# HERE THEY ARE: The Macquarie quant team's tips for the 2016 Melbourne Cup

2016 Melbourne Cup entrant Wicklow Brave, ridden by track rider David Casey. (Photo by Michael Dodge/Getty Images)

The brainiacs at the Macquarie Research quant team are out with their picks for the 2016 Melbourne Cup.

They have made a significant update to their model after what they admit have been “a couple of lean years” in their annual assessment of the field. This note became legendary in its early years, having picked a box trifecta in 2007 and again in 2010, and the winners in 2009 and 2010.

The updated model is inspired by the rise of value as a theme in equity markets, spurring the team to use quant techniques “to capture how over- or undervalued horses are relative to their odds”.

It’s called the Halpha model. The team explains:

An under-valued horse will, on average, win more frequently than its odds imply, while the opposite holds true for an overvalued horse. For a risk-neutral punter interested in maximising returns, the optimal strategy is actually to consistently bet on the most undervalued horse, rather than the one with the highest probability of winning. This results in less frequent, but much bigger wins. However, we can also use the Halpha Model to “correct” the stated odds, and provide a rank prediction as we have done in prior years.

When tested with \$1 bets on 1000 actual races, its return was \$292. Not bad!

There are some important caveats, given that the focus is on value.

Rather than maximising the chances of picking the winner, the aim of this model is to pick the most undervalued horse; that is, the horse with the highest expected returns. Since the purpose of this model is maximise returns rather than hit rate, readers should note that direct application of this model will predict the winner far less frequently than relying on just the odds. The model depends on the fact that when it does pick the winner, the payoff should more than compensate for its low hit-rate.

They use a multifactor Ordinary-Least-Squares (OLS) regression model which looks at the following variables:

• Pre-race Odds
• Form Rating
• Last Five Race Outcomes
• Age
• Handicap
• Barrier Number
• Sex of the Horse and Jockey

The last point is included because they identified a “Prince of Penzance effect”, in that female jockeys on male horses win statistically more races after controlling for other variables. For the 2016 Cup field, this is appointed to Assign, ridden by Katelyn Mallyon.

The specification of the model is then:

And (yeah, yeah – we’re getting to it) their ranked picks are:

1. Hartnell
2. Jameka
3. Big Orange
4. Oceanographer
5. Bondi Beach

However, they also note that if you are looking for a value bet, the most undervalued horses are Assign, Curren Mirotic and Almoonqith.

One important feature of the model is that it is best used as a strategy over a range of races, so there are limitations to its application to a one-off event like the Melbourne Cup. Also, the team of Zhe Chen PhD, John Conomos CFA, Jeremy Lamplough, Fabrice Schloegel PhD, and Lachlan Palmer note that “despite all our analysis, we actually know very little about horses” and, naturally, that “past performance is not indicative of future performance”.

Happy punting.