-
Create view
chao_draft.lmt_testYears_CSS_null_norm_prob_for_pointsby adding numeric columns for position_L/R/D/C and country_USA/CAN/EURO. -
Calculate sum of positive weights multiplying attribute difference, sum of negative weights multiplying attribute difference, and individual attribute difference multiplying weight for two cohorts.(player drafted in 2001/2 ; player drafted in 2007/8). code can be found here: https://github.com/sfu-cl-lab/Yeti-Thesis-Project/blob/master/Decision_Trees/LMT/calculate_diff_sum_with_pos_neg_weights.py , https://github.com/sfu-cl-lab/Yeti-Thesis-Project/blob/master/Decision_Trees/LMT/calculate_individual_attribute_diff_times_weight.py, https://github.com/sfu-cl-lab/Yeti-Thesis-Project/blob/master/Decision_Trees/LMT/calculate_weighted_mean_in_leafNode.py
-
Store results in view
chao_draft.individual_player_01_02/07_08_check. Some fields in the view are explained as follows:
| Field | Details |
|---|---|
| player_diff_total_sum | ∑wi(sign='+')(xi- x̄) + ∑wi(sign='-')(xi- x̄), sum of all attributes difference compared to the mean value of each LeafNode |
| weight_pos_val | ∑wi(sign='+')(xi- x̄), sum of attributes difference with positive weights |
| weight_neg_val | ∑wi(sign='-')(xi- x̄), sum of attributes difference with negative weights |
| weighted_mean_in_LeafNode | ∑wx̄, sum of attributes mean value in each LeafNode |
| weighed_$attribute$_norm_diff | wattribute(xattribute - x̄attribute), e.g. weighed_DraftAge_norm_diff |
- Strongest player in strong LeafNodes(if the rate of GP_greater_than_0 > 50%)
| id | PlayerName | DraftYear | LeafNode | GP_greater_than_0 | player_diff_total_sum | lmt_prob | weight_pos_val | weight_neg_val |
|---|---|---|---|---|---|---|---|---|
| 8469455 | Jason Spezza | 2001 | 3 | yes | 2.573201971432897 | 0.9282017673222154 | 1.5149518724634254 | 1.0582500989694714 |
| 8469474 | Colby Armstrong | 2001 | 3 | yes | 2.3783673218444847 | 0.9140841746450182 | 1.733848766183595 | 0.6445185556608896 |
| 8469454 | Ilya Kovalchuk | 2001 | 3 | yes | 2.191565712215148 | 0.8982338589620239 | 1.3044223178129801 | 0.8871433944021676 |
| 8474040 | Sam Gagner | 2007 | 1 | yes | 4.0264360399140795 | 0.9969284587291163 | 3.607425146070099 | 0.4190108938439805 |
| 8474141 | Patrick Kane | 2007 | 1 | yes | 3.575510915857473 | 0.995186829189845 | 2.8495207008828474 | 0.7259902149746256 |
| 8474102 | David Perron | 2007 | 1 | yes | 3.2056610480408914 | 0.9930478269051884 | 3.150558241377556 | 0.05510280666333528 |
| 8474059 | Linus Omark | 2007 | 4 | yes | 1.3971131080309775 | 0.7998548932406279 | 0.3747145573109591 | 1.0223985507200184 |
| 8474114 | Oscar Moller | 2007 | 4 | yes | 1.390128255222534 | 0.7987343664858106 | 0.4388812508432919 | 0.951247004379242 |
| 8474627 | Jori Lehtera | 2008 | 4 | yes | 1.3543511587556547 | 0.7929214120046657 | 0.7584214819873075 | 0.5959296767683472 |
| 8474146 | Spencer Machacek | 2007 | 6 | yes | 1.8902042170864255 | 0.9900950944178386 | -0.021663440345763063 | 1.9118676574321887 |
| 9347 | Justin Azevedo | 2008 | 6 | no | 1.7695379029302973 | 0.9888389841358178 | 0.0913124006850809 | 1.6782255022452164 |
| 8474631 | Marc-Andre Bourdon | 2008 | 6 | yes | 1.6371760954445473 | 0.9872795553672469 | 0.16824162116048047 | 1.4689344742840669 |
- Some players have high weight_pos_val(∑wi(xi(sign='+')- x̄)) but lower player_diff_total_sum(∑wi(xi- x̄)). It implies they might be exceptional in some fields but also have obvious weakness in other fileds. e.g.
| id | PlayerName | DraftYear | LeafNode | GP_greater_than_0 | player_diff_total_sum | lmt_prob | weight_pos_val | weight_neg_val | weighed_DraftAge_norm_diff | weighed_Weight_norm_diff | weighed_CSS_rank_norm_diff | weighed_rs_P_norm_diff | weighed_country_EURO_diff | weighed_rs_GP_norm_diff | weighed_rs_PIM_norm_diff | weighed_rs_PlusMinus_norm_diff | weighed_po_A_norm_diff | weighed_po_P_norm_diff | weighed_po_PIM_norm_diff | weighed_country_CAN_diff |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 8469542 | Stephane Veilleux | 2001 | 3 | yes | 8.641748262585226 | 0.814777948388665 | 3.348573604539732 | -1.8534006668545049 | 0.0848315345122045 | -0.061360691633222 | -1.0909036201709423 | 0.8416343954816107 | -0.00243652510311 | 0.6469990348517232 | -0.06601310452150307 | 0.000000000000024159255961331284 | -0.6964839421620836 | 1.0981102334175368 | 0.5179770618781889 | 0.2228185611348 |
- Some players are ranked high in our prediction(in terms of wx_sum, lmt_prob, player_diff_total_sum) but have GP_greater_than_0 = 'no'. The reason can be that they are drafted but never signed a contract with any NHL team. Instead, they play in other competitive leagues. e.g.
| id | PlayerName | DraftYear | LeafNode | GP_greater_than_0 | player_diff_total_sum | lmt_prob | weight_pos_val | weight_neg_val | career league |
|---|---|---|---|---|---|---|---|---|---|
| 9347 | Justin Azevedo | 2008 | 6 | no | 1.7695379029302973 | 0.9888389841358178 | 0.0913124006850809 | 1.6782255022452164 | AHL/KHL |