Thursday 1 July 2010

Unreasonable Expectation


A few comments to address.

First, on the subject of Slicer and his Holy Grail, Brian Bee wrote: Maybe I imagined it but I'm sure around the time of the thread there was a few matches that would fit the criteria whereby the FT 0-0 would be so high and the half time 0-0 not high enough so that come half time you could green up the full time 0-0 to cover your lay loss on the half time market thus giving you a risk free punt on their being a goal in the 1st half. However once it was realised the FT 0-0 would get backed in so much the edge vanished.

The vanishing edge, if it really existed, is just one problem with this Holy Grail. The market is efficient and will soon correct if any price in a market is out of line. In the scenario outlined above, there would never be any need for the third bet anyway, and something as obvious as that is unlikely to have ever existed. It's not uncommon for the price of the same event to be slightly different in separate markets (Correct Score 0-0, No Next Goal, No First Goalscorer, Total Goals < 1 etc.) but it's usually only briefly and rarely for a big enough differential to make it a viable proposition.

Anonymous, but a reasonably eloquent one and only slightly patronising, so I will address the comment, had something to say on the R16 results and my Quarter-Final write-up.

Whilst I don't agree at all with your reasoning on the penalty shootout, there probably was some small value on the 12-1 shots.

A far better way of assessing the true odds (compared to your over-simplistic method based on a limited sample size) is as follows:

If, say, the draws were around 30% in each match - which is probably roughly in keeping with betfair prices - then approx 60% will then go to penalties (read that somewhere, may have been here?) then around 18% can be expected to go to pens. So 9% for either team if shootout assumed to be 50-50. Therefore some value at 12-1

NB - I am assuming the 60% is correct but haven't checked it. First impression is it seems a little high but on reflection it's probably about right given the pressure in extra-time and a probably reluctance to attack as much as usual.

Your stats on the tightness of 1/4 final games make interesting reading given the huge difference in draws and 0-0s compared to "normal" matches. If under 2.5 goals is 58% then the "normal" draw %, assuming equally matched teams, would be in the region of 30%. The 45% you suggest being the "true" price is way too high. But, 30% is perhaps a little too low due to "abnormal" factors in the games and hence there may well be some value in backing the draw.

(and if the draw is greater than 30% in the 1/4s it can reasonably be expected to be higher in last 16 as well. Which would mean there was greater value on the 12-1 than suggested in the earlier explanation)
My 'over-simplistic method based on a limited sample size' isn't from choice. There is only a small amount of data to work with when investing on World Cups, but what data there is appears to be solid, at least in the knock-out stages. We can either while about the lack of data, or make use of the limited data we have. At least teams need to advance and can be relied upon to do their best to do that.

The statement that "if the draw is greater than 30% in the 1/4s it can reasonably be expected to be higher in last 16 as well" is totally incorrect however. In the R16, the probability of a draw is 29%, but the careless statement does help to illustrate the importance of drilling down into the data rather than simply 'reasonably assume' that data will be consistent across rounds. It isn't. We shouldn't expect it to be.

I did try to highlight previously how the knock-out rounds are all a little different. While the draws at the R16 stage are 'only' 29%, probably due in part at least to some relative mismatches with weaker teams making it out of the group stage and facing one of the big boys, the probability of a draw in the Quarter-Finals is much stronger, and the 'over-simplistic reasonable expectation' that the R16 should reflect the QF stage in draws is very wrong.

The two are close however, when it comes to the Under / Over 2.5 markets.

The HT 0-0 is also, unsurprisingly, much more likely in the Quarter-Finals - 29% v 11% in the R16, so if anyone is thinking of trying Slicer's bet on this round, good luck.

I might also mention that only two matches at the QF stage have been won by more than one goal. Italy beat a weak Ukraine team in 2006 by 3-0 and Croatia beat Germany by the same score in 1998.

Then Googly asks:
I was just wondering by using your rating systems exactly how you came up with the fact that the difference in rating, equates to 1/2 or more goals?
My "superiority predictions" are based on the result needed for the Elo ratings of the two teams to remain unchanged, so usually the result needed is not a whole number, but a fraction. One of the improvements I'll be working on this summer is a more accurate measure of superiority. Last season I lumped the +1.5 teams into the same bracket as the +2.49 teams which is clearly not ideal. Hard to believe the new season is just around the corner!

6 comments:

Anonymous said...

What total drivel about the draws in the last 16 and 1/4 finals.

You openly admit your sample size is too small but is all you have to work on.

The comment about reasonably expecting draws to be higher than "normal" in the last 16 as well as 1/4 finals is absolutely fine. It's relating to teams' fear of failure not any of your insiginifcant sample analysis. Your 29% means nothing, nor your 45% - which is way out.

I look forward to discovering that 70% of all semi-finals are draws.

Anonymous said...

Forgot to say. You are making an incredible amount of assumptions and stating them as fact in your World Cup musings. This, despite admitting you don't have enough data.

e.g. when dismissing a different statement you state "In the R16, the probability of a draw is 29%". Why? Because 29% of your small sample have finished in a draw? And you have the balls to use this statement to state that something else, suggested to be able to be reasonably assumed, is "totally incorrect".

Brian Bee said...

The 3rd bet in that scenario was the "lay" bet of the FT 0-0.

Cassini said...

East London Anonymous – Read what I actually write, even if you think it’s ‘total drivel’. I KNOW the sample size is small. We all know it is small, but as I have written at least twice now, there’s nothing we can do about that. We have to work with what little we have. To me, the data is not ‘insignificant’, there just isn’t much of it, but there is no evidence that the data we have is not reliable. If you don’t want to take advantage of my observations, then don’t, but the fact that you keep reading and commenting tells me you ARE rather interested, so either stop with the negativity and childish comments and try to add something to the discussion for once or stay away.

Brian - I believe Slicer claimed that Bet 3 consisted of a tweak of Bet 2 (presumably laying off all or part of it) but some other bet as well, all made at HT, but then other times he said the bet was made after 70 mins. It's all nonsense anyway, for the reasons explained previously.

Anonymous said...

It's not the only data you have to work with. It's the only World Cup data you have to work with. There's no need to base everything purely on World Cup games. If you added other tournaments of similar nature then you may be able to draw some more useful conclusions.

Cassini said...

The only similar championship is the Euro tournaments, and you will find those results hold up well when you include those too. However, any tournament that is held every two / four years is always going to be short on data.