |
21.45% (82.10%) |
25.83% (74.00%) |
50.23% (102.24%) |
26.58% (74.65%) |
18.95% (87.35%) |
42.58% (81.30%) |
| 馬単 | 4.83% (63.26%) |
4.85% (70.91%) |
4.80% (57.39%) |
5.00% (36.08%) |
13.24% (111.96%) |
4.94% (43.42%) |
4.04% (65.47%) |
20.67% (68.61%) |
| 三連複 | 5.62% (81.04%) |
6.34% (72.21%) |
4.80% (87.15%) |
6.67% (121.58%) |
18.72% (149.04%) |
7.42% (78.01%) |
3.83% (75.92%) |
14.70% (69.06%) |
| 三連単 | 1.30% (68.64%) |
1.12% (60.04%) |
1.40% (71.16%) |
2.50% (155.08%) |
5.48% (315.71%) |
1.73% (80.53%) |
0.78% (42.38%) |
14.70% (65.17%) |
| 総合 | 54.56% (76.11%) |
55.54% (76.04%) |
53.34% (75.64%) |
57.50% (83.59%) |
74.89% (136.18%) |
59.83% (69.19%) |
50.97% (73.07%) |
87.73% (69.84%) |
という昨年の検証結果を出してますが、これがどこまで伸びるか今晩遅くにはここに追記しますね。現時点で13時間程度過ぎた所ではまだ\(R^{2}\)値は0.2611までです^^;
0 件のコメント:
コメントを投稿