Golf finishing odds stats

[AussieVamp2]

Ok, dug up the file I was working on, had not finished it, had looked at a couple of years from CorrectOdds - note that they do not always pick all the place getters, so odds not always listed.

Not too big a sample yet, so the odd 200/1 can make things whacky. More data points in 1/2 as you would expect as well.

So from 98/99 in Europe

Median prices

1 25
2 40
3 40
4 50
5 40

(includes ties for places where they had odds)

Average prices

1 42
2 50
3 61
4 62
5 56

longest priced winner in that time was some British Open or other, 6 100/1 winners though, and that was with some winner prices missing....

when get time will have to go through and look at some more..... Not that interesting perhaps, but anyway.

 

[Stanley]

Continuing on from the discussion in the Scottish Open thread ...

Actually, the bit about the uniform distribution is not important for my figures in that thread, other than they are based on the player finishing 1st 20% of the time they finish in the top-5. Because 1/4 odds are paid on 2nd, 3rd, 4th and 5th place finishes alike (assuming no ties), then it makes no difference how the probability distribution is shaped for those places.

For Garcia, it is an appropriate figure. I use 3-years as the appropriate time span for relevant player stats on my site. So, in the past 3 years, Garcia has finished in the top-5 on 18 occasions and won 4 times.

It brings me on to another point which is much more important. When should the ew bet be placed and when should the win-only bet be placed. Regardless of odds, it should differ greatly among players. Mine is a simple rule-of-thumb (read "lazy") approach that simply says play e/w when double figure odds are available. That approximates to a 35% floor (for odds of 10/1) for finishing in the top-5 as outlined in the Scottish Open thread and provided I think the player has at least that much chance of finishing in the top-5 I will make the bet else they will be passed over.

Okay, my record on outright plays suggests that it is profitable to do so, but could it be better with more homework?

Looking at my selections, for the past three years Garcia has won 4 times from 18 top-5 finishes (22%); Lehman has won 1 time from 11 top-5 finishes (9%); Parnevik has won 5 times from 13 top-5 finishes (38%).

It should be clear that an e/w bet is essential to betting on Lehman, but that a win-only bet would be far better for Parnevik.

It is possible to provide a formula to differentiate between when it is profitable (based on historical stats) to bet win-only and when to beat e/w.

Assume x are the odds available on a player. So x=10 when 10/1 are the odds.

Assume y is the % wins from top-5 finishes. From the figures above, y=0.22 for Garcia, y=0.09 for Lehman and y=0.38 for Parnevik.

Assume, as with all my outright plays, that each outright playis 1 unit. That is, 1 unit on a win-only bet or for an e/w bet, that is 1/2 unit on each of the win and place components.

The profit from a win on a win-only bet is x

The profit from a win on a e/w bet is x/2 + x/8

The profit from a place-only win on a e/w bet is x/8 - 1/2

We therefore need to find the value of y such that it is better to place an e/w bet rather than a win-only bet.

So the formula is

y(x/2 + x/8) + (1 - y)(x/8 - 1/2) > yx

which reduces to

(x/4 - 1)/(x - 1) > y


For odds of 10/1 (x=10), if the probability of winning given a top-5 place is less than 1/6, then the e/w bet is appropriate. If it is greater than 1/6 then the win-only bet is appropriate. For odds of 16/1 (x=16), then if y > 0.20 then the win-only bet is more appropriate.


Of course, this is still no more than a rule of thumb. Changes in the quality of field across the 3-years, the choice of sample period, current form, course form and the rest will affect the appropriateness of this formula.

Which brings me back to the original point. Increased sophistication of modelling to make it more player-specific does not necessarily make the result any more reliable. Simple rule-of-thumb models can work just as well and they tend to be more robust.

Sounds like you have the beginnings of an interesting dataset there AV to look at this. It is flawed as you say, but if you did want to look at this more closely, then I have read a number of academic articles, including some for golf, where the authors have been able to get historical odds from Intertops for their dataset

 

[doonhamer]

I know you guys all have your own way of picking your selections am I the only one that makes my selections through what I have seen on Sky golf coverage ,what I have remembered through years of watching golf ,and a little research into the players and tournament statistics you seem to be making things very complicated Doonhamer

 

[AussieVamp2]

Stan,

Always interested in academic articles on sports stats, if you have references you can pass along, or even better urls/pdfs or whatever, that would be cool.

I was thinking more along a 'generic model' than a player specific one cf Parnevik et al., which has the problem as you say of courses, small samples, etc., etc. One bookmaker at least was being nice to an academic, good for thme.

Field size another interesting issue I guess.

 

[AussieVamp2]

Doonhammer, don't watch golf, even if I did, not on tv except locally and the odd major at horrible times. So that would not help me too much.

 

[Stanley]

Primary journal for these articles is 'Perceptual and motor skills' which is a psychology journal but deals with sports a great deal. Will dig out some more articles for you next week - I won't be in the office tomorrow. Don't know of any URLs or pdfs at the moment, but again I'll dig around next week.

Doonhamer, I agree with you. I was trying to show that increasingly "sophisticated" models look good, but really cannot offer a great deal more than the simple rule-of-the-model guidelines which I use to make my plays

 

[AussieVamp2]

geez, days off, I dunno

Thanks!