As is tradition now with the Ultimus Week point tasks (doing it twice makes a tradition right?), I’m going to look at the Ultimus Week from a statistical stand point. From my recent articles, which you can read here and here, I’ve created a machine learning model that can predict the likelihood of a win based on the team’s roster and whether the game is played at home or away. The article goes into a lot of details about how this is possible, but in short I use some data scrapping from the index to pull the current rosters for every team as well as their schedule. A machine learning model learns from these rosters whether or not they end up winning against a certain opponent’s roster, a binary classification task. It uses some pretty interesting ML trends in convolutional neural networks and long short term memory layers, and it comes out around 75% accuracy.
So, for my Ultimus Week task, I’m going to look into the likelihood that each of the playoff teams, from the outset of the playoffs, would win the Ultimus Trophy. It will involve a few different branches, but it should be pretty fun. The goal is to understand, from a deep analytical perspective based on a machine learning model analyzing the rosters of all playoff teams, what each road to victory would look like (plus what has already occurred!). Let’s get started.
Colorado Yeti – Started Playoffs with 34.9% odds of Ultimus – Now 63.7%
We’ll start with the Colorado Yeti, the favors of the model to win the final trophy – and compared to Orange County (another favourite) they’re still alive. So how do we get to 34.9% odds? They were predicted to win the matchup against Sarasota with 54.9% odds and are expected to have an even greater chance of beating San Jose at 63.7% odds. As any mathematician can tell you, the odds of two independent events occurring is the multiplication of their probabilities. In that case 54.9% x 63.7% = 34.9%, giving Colorado a very nice look for the final award. Interestingly, if Orange County had beat San Jose (as they were predicted to do), Colorado’s odds would’ve decreased to 26% per the model. Now that Colorado is in the final, they’re predicted to win the final at 63.7% odds.
San Jose Sabercats – Started Playoffs at 1.96% odds of Ultimus – Now 27.3%
A great underdog story here as the model predicted San Jose of winning the Ultimus Trophy, at the beginning of the playoffs, with 1.96% odds. Compare that to their Ultimus opponents, the Colorado Yeti, who started with 34.9% odds. How do we get 1.96% odds? Well any team facing a play-in were up against the odds as it dramatically reduces their likelihood of getting to the final. The importance of the bye-week cannot be understated. The odds for San Jose across the playoffs were 25.4% against Arizona (away), 28.3% against Orange County (away) and now 27.3% against Colorado – also away. Multiplying those odds together gives us the 1.96% total playoffs odds for the Ultimus. Incredibly, they are now 27.3% underdogs – but given the games so far, who’s betting against them?
Orange County Otters – Started Playoffs at 33.8%
The one that got away, Orange County started the playoffs with the second highest odds per the machine learning model, of winning the Ultimus itself, only 3% back from Colorado. They were 69.4% favourites against San Jose, and would’ve been 48.8% even-line against Colorado. Unfortunately, things got off track in the first game, and San Jose stole one and erased all those glorious odds.
Arizona Outlaws – Started Playoffs at 5.7% Ultimus Odds
Arizona started with a significant advantage over San Jose, mostly on the back of an expected advantage on home field turf at 53% odds in that first game. Beating Orange County, then Colorado, both were around 33% and 32% respectively, given total odds almost at 6%. Unfortunately San Jose won that opener, and dashed these 1 in 20 odds.
Sarasota Sailfish – Started Playoffs at 7.5% Ultimus Odds
This is assuming that Sarasota would’ve had homefield advantage against San Jose, and that San Jose beats Orange County. If Orange County prevails, the odds actually increase to 8.3%. Interestingly, the model prefers Sarasota at home against Orange County than Sarasota at home versus San Jose. In any case, Sarasota were 61.9% favorites against Chicago and 26.5% underdogs against Colorado. In the latter, their streak was ended by the dominant Yeti before the Sailfish could get to some juicy opponents and favourable Ultimus matchups.
Chicago Butchers – Started Playoffs at 2.8% Ultimus Odds
With the second longest odds, but still better than the Ultimus-final bound San Jose Sabercats, Chicago came into the playoffs with everything stacked against playing away for the entirety of the playoffs. The odds, in each matchup, got progressively worse as well 36.8% against Sarasota, and would’ve been 29.2% against Colorado and 26.2% against San Jose. But, unfortunately, the 36.8% odds failed to come through, and Chicago was ousted in the play-in round.
