Do takeaways and giveaways matter?

[studentized 166]

A couple weeks back I made a fairly simple post tracking giveaways and takeaways in the VHL (see here). I cherry picked some of the leaders in some categories I thought were interesting, but that didn't really tell the whole picture. So to start off this media spot, I thought I'd include a full individual leaderboard below. I've haven't been successful getting this html to embed in one of these forum posts so I've ended up just adding it as an attachment. Let me know if its broken for anyone, but it should be viewable in just about any browser.

 

takeaways-leaderboard.html

 

A couple of good points were brought up to me regarding that past article. The one that stuck with me the most is the whole "Why should we care?" aspect. Are giveaways and takeaways even worth tracking in the VHL? Do more takeaways make a better player? More giveaways a worse one? More importantly, how do they affect wins at a team level? To start answering this last question, let's just look at the team rankings below combined with their place in the standings.

 

Team	GP	W	L	OTL	P	Takeaways	Giveaways	Diff
Malmo	72	46	20	6	98	9803	9541	262
Riga	72	46	22	4	96	9701	9679	22
Helsinki	72	45	22	5	95	9797	9851	-54
Seattle	72	43	24	5	91	10051	10036	15
HC Davos	72	40	22	10	90	9852	9732	120
Moscow	72	40	30	2	82	9701	9782	-81
Vancouver	72	35	28	9	79	10030	9974	56
Calgary	72	35	32	5	75	9658	9602	56
Toronto	72	29	29	14	72	9930	9930	0
New York	72	30	35	7	67	10048	10135	-87
Prague	72	23	40	9	55	10058	10083	-25
D.C.	72	20	49	3	43	9349	9633	-284

From looking at the above you can argue that some relationship probably does exist, but it's not clear how strong or how important that relationship is. The top and the bottom of the standings happen to be the best and the worst in turnover differential, but the middle is not as nice. So how should we look at his? One thing we could try is to compare the accuracy that turnover differential has on predicting wins to the accuracy of another, more common stat. It's widely believed that the team that shoots the puck more, scores more (and so wins more), so shot differential can serve as a nice comparison stat. In S68, the team that won the shots differential won the game 66.1% of the time. I wanted to compare this same metric to teams that won the turnover differential. Turns out that teams that won the turnover differential won only 52.1% of the games. At first glance, this feels low but the more you think about, the more it makes sense. Keep in mind that the more a team possesses a puck, the more chances for giveaways (and less chances for takeaways) they have. Considering the extreme case where one team possesses the puck for close to the entireity of the game, said team would almost certainly win but also lose the turnover battle. So a better metric would measure "takeaways per some unit of possession allowed". The closest metric measuring puck possession that I can use is shots. With this new metric of "takeaways per shot allowed", we get that teams winning this differential end up winning the game 65.5% of the time.

 

One thing lost in both these stats is how badly a team won/lost the turnover differential. A single more takeaway than giveaway might not sway a game much, but 20 more might. This is represented in the data above; when looking at all the games where the team won the turnover battle (as defined by my new stat) but still ended up losing the game, we find that the average turnover difference is 20% smaller than normal. This supports the notion that games that are closer in turnover differential, often end up going either way in terms of the final result. One good way to actually model this would be with a linear logistic regression which would try to model how the magnitude of turnover difference affects the probability of winning. I might end up doing this another week, but for now I'm out of words to write.

 

608 words