[golfore]

i've recently been trying to bet golf head-to-heads and have been swimming upstream, partly because i'm a hack, but mostly because i'm struggling to set my own lines. oh, and also: because i have no idea what i'm doing. does anybody here bet golf, and if so, any tips for what to look for? of particular interest to me is how to convert data sets (even something simple like sagarin) to money lines.

for instance (link)...

jason dufner 68.66
zach johnson 69.66

if i have golfer X as 1 shot better per round than golfer Y, how might this translate into a money line price?

thanks in advance for the help.

[Gradius]

I'm not sure if it works in golf, but you could try to use the Pythagorean theorem.

Dufner: 68.66^2 = 4714.2
Johnson: 69.66^2 = 4852.5

4852.5/9566.7 = 50.7%, Moneyline = -103

Dufner should be the favorite of a ML of -103, since he has a lower score than Johnson, according to the Pythagorean theorem. Although, I'm not as familiar with golf; you should probably wait for someone else more qualified to post here.

[golfore]

Thanks for the effort, Gradius. I appreciate it.

I don't know if that would work or not. I'm assuming it can't be that easy, but yeah, hopefully someone with a bit more golf expertise can set me straight.

[brettd]

Determine win % empirically through regression.

IE: How often does a predicted one shot favorite beat his opponent? Do this for a differential of the full spectrum in shot difference, create a distribution, and then draw an equation from it.

[golfore]

In theory, how would you price a player who is a 1 shot favorite (72 holes) against his opponent?

So, if Player X is a 0.25 shot favorite per round over Player Y ... i.e., a 1 shot favorite over 72 holes ... what would that correspond to on the money line, as a general rule?

Example (72 holes)...
1 shot favorite = -125
2 shot favorite = -150
3 shot favorite = -175
Etc.

[golfore]

I did find this very useful thread: http://www.sportsbookreview.com/forum/handicappe...rediction.html

Some of the concepts are still a little too advanced for me (damn, I should have taken a few more math courses in grad school!), but hoping that my basic question - "how to create a line?" - is easily answerable.

[bubblebuttluv]

I just bet Phil Mickelson to f.u.c.k. up constantly, and I like Phil.

[golfore]

I just bet Phil Mickelson to f.u.c.k. up constantly, and I like Phil.

[oddsfellow]

but hoping that my basic question - "how to create a line?" - is easily answerable.

Not sure how basic your knowledge is but if you are going to create lines you are going to have learn some statistics and functions as its not easily answerable else. In your example assume Jason Duffner is one shot better than Zach Johnson over 18 holes. However lets just say that Duffner is capable of shooting anything between 62 and 74 with the mean at 68. Johnson on the other hand is far more consistent but does not score so low. His spread is 65 to 73 with a mean at 69. Then you can compile a matrix showing all the different combinations of score that are feasible. For example Johnson shoots 67 Duffner shoots 68 Johnson wins, Johnson shoots 67 Duffner shoots 66 Duffner wins. However as the mean is 68 and 69 there will be more scores in and around that area than at the ends of the spectrums. What you need is a means of simulating all these scores. The spread of scores is called the standard deviation which if you can calaculate that and know the mean you can do whats called a monte carlo simulation (google it) that will simulate the two scores over thousands of times and produce a percentage chance of each player winning or a tie. There is even a free excel tool for it. That i am afraid is about as easily answerable as it gets.

[golfore]

Not sure how basic your knowledge is but if you are going to create lines you are going to have learn some statistics and functions as its not easily answerable else. In your example assume Jason Duffner is one shot better than Zach Johnson over 18 holes. However lets just say that Duffner is capable of shooting anything between 62 and 74 with the mean at 68. Johnson on the other hand is far more consistent but does not score so low. His spread is 65 to 73 with a mean at 69. Then you can compile a matrix showing all the different combinations of score that are feasible. For example Johnson shoots 67 Duffner shoots 68 Johnson wins, Johnson shoots 67 Duffner shoots 66 Duffner wins. However as the mean is 68 and 69 there will be more scores in and around that area than at the ends of the spectrums. What you need is a means of simulating all these scores. The spread of scores is called the standard deviation which if you can calaculate that and know the mean you can do whats called a monte carlo simulation (google it) that will simulate the two scores over thousands of times and produce a percentage chance of each player winning or a tie. There is even a free excel tool for it. That i am afraid is about as easily answerable as it gets.

