tXtFL Bowl 8.0
The Big Game is upon us! The tXtFL Bowl, that is. The Seahawks and Broncos are suiting up for no less than 10,000 x3 games simulated on the tXtFL football engine.
This year, we've run three separate simulations. The first set is based on regular season stats alone, while the second uses only postseason stats. The final set uses a combination of the two, specially weighted to balance evenly the entire regular season with the postseason stats to give greater emphasis to the more recent, more competitive postseason games.
Using regular season stats alone
Seahawks 23, Broncos 19.
The scatter plot demonstrates how variable one game can be from the next but with a score trend tilted in favor of the Seahawks. The box-and-whisker plots confirm that the median scores put the Seahawks ahead, similar to the mean scores listed above.
Using postseason stats alone
Seahawks 23, Broncos 20.
The Broncos tighten things up a bit in the postseason. As you can see in the scatter plot, the trend inches ever so slightly to even, and the median scores pull tighter together in the box plots. The Broncos did make some solid things happen in the postseason, and if they can continue to the trend, they may just upset the Seahawks.
Using weighted regular-postseason stats
Seahawks 24, Broncos 16.
We were a bit surprised to find that bringing the postseason stats to the regular season made life so much worse for the Broncos, despite the fact that the Broncos appeared to do better in the postseason, and the weighting even put less emphasis on the regular season. The end result may have to do with the additional complexities of applying regular season stats to the postseason depth charts (the ones used in these sims) and the individual offensive-defensive match-ups throughout the game, where the powerful Seahawks defense may have taken advantage of deficiencies in particular Broncos' offensive players.
Looking back at tXtFL Bowl 7.0 from last year, we made several key changes under the hood, though overall they're unlikely to have affected these outcomes. We finally restored multiplayer gaming, and we overhauled the underlying models for late-game decision-making, including punt-or-play and when to try a field goal. Multiplayer of course won't affect the Bowl, and the late-game changes will probably offset one another. Another change coming into tXtFL 3.8, on which these tXtFL Bowl 8.0 sims were run, will be home-field advantage, though this again won't affect the Bowl given the NFL neutral site.
As pure Niner fans here, it breaks our heart to see the Seahawks take the simulations so handily, but their numbers do speak to the Seahawks' formidable game play. Of note, our predictions based on the regular or postseason stats individually are fairly in line with current odds from Vegas--though with scores flipped. As the ESPN Insider analysts say, "The numbers tell me Seattle, but my heart tells me Number 18." Perhaps if we had weighted Number 18, Peyton Manning, with the extra presence he always brings, we would see some very different numbers.
Update: The Vegas odds were corrected with the Broncos as most analysts' favorite to win, with tXtFL predicting the opposite result.
Update 2: We've posted the post-game analysis!
This year, we've run three separate simulations. The first set is based on regular season stats alone, while the second uses only postseason stats. The final set uses a combination of the two, specially weighted to balance evenly the entire regular season with the postseason stats to give greater emphasis to the more recent, more competitive postseason games.
Using regular season stats alone
The scatter plot demonstrates how variable one game can be from the next but with a score trend tilted in favor of the Seahawks. The box-and-whisker plots confirm that the median scores put the Seahawks ahead, similar to the mean scores listed above.
Using postseason stats alone
The Broncos tighten things up a bit in the postseason. As you can see in the scatter plot, the trend inches ever so slightly to even, and the median scores pull tighter together in the box plots. The Broncos did make some solid things happen in the postseason, and if they can continue to the trend, they may just upset the Seahawks.
Using weighted regular-postseason stats
We were a bit surprised to find that bringing the postseason stats to the regular season made life so much worse for the Broncos, despite the fact that the Broncos appeared to do better in the postseason, and the weighting even put less emphasis on the regular season. The end result may have to do with the additional complexities of applying regular season stats to the postseason depth charts (the ones used in these sims) and the individual offensive-defensive match-ups throughout the game, where the powerful Seahawks defense may have taken advantage of deficiencies in particular Broncos' offensive players.
Looking back at tXtFL Bowl 7.0 from last year, we made several key changes under the hood, though overall they're unlikely to have affected these outcomes. We finally restored multiplayer gaming, and we overhauled the underlying models for late-game decision-making, including punt-or-play and when to try a field goal. Multiplayer of course won't affect the Bowl, and the late-game changes will probably offset one another. Another change coming into tXtFL 3.8, on which these tXtFL Bowl 8.0 sims were run, will be home-field advantage, though this again won't affect the Bowl given the NFL neutral site.
Update: The Vegas odds were corrected with the Broncos as most analysts' favorite to win, with tXtFL predicting the opposite result.
Update 2: We've posted the post-game analysis!
Comments
Post a Comment