QU’EST-CE QUE C’EST?!? THAT GARETH DAVIES KICK…
Welcome to the inaugural post of “Qu’est-ce que c’est?!?”. This is the first of what I hope to be a recurring segment where we look at dubious or controversial game decisions from recent rugby matches and use the analytics to determine if the decision made was the right one. For those of you who watched the opening round Six Nations Wales versus Ireland match this past weekend, there is a lot of online buzz about Gareth Davies’ decision to kick the ball away on what could potentially have been the last play of the game.
For those of you who missed the decision and how it potentially cost Wales the game, here it is:
At the time of the scrum, Wales had a 5 point lead and possession of the ball - their win probability was 92.2%. This may seem a bit low given that there were ~20 seconds left in the game, but our win probability model also takes extended play (i.e. game play that extends beyond 80 minutes) into consideration so WP at 80 mins does not necessarily equal 100% for the team that’s in the lead. Regardless, logic would dictate that all Wales needed to do was win the scrum and then maintain possession of the ball until time expired. When the final horn blows, kick the ball off the pitch and they win the game.
While it would be easy to disparage Davies for his decision, perhaps this could be an opportunity to educate players using the numbers? As you can see, after winning the scrum, Davies elected to attack the blind side of the scrum and put a grubber into the Ireland 22m. While the kick did gain 36 metres of territory, by giving up possession of the ball Wales’ win probability dropped from 92.2% to 84.3%. Thus, by kicking the ball to Ireland from the 55m to the 19m line, Wales gained 1.47 Expected Points Added (EPA), but also lost 7.9% of Win Probability Added (WPA). In terms of WPA, you want your contributions to increase your team's chances of winning the game - not decrease.
While Ireland didn’t have a great chance of winning the game when they were kicked possession of the ball, a small chance is better than no chance. If we follow the sequence of play all of the way to the end of the game, when Ireland wins the penalty on the 67 meter line, Wales’ win probability dropped further from 84.3% to 79.9%. Had Billy Burns been able to find touch on the 5 metre line, Wales’ win probability would have fallen even further to 65.9%. For you rugby nerds doing the mental math, yes this means that Billy Burns missing touch was worth -14.0% of WPA. But since it ended up being the last play of the game, it could also be argued that the kick error was actually worth -34.1% WPA. Regardless, it’s clear that it was a costly error, and I’m sure Burns feels badly about the mistake so let's not dwell on it any further.
This is an example of how advanced analytics can be used as a teaching tool to help educate players about decisions that were made and how they can make better decisions in the future. In this particular case, the correct decision was pretty obvious, but what about cases where the correct decision is more subtle and less obvious? Let's say the score was tied in the above scenario? Should Wales try and scrum for the penalty? Or should they play away and try to score while also risking a turnover or conceding a penalty?
Do you see how this type of analytics can be used to optimize decision making at critical times in games?