# Mitch Robinson Versus A Coin: How I Made a Tipping Algorithm

Friday Night Football. It’s time to put in your tips. One team pays \$1.70, the other \$2.15. It’s moderately close. You study recent form, the ins/outs. You have a feeling about how things are going to go down. But what if you don’t? What if you don’t have a working knowledge of either team? How do you settle on an answer?

You could use a coin.

The coin is quite elegant. A binary solution for a binary problem. Head, tails. Win, loss. You could definitely use a coin. Or you could realise that football is fucking chaos. To predict outcomes in football, a coin isn’t enough. You need chaos to understand chaos. Look into your heart. You already know what you need. The most chaotic of chaotic forces.

You say his name into the winter air. “Mitch Robinson,” you whisper. (It’s Mitch Robinson.)

Consider Robinson with the ball. He could kick it to a teammate, or he could handball it. Or he could fumble it. Or he could shank it to the fourth row. He could over-kick it and accidentally score a goal. Mitch Robinson is beyond random. It’s like you flip a coin, and it explodes.

How do we use Mitch Robinson in our calculations?

Mitch Robinson is the perfect barometer. The quality of his football is equal parts exhilarating and frustrating. His performance may be the most neutral force in the competition. For example Mitch Robinson is a best and fairest winner, but a best and fairest winner in Brisbane.

To turn my dreams into reality, I consulted Robert Younger. Rob runs a popular AFL statistics blog (www.figuringfooty.com) and is an expert on everything football and numbers. (I’m more of an expert on everything Mitch Robinson. Did you know that his son is named Chance? I did not need to Google this.)

Rob helped me build a ranking system based on Mitch Robinson’s performances. We defined the quality of Robinson’s performances using fantasy points and set the sample at 5 years. It works like this: if Mitch has played well against a team, they are more likely to lose. If Mitch played poorly, they are more likely to win.

And just like that, we have a Mitch Robinson tipping algorithm.

Mitch Robinson vs. the AFL

This is the competition, seen through the prism of Mitch Robinson’s performance. Already, pretty strange things are happening. Carlton are the worst team (OK so far), but Essendon are the best (weird). Brisbane are the 5th best team. (No.) But let’s remind ourselves about predictive models. They aren’t supposed to sound right, they just have to be right.

Rob helped me run the numbers, using our algorithm, to tip all the games this year. We went head to head with the simplest tipping strategy- a coin toss. So, assuming heads are wins and tails are losses, does Mitch Robinson beat a fucking coin?

Mitch Robinson vs. A Fucking Coin

“So I have some good news and some bad news,” Rob tells me. “The good news, Mitch Robinson beats the coin pretty easily”. This is excellent news. I google the nearest betting shop to my house. It’s a seven-minute walk.

“The bad news, [it beats the coin] not quite easily enough to be statistically significant.” I close Google Maps immediately. What does that mean? “Usually scientists use a 5% cutoff for what they deem significant,” Rob said. “There’s about an 8% chance that you’d flip as many or more heads than we have had wins”. I think I know where this is going.

“Strictly speaking, we could see this result from chance,” Rob concludes. I want to throw my laptop into a large body of water. Google Maps tells me Albert Park Lake is four kilometres away. I am distraught.

Well, I am distraught until Rob tells me the next thing.

Mitch Robinson vs. Other Tipping Algorithms

Turns out the Mitch Robinson algorithm is not all trash. “If you use this to tip you’d get 75 correct from 135 games,” Rob says. “Most tipping leagues would give you the draws, so you’d have 77 points. Which for this year isn’t too shabby.”

Where does this put me against other tippers? “It’d have you in at least the top half of any tipping league most likely,” Rob says. He directs me towards a league of predictive models, run by Monash University. Rob’s own predictive model is currently ranked 10th out of 73 active tippers.

“That [league] rewards points for margin, so it’s not an entirely fair comparison, but you’d probably be in the top 30,” he says. From 73 tippers. All running sophisticated predictive models.

“I reckon old Mitch still did pretty well.”

What now?

In conclusion, I learn that gambling is generally a bad idea for guys like me. And that’s purely from a mathematical point of view. If some satirical algorithm places me in the top half of a university tipping league, it is somewhat instructive about predicting and predictions.

It’s pretty easy to be above average, flukey really. It must be incredibly complicated to be great. I’ll leave that to experts like Rob Younger.

Mitch Robinson continues to provide joy to my life, even as he recovers from injury. And until he does, we’ll always have his algorithm.

(P.S. Friday night, St. Kilda plays Essendon. If you’re a believer, Mitch wants you to tip Essendon, the best club in the competition.)