The simulated 2020 Championship: who would have reigned supreme?

New data analysis tool crunches the numbers from thousands of previous games to calculate how the original All-Ireland series would have been won

WE SHOULD be heading for the business end of things now. We should know who has come out of the notoriously competitive Leinster and Munster groups in the hurling championship and who are the kings of football out west. We should be hurtling towards the opening round of the Super 8s and looking forward to the high wire act that is knockout hurling. But whatever happens towards the end of this year, the 2020 championships, in their original incarnations at least, will never come to pass.

How they might have panned out is a matter for debate but one data analyst, working in conjunction with DeelySportScience.com, has developed a piece of software that simulates how the championship, according to the statistics and form, might have been.

Diarmuid Whelan is a data analyst who has worked for Apple and Paddy Power Betfair and his model is similar to those relied upon by bookmakers to simulate tournaments such as the World Cups in rugby and soccer before settling on their prices.

By analysing ten years of results, which covers more than 4,000 league and championship matches across hurling and football, Whelan has developed a model that has simulated the entire championship as it was meant to run this summer.

Each team is rated on their ability to score goals as well points and their ability to keep them out. Those ratios are adjusted whether they are at home and away with allowances built in to the simulation for the strength of the opposition. A time decay component is applied, meaning more recent results carry greater significance.

Attack

“For example, take Dublin’s goalscoring record in say 100 games. Their goal attack strength gives you a rating of how much better or worse Dublin are than the average in terms of scoring goals,” Whelan explains.

“In terms of points, it’s how good they are in terms of scoring points. The average might be one. And a good team might be 1.5 which means they are 50 per cent better than the average at scoring points over last ten years. How good are they defensively is calculated too in a similar way… If Dublin are only conceding 70 goals over 100 games and 100 is the average, they are 0.3 below the average. Dublin are strong in terms of scoring goals and if they play a team that is weak at preventing goals that will multiply up the factor and you’d project them to score more goals in that scenario.”

There are areas the simulation can’t allow for. Whelan points out that there is no way allowances can be made for a scenario where a team needs only a very specific result to progress.

“The weaknesses are probably the more abstract things, the ‘needing to get a result’ stuff. Take the Super 8s or the Munster hurling round robins

where say Tipperary might know in the final round all they need to do is not lose by four points or more, you can’t factor in the more behavioural and emotional parts of these things.

“In that case, what a bookmaker will do is throw a margin on it to cover that up, they have that room for error and they might tweak things if say, in the last round, Waterford have nothing to play for, they mightn’t trust completely the simulation in that case.

“In straight knockout games, those issues aren’t there.”

In this simulation, all five Munster teams go into the final round with a chance of progressing, closely aligning with expectations. Galway and Wexford reach the Leinster final with Kilkenny taking third.

The biggest shock across both championships sees Division 4 Wexford beat Division 1 Meath in Leinster.

The final two games of the league were simulated and saw Meath head into the championship on the back of seven losses, while Wexford gained promotion from the basement division.

“That was kind of the standout result. Meath would have played Dublin in round six of the league and Monaghan as well so their trajectory was down. And Wexford in football ended up getting promoted so they were on the up. Meath were still rated in the high 80s in percentage for turning them over but Wexford managed to win.

“This was just one of those occasions where form lines were trending towards each other and it’s just one of those things. These things happen, nothing is 100 per cent, that’s the great thing about sport so it could happen and they do happen in sport all the time.

“It is a shock, and if you look further on, Meath play more to the form you might expect from them.

“Overall, the first thing about the simulation for me was, ‘Great, these look realistic’. I was struck by how probable they were. When you’re running the simulation you’re going, ‘I hope Dublin don’t get beaten in the first round’ – that sort of thing.

“In the Super 8s you had the group of death. Dublin, Kerry, and Donegal got into that group through the back door.

“And in hurling you had the last day where all five teams in Munster could have ended up in the All-Ireland series with Limerick sitting there watching results hoping they stay in the Munster final. The narrative of it all was interesting.”

In the simulation, Dublin end up winning the six-in-a-row while Tipperary also retain their hurling title.

Trends

Data analysis, which can help identify team-, player- and position-specific trends and requirements, is already part of the day-to-day of professional sports teams and is the latest addition to a suite of offerings on DeelySportScience.com which is the brainchild of sport scientist and former London manager Ciarán Deely (right).

Deely has pulled from his experience as an inter-county footballer with Wexford, his stint in charge of the Exiles as well as his day job with QPR, to put together an online resource aimed at assisting coaches across Gaelic games and soccer.

The site is still in its infancy but is already outperforming expectations with subscribers given access to coaching tools, sports science, strength and conditioning psychology as well as game-based session plans.

“The idea is, this is what we have done with QPR and this is what we have done at intercounty and you can adapt then to hurling and Gaelic football and we are introducing a lot of soccer stuff to.” While data analysis has been used to simulate the 2020 championship in this case, the science can also be used to identify match trends, player habits and heat maps and it’s envisaged it will aid coaches make informed, realtime, in-match decisions in the future.

“This is approaching sports science and performance from a data analysis point of view rather than just video clips.

“And that is very normal in professional soccer and with ‘Moneyball’ in baseball and American sports. So I wanted to introduce this into the Gaelic world. And from what I know (Diarmuid) is the only data analyst involved in the GAA.

“I would expect within the next few years that Dublin, Kerry and Mayo would be foolish if they don’t have a data analyst because I’m pretty sure they don’t at the moment but they should.

“And the more data you put in the more it will spill out and it will become clear what is important; what are the critical points … we are identifying the important things to come out of games, why is a team winning so much. It’s the next stage of analysis.”