Golf Editorial

20130306 The Rise & Fall of a Golf Prediction Software


Picture this. It’s 2002, in Alice Springs. Not much to do in the evenings after long days at the betting factory. You could only play some sport, watch tv, surf the ‘net, go out and get drunk and / or try to score with a European backpacker.

Having had my fill of beer and backpackers in earlier years, my thoughts turned nightly to something completely different, golf pricing and, in particular, discussions I’d had with a mathematician friend who was also a frustrated golf punter like me.

‘It’s all about scoring versus par’ he’d said. ‘All I need sufficient data points for each player and a few executive decisions on how to age data, and I can run this in Excel. I can enshrine in a simple program everything we’ve discussed!’

So, on those quiet and lonely nights I’d download scores, thousands of thousands of them each night from PGA Tour events, and laboriously paste them into my newly-discovered MS Access database along with course notes on things like: distance, par, any left-to-right or right-to-left bias, grass used on greens, etc.

On the search for the Holy Grail; a golf betting prediction software

The Brilliant Logic

My friend’s translation of our golf scores logic into a simulator program took every player’s career rounds on the type of course that’s next being used on the PGA Tour (eg bermuda greens, can be windy, bombers track).

Then he’d run the program and ‘play’ the golf tournament in Excel 20,000 times and average the outcomes; the outcomes being: 1st & 2nd round scores for the field followed by a cut and then 3rd & 4th round scores for those who’d made the cut.

He reckoned this high number of simulations would give us valid data in the form of a set of theoretical scores for every player – a final scoreboard!

Then we’d be able to compare these theoretical finishing positions with the betting market on the tournament to identify players who the simulation predicted would perform significantly better than their prices suggested.

The First Simulation

Our laptop was not exceptionally powerful and the first simulation was only part way through by the time the tournament teed-off! We’d started on the Wednesday.

So the next week, freshly armed with a field for the PGA Tour’s HP Classic of New Orleans, we started our simulation on Tuesday morning. If finally finished late Wednesday night to the delight of two anxious golf punters hovering like overly concerned new parents!

Simulation Outcome

From memory, Davis Love was the 10/1 favourite and he’d won a more or less expected 1,900 or so of the simulated tournaments – meaning his price was correct. And so on, and so on. Then paydirt!

The exciting (orgasmic even!) outcome was that one player really stood out. He had also won almost 2,000 ‘tournaments’ but his market price was 50/1. 50/1 when he should be 10/1!

We stayed up all night trying to find a flaw in our logic but there was none! It was perfect, unimpeachable, brilliant! We had to back this Steve Flesch guy.

The Betting

I opened with $50 each way with StanJames. My friend got the same price with Pinnacle. We also bet with VictorChandler, Sportingbet and gave our modest credit cards a pounding. We had no real plan but took mostly outright win bets and a few each ways. Big money for us, though small potatoes to a professional gambler.

More significantly, as all tipsters will appreciate, we tipped Flesch to friends, family and work colleagues and many backed him, albeit somewhat dubious of our enthusiasm for our simulated output!

The Tournament

Flesch was a long way back after two rounds but scored 65 on Saturday to keep hope alive though he was still 7 shots back, an almost hopeless winning chance. Still a place possibility though, to at least reward the place component of our each way bets?

But then, miracle of miracles, another 65 on Sunday and into a playoff with Bob Estes! Oh, the drama! Flesch had never won; could he?

Well, he won that playoff! Vindication! New careers as professional gamblers! The most brilliant golf prediction tool on the planet! Man, we were full of ourselves; and pretty flush with cash too. And all those who’d been worn down by our rantings and had backed Flesch were pretty happy too.

The Next Week

Same scenario. A PGA tournament, Quail Hollow, massive pressure on the old laptop running for 36 hours! And finally, the output, the good stuff. An outcome we liked the look of as well, featuring just one quadruple overlay. A 15/1 shot paying 60/1 with the bookies! Oh joy!

Then it was a simple case of getting into the betting, and we got stuck in like a wolfpack descending on an injured animal. But our man missed the cut!

The Ensuing Weeks

Flushed with success, I’d invested nearly $4,000 in week two. I was a god; I didn’t need a staking plan!

Week three was a little more cautious at around $3,000 but approached very much in the belief that week two’s missed cut was simply a statistical anomaly, a minor bump on the road to success.

Week four delivered two missed cuts. Ridiculous, I was the unluckiest gambler on the planet to suffer such statistical anomalies at this time!

Undaunted, I continued right up to the time I ran out of money and by week ten I was broke, in fact in debt as some of the bets had been on credit!

The Aftermath

There may be merit in basing future golf performance predictions purely on historical scoring but, if so, I couldn’t find it!

My quest for prediction software subsequently turned to world ranking points but they were calculated over too long a period to be truly useful.

Then on to a mix of recent & course form but that was skewed if a player had not played much in recent weeks and as it turned out course form was, perhaps surprisingly, a very unreliable predictor of future performances on the same course!

And so on to the software I use today which applies a weighted mix of: ranking points, recent form, course form, stats and strength of fields.

Still not the Holy Grail; definitely closer but just another evolution! The quest continues ……

Good luck with your golf punting!


© Copyright MJ Miller (Mike) 6 March 2013