Wednesday, 29 September 2010

Drawing Hope

The draw debate continues, although it is spilling over into a broader criticism of Peter Nordsted’s book Premier Football Betting Handbook 2010/2011 which I haven’t actually read, so I can’t really make any comment about.

I do find it interesting though that some people are missing the obvious, which is that however he finds them, these draw selections of Peter's are value if backing them is consistently profitable, which his method appears to be.

When it comes to the draw price in matches “involving vaguely similarly matched teams” the price is in a fairly narrow range, with slight differences between leagues, so if the Elo ratings continue to predict draws at anywhere close to their current strike rate (37.84%), then a draw price in the 3.3 to 3.5 range is clearly huge value.

It is suggested that “If there has been value on these selections over this long period of time then you can be (fairly) sure the market will be, and has been, reacting”. Doesn't this rather presume that the market would somehow know what my selections were? Only my spreadsheet knows, and it only tells me when I enter in the appropriate numbers each week. Draws will continue to hit overall at the rate that they have hit over many years. Generally speaking, the price will not move unless the overall rate of draws increases for some reason.

Also, the draw price is arguably more stable than prices on individual teams, and is almost a pivot around which the home and away prices adjust. It's also been said that the draw is unpopular with backers, and although that may have been the case a few years back, I doubt that it is true now, especially on the exchanges. The percentage of drawn games has remained constant over the years, but even if the draw price on my spreadsheet’s selections were to drop, it would need to be a significant drop in price to make it a losing proposition.

Anonymous says "Your method of selecting draws is never going to select a draw in a "mismatch" or anything approaching a mismatch".

This is, of course, correct. Any objective ratings system that predicted a draw between two mismatched teams, after accounting for form and home advantage, would be one strange rating system. Maths doesn't work like that.

But having said that, by comparing the expected margin of victory with the prices available, the spreadsheet does often suggest that an expected winner isn’t value, so while not actually predicting the draw, we can infer that the draw and underdog prices are value, and lay the favourite. There is more than one way to skin a cat.

Looking ahead to tonight's big games in the Conference, and there is one draw predicted - Eastbourne Borough v Rushden and Diamonds at 3.55. Draw predictions in this league are hitting at exactly 33.33% this young season, up from last year hopefully the result of more accurate ratings!


Anonymous said...

No, the market doesn't need to know your selections. If the price on a certain outcome is consistently incorrect, which you infer by the insistence your picks offer value, then you can expect the market to "correct" over a course of time (the old chestnut of market efficiency, or a form of).

It's an almighty IF to suggest your picks will continue to return 37% or so. Historically speaking, it's almost certain they won't. I don't know your sample size, but that figure seems to be based on just part of last season in which case the sample size is, of course, insignificant.

A big factor in draw expectations is obviously the total goals expected. I don't know for sure, but I suspect your projections fail to take this into account. The lower the goals the more likely the draw - obviously.

A salient point you make is "Draws will continue to hit overall at the rate that they have hit over many years". Yes, pretty much true, unless a big factor in the game were to change (e.g. extra points for goals scored or no points for a draw (ain't going to happen but I remember in the 1980s when the equivalent of the Conference, the Gola League I think it was called, offered 3 points for an away win and 2 points for a home win which should have changed the dynamics of a game. As did the introduction of three points for a win - which saw a pronounced increase in home advantage and, probably (haven't checked) an increase in goals and hence decrease in draws. Anyway, the point I was going to make is that the draw should indeed continue to hit at the same (well, similar) rate. And that rate sure as hell ain't 37%, even in the type of match you are predicting the draw in.

It's almost unthinkable that you can consistently hit at a % equivalent to a price of 2.66 on an outcome generally priced in the 3.3 - 3.5 range you mention. Sorry, but the market is never going to be that wrong in the modern age.

Now, I'm not saying the draw isn't ever value. I think it is. But rarely is it the enormous value you make out. There are certain match situations - not confined to dodgy Italian results - which contribute to a likely (relatively) higher probability of a draw. But, in the main, this would probably involved getting a price of 3.25 or so on a 3.00 shot. Nice value for sure.

An important point which you haven't addressed is how you price the draw when your system predicts a draw. You can't simply back an outcome blind without looking at the price and, to the best of my knowledge, your system has shown no price sensitivity whatsoever.

Hope that helps.

Anonymous said...

One error (at least)- apologies. "Sorry, but the market is never going to be that wrong in the modern age"

Not strictly true. But certainly not on a consistent basis.

Anonymous said...

Dont bother reading the book its a sales pitch for his service.