toffee
Well-known member
We can win out the rest of the season, it may not be enough to make the playoff. But it's entirely possible to win the rest of the games.
Sent from my IN2017 using Tapatalk
Sent from my IN2017 using Tapatalk
Shanegotyou11
Well-known member
- Joined
- Oct 9, 2017
- Messages
- 5,371
- Reaction score
- 381
toffee":2wxt505p said:
Win tomorrow and 1 back of wc.
Shanegotyou11
Well-known member
- Joined
- Oct 9, 2017
- Messages
- 5,371
- Reaction score
- 381
toffee":ohf449d9 said:
Win tomorrow and 1 back of wc.
SoulfishHawk
Well-known member
It's just too bad they gave away so many games this year. Last year, winning close games was the norm.
Complete opposite in 2021.
Complete opposite in 2021.
A.D.I.D.A.S.
Member
- Joined
- Jan 16, 2018
- Messages
- 143
- Reaction score
- 16
Is this you, OP? 
[youtube]A4XQmnN1Rkk[/youtube]
[youtube]A4XQmnN1Rkk[/youtube]
HomerJHawk
Well-known member
^^^^^^^^
Nice!!!
Nice!!!
Our actual chances are bound to be higher than that, probably somewhere around 6%.
538 gives each team an ELO and then runs a Monte Carlo where they play the season out a large number of times and teams win or lose each game based on randomness, the difference in ELOs, and home field advantage. The percentage assigned to each team is based on how many simulations they made the playoffs compared to the total, so perhaps they did 10,000 runs over the remaining season and the Seahawks made the playoffs 421 times.
The problem here is that they are limited to using current ELO ratings, and that makes the remaining simulation only reasonable if they are all independent events going forwards. That is decidedly not true in football, where teams go on streaks for causative reasons such as key injuries or having parts of their schemes figured out. The way that you would actually want to do this is to re-calculate future ELO ratings after each fictional week based on what you observed to with some kind of correction factor for how likely upsets were to be random or based on an ongoing factor. That's too much arm waving for 538 and so they stick with a simplified version, but it means they overrate favorites and underrate underdogs as a result.
To look at a specific example, the Cowboys are "85% likely to make the playoffs" because they are #9 in ELO and play a lot of teams with a much lower rank. They are at the Saints on Thursday where they are favored to win 64% of the time. In the 36% of occasions where they end up losing, some of those will be due to bad luck such as unlucky bounces, good/poor guesswork, and officiating calls. However, some may be because key players go out with serious injuries, or they get exposed in some manner that future teams will take advantage of. In that case their chance of winning in weeks 14-17 drops across the board because they are dependent events; perhaps the Cowboys could end up as underdogs in all of those weeks. The 85% figure doesn't account for that possibility.
538 gives each team an ELO and then runs a Monte Carlo where they play the season out a large number of times and teams win or lose each game based on randomness, the difference in ELOs, and home field advantage. The percentage assigned to each team is based on how many simulations they made the playoffs compared to the total, so perhaps they did 10,000 runs over the remaining season and the Seahawks made the playoffs 421 times.
The problem here is that they are limited to using current ELO ratings, and that makes the remaining simulation only reasonable if they are all independent events going forwards. That is decidedly not true in football, where teams go on streaks for causative reasons such as key injuries or having parts of their schemes figured out. The way that you would actually want to do this is to re-calculate future ELO ratings after each fictional week based on what you observed to with some kind of correction factor for how likely upsets were to be random or based on an ongoing factor. That's too much arm waving for 538 and so they stick with a simplified version, but it means they overrate favorites and underrate underdogs as a result.
To look at a specific example, the Cowboys are "85% likely to make the playoffs" because they are #9 in ELO and play a lot of teams with a much lower rank. They are at the Saints on Thursday where they are favored to win 64% of the time. In the 36% of occasions where they end up losing, some of those will be due to bad luck such as unlucky bounces, good/poor guesswork, and officiating calls. However, some may be because key players go out with serious injuries, or they get exposed in some manner that future teams will take advantage of. In that case their chance of winning in weeks 14-17 drops across the board because they are dependent events; perhaps the Cowboys could end up as underdogs in all of those weeks. The 85% figure doesn't account for that possibility.
Similar threads