Great tips. This is very helpful. Will look closer into what you suggested and possibly ask a few follow ups at another time.

Does anybody here bet golf H2H's?

[evo34]

In theory, how would you price a player who is a 1 shot favorite (72 holes) against his opponent?

So, if Player X is a 0.25 shot favorite per round over Player Y ... i.e., a 1 shot favorite over 72 holes ... what would that correspond to on the money line, as a general rule?

Example (72 holes)...
1 shot favorite = -125
2 shot favorite = -150
3 shot favorite = -175
Etc.

For a tournament matchup, each 0.1 strokes per round of expected advantage will correspond to ~5 cents of money line value. E.g., a golfer projected to score 69.8 per round should be -110 (no vig) or so to beat a 70.0 golfer for the tournament. Note: I haven't bet golf in over five years, so don't take this as gospel. But it's what I calculated in my old spreadsheet.

One other finding: the best lookback period to use for rating golfers was 20 events. That is, if you had to look only at every golfer's last x events to handicap a tourney, the average error was lowest when using last 20. Obviously to refine that, you'd want to weight each event according to how long ago it was, and customize it a bit for individual golfers who are more momentum/reverting than normal.

[golfore]

Yeah, figuring out how to make all the adjustments - i.e., changing courses, conditions, fields, etc., is what I'm struggling with most. There's a lot to absorb, obviously, I'm but hoping to at least lay a foundation with some basics that will get me headed in the right direction.

Again, thanks to those who have responded.

[oddsfellow]

This tool may help http://www.bmandmsolutions.co.uk/gol...em/predict.asp it spits out European odds so you will need to convert them but you can change the weightings and create your own ratings.

[thedirtyscreech]

Where did you get your data? I've been looking for a good place to scrape information on tournaments, courses, player results, etc.

[floridagolfer]

I might be wrong, but I don't think those Sagarin numbers necessarily equate to X strokes per round.

[thedirtyscreech]

I might be wrong, but I don't think those Sagarin numbers necessarily equate to X strokes per round.

That's 100% correct. It is a combination of win/loss/tie record, stroke differential, and connections to other players (how well each did on the same course on the same day, etc.). Players need at least 10 events in the past year to qualify to be rated, and it's an attempt to help quantify what the likely outcome for each player would be if they played the same imaginary course on the same imaginary day.

To answer OP's question of how to convert these to the probability of a given player winning in a heads-up scenario is a bit trickier. Using the pythagorean theorem method is possible, but you'd likely need to run a regression to figure out the correct exponent. Another method would be to take a more frequentist approach to it, and compare what the outcome has historically been when the stroke differential is 0.5, 1, 1.5, etc. Thus, if the player with a 1-stroke rating advantage wins 60% of the time, it's easy to transform that into a proper moneyline. Of course, you'd need a player's historical sagarin rating going into each tournament to do this, and I'm not sure where one would get this. Pythagorean would be more attainable if this info isn't available. There are other methods similar to the pythagorean approach that could conceivably get you something workable, but ultimately, I'd be more inclined to trust the historical outcomes (provided your sample size is large enough).

Golfore, where do you get your historical data? I've been looking to start golf handicapping, but am dreading the annoying task of merging multiple sources of data when scraping. Invariably, sites always display names differently ("T. Woods", "Woods, Tiger", "Woods, T.", "Woods T", etc.), and given the amount of professional golfers on all the worldwide tours, it seems there's bound to be naming collisions (multiple "Smith, J" for instance).

[mr.ed]

You might want to take a look look at tour-tips.com. Some of it is free...other stuff you have to pay for.

Lines have really sharpened the last few years....and you will have your work cut out for you if you think you can beat them with some type of model. A model might give you a base to work off of, but you will need to do a lot of massaging of the numbers to find value. But the tour-tips site will give you numbers on a lot of variables, such as wind/weather, and how pros play when they are in and out of contention.