03-09-2013, 12:59 PM
#964
overpass
Registered User

Join Date: Jun 2007
Posts: 3,517
vCash: 500
Quote:
 Originally Posted by Hawkey Town 18 I think the Iginla factor is something that should be considered. Does anyone have an idea on how much Conroy should be discounted here? I have taken a look at those seasons and in all but one Iginla outscored Conroy by a good amount at even strength. The one year where they were close was 2003, where Conroy had 53 adjusted ESP and Iginla 58 in 79 and 75 games respectively.
OK, I'm going to post a new statistical method I just tried out. This may be completely out to lunch but I'll just throw it out there. Feel free to ignore if it doesn't make any sense.

Let's look at Conroy's prime (1997-98 through 2008-09). Oddly enough he played every single season with either Jarome Iginla or Pavol Demitra at a time when both were among the best scorers in the league. I'm going to see how well Conroy's scoring tracks the % of time he played with them, and try to find the bump in his scoring that could be attributed to that.

Major assumption: % of even strength points scored with Iginla or Demitra is the independent variable and overall even strength points (adjusted and per game) is the dependent variable. Meaning that if Conroy plays more with either Iginla or Demitra at even strength his scoring will increase. Of course the reverse could be the case - if his scoring increases he will play more with Iginla or Demitra. I'll ignore that for now.

How much did Conroy play with them? I'll estimate that by looking at the percentage of Conroy's even strength points in which either Iginla or Demitra also received a point. This will be shown as Iginla/Demitra %

 Season \$ESP GP \$ESP/82 Iginla/Demitra % 1998 51 81 52 3% 1999 48 69 57 8% 2000 27 79 28 18% 2001 32 83 32 4% 2002 62 81 63 66% 2003 53 79 55 63% 2004 40 63 52 45% 2006 51 78 54 41% 2007 33 80 34 46% 2008 39 79 41 43% 2009 57 82 57 40%

Then I'll run a regression in which Iginla/Demitra% is the independent variable and \$ESP/82 is the dependent variable.

Result: Craig Conroy's \$ESP/82 for a given season can be predicted by the equation 40.5 + 20.9*(Iginla/Demitra%)

So 40.5 \$ESP/82 is the baseline level for Conroy over this time if he doesn't play at all with Iginla or Demitra. If they share in 100% of his EV points then we'd expect him to score 61.4 \$ESP/82. For all actual seasons we will predict something in between.

Let's add back in the actual seasonal variation for Conroy, but subtract the "Iginla/Demitra factor" and see what it looks like.

 Season \$ESP GP \$ESP/82 Iginla/Demitra % Predicted \$ESP/82 (given Iginla/Demitra%) \$ESP/82 minus Iginla/Demitra factor 1998 51 81 52 3% 41 51 1999 48 69 57 8% 42 55 2000 27 79 28 18% 44 24 2001 32 83 32 4% 41 31 2002 62 81 63 66% 54 49 2003 53 79 55 63% 54 42 2004 40 63 52 45% 50 43 2006 51 78 54 41% 49 45 2007 33 80 34 46% 50 24 2008 39 79 41 43% 49 32 2009 57 82 57 40% 49 49

Basically I used the prediction equation to predict \$ESP/82 for Conroy in each season, based on his Iginla/Demitra%. I took the difference between that prediction and 40.5 (his baseline level) and subtracted it from his actual \$ESP/82. This gives the result shown in the rightmost column - an estimate of Conroy's \$ESP performance with the Iginla/Demitra factor removed.

This method discounts Conroy's production by up to 20%, which may be a bit high. Subjectively I would have said 10% or maybe 15% in the top season before I ran these numbers.

Quote:
 Originally Posted by Sturminator edit: Kesler and Richards have only three seasons that are good scoring years by ATD standards, and Bergeron has four. The biggest issue here is not Iginla's influence, but their relative lack of meaningful prime seasons.
Looking beyond just "scoring years" to general contribution, I think they each have five meaningful seasons (working on their sixth, or 5.5th, or whatever.) Conroy has over twice that, which is definitely an advantage in his favour and very possibly enough to give him the edge overall. But I don't think he has any kind of peak advantage.