Best of luck to Colorado and San Jose in the final! Even though the model has an edge to Colorado, 27.3% is nothing to sneeze at for San Jose’s back to back attempt.
873 words
So, for my Ultimus Week task, I’m going to look into the likelihood that each of the playoff teams, from the outset of the playoffs, would win the Ultimus Trophy. It will involve a few different branches, but it should be pretty fun. The goal is to understand, from a deep analytical perspective based on a machine learning model analyzing the rosters of all playoff teams, what each road to victory would look like (plus what has already occurred!). Let’s get started.
Colorado Yeti – Started Playoffs with 34.9% odds of Ultimus – Now 63.7%
We’ll start with the Colorado Yeti, the favors of the model to win the final trophy – and compared to Orange County (another favourite) they’re still alive. So how do we get to 34.9% odds? They were predicted to win the matchup against Sarasota with 54.9% odds and are expected to have an even greater chance of beating San Jose at 63.7% odds. As any mathematician can tell you, the odds of two independent events occurring is the multiplication of their probabilities. In that case 54.9% x 63.7% = 34.9%, giving Colorado a very nice look for the final award. Interestingly, if Orange County had beat San Jose (as they were predicted to do), Colorado’s odds would’ve decreased to 26% per the model. Now that Colorado is in the final, they’re predicted to win the final at 63.7% odds.
San Jose Sabercats – Started Playoffs at 1.96% odds of Ultimus – Now 27.3%
A great underdog story here as the model predicted San Jose of winning the Ultimus Trophy, at the beginning of the playoffs, with 1.96% odds. Compare that to their Ultimus opponents, the Colorado Yeti, who started with 34.9% odds. How do we get 1.96% odds? Well any team facing a play-in were up against the odds as it dramatically reduces their likelihood of getting to the final. The importance of the bye-week cannot be understated. The odds for San Jose across the playoffs were 25.4% against Arizona (away), 28.3% against Orange County (away) and now 27.3% against Colorado – also away. Multiplying those odds together gives us the 1.96% total playoffs odds for the Ultimus. Incredibly, they are now 27.3% underdogs – but given the games so far, who’s betting against them?
Orange County Otters – Started Playoffs at 33.8%
The one that got away, Orange County started the playoffs with the second highest odds per the machine learning model, of winning the Ultimus itself, only 3% back from Colorado. They were 69.4% favourites against San Jose, and would’ve been 48.8% even-line against Colorado. Unfortunately, things got off track in the first game, and San Jose stole one and erased all those glorious odds.
Arizona Outlaws – Started Playoffs at 5.7% Ultimus Odds
Arizona started with a significant advantage over San Jose, mostly on the back of an expected advantage on home field turf at 53% odds in that first game. Beating Orange County, then Colorado, both were around 33% and 32% respectively, given total odds almost at 6%. Unfortunately San Jose won that opener, and dashed these 1 in 20 odds.
Sarasota Sailfish – Started Playoffs at 7.5% Ultimus Odds
This is assuming that Sarasota would’ve had homefield advantage against San Jose, and that San Jose beats Orange County. If Orange County prevails, the odds actually increase to 8.3%. Interestingly, the model prefers Sarasota at home against Orange County than Sarasota at home versus San Jose. In any case, Sarasota were 61.9% favorites against Chicago and 26.5% underdogs against Colorado. In the latter, their streak was ended by the dominant Yeti before the Sailfish could get to some juicy opponents and favourable Ultimus matchups.
Chicago Butchers – Started Playoffs at 2.8% Ultimus Odds
With the second longest odds, but still better than the Ultimus-final bound San Jose Sabercats, Chicago came into the playoffs with everything stacked against playing away for the entirety of the playoffs. The odds, in each matchup, got progressively worse as well 36.8% against Sarasota, and would’ve been 29.2% against Colorado and 26.2% against San Jose. But, unfortunately, the 36.8% odds failed to come through, and Chicago was ousted in the play-in round.
Best of luck to Colorado and San Jose in the final! Even though the model has an edge to Colorado, 27.3% is nothing to sneeze at for San Jose’s back to back attempt.
873 words