Comments on: College Basketball’s Unmentionable: the Issue of Point Shaving?

By: David Hess

David Hess — Tue, 21 Sep 2010 19:05:44 +0000

I have a couple more quibbles:

1) This is related to Ben’s point about the EMH assumption. I don’t see why line shading cannot cause a double peak. There are a variety of types of teams that will draw more bettors than usual, not just favorites. For example: teams on winning streaks, road favorites, big name schools, teams from BCS conferences (when playing minnows). Having just listed all those, I now see that MOST of them should be favorites, so I guess the shading should GENERALLY be in the same direction. But not always. If we have overlapping distributions of shaded-down, shaded-up, and not-shaded, couldn’t it be reasonable to see a multimodal distribution?

2) I’m not sure if this one is valid, but intuitively it’s an issue to me. Their sample size for the 19+ point spread postseason games is 99. As someone who has gambled on sports in the past, and seen results/trends fluctuate from one year to the next (particularly I’m thinking of data derived from the NFL, which has 296 games each year), this seems a dangerously small sample from which to draw a conclusion. This is especially true when their conclusion is essentially “despite the fact that you can visually see the same double-peak pattern in the postseason data, the significance test says it’s not really there.” Aren’t they claiming that the pattern is caused by chance in the postseason, but by point-shaving in the regular season? Maybe I am misunderstanding or misinterpreting this section.

3) I guess this is just another restatement of the EMH assumption problem, but they seem to be claiming that any deviation from their assumptions that they haven’t explained must be due to point-shaving. I just thought of this, and no time to go back and re-read the article, but did they account for simple variables like home/away, conference/nonconference in their analysis? I don’t recall seeing that.

That said, I am not arguing that there is NOT point-shaving. I am just not sure that their evidence shows that there IS.

By: rtmsf

rtmsf — Tue, 14 Sep 2010 01:33:53 +0000

Doesn’t sound stupid to me, Ben. I think the authors will check in on the post, but I’ll make sure they see your comment separately. Thanks.

By: Ben (Dear Old UVa)

Ben (Dear Old UVa) — Mon, 13 Sep 2010 17:43:23 +0000

I know this sounds stupid, if I were reviewing this for a journal, I’d ask to see well-reasoned argument as to why I’d expect sports betting to be an efficient market. Are there betting volume numbers to back that up?

There are multiple EMHs (something about strong and weak, but I can’t recall the distinction), does this assume one versus the other? etc. etc.

The authors give this part of the paper the short shrift.

By: rtmsf

rtmsf — Mon, 13 Sep 2010 16:21:25 +0000

In reply to Ben (Dear Old UVa). Ben, given your background and experience, I'm sure that's a valid criticism. What could the authors do to mitigate or eliminate that concern, if anything?

By: Ben (Dear Old UVa)

Ben (Dear Old UVa) — Mon, 13 Sep 2010 12:39:10 +0000

Hi guys –

Fascinating work here.

My one quibble with the work not mentioned in the above article is this: all of this relies heavily on the efficient market hypothesis. Financial economists have long relied on the EMH for analyzing stock returns. However, there are many economists that have disputed whether or not the EMH actually exists for stocks.

My question is this: if we can’t necessarily believe that the EMH holds for something relatively efficient as the stock market, how can we assume that sports betting is an efficient market?

I’m not saying that I don’t believe their findings or that there isn’t point shaving in college bball. I just don’t think the authors have done a sufficient job of convincing me that sports betting is 100% efficient market.