Post by Ramen on Sept 24, 2014 16:22:51 GMT
Transferring this from the other boards, and will be posting some forecasts when I get the time, but what do you think our regular season record is going to be at the end of the year?
My best guess right now is somewhere around 9-11 wins.
------------------------------
Nerd Talk - preseason wins do not predict regular season record.
Yeah, I'm sure most people already know this. But it's sometimes good to put hard numbers to questions like these, so I collected data on regular season win-loss records for every team since 2010, as well as preseason win-loss records during that time.
Without going into all of the boring details, preseason win-loss record has virtually no predictive power when it comes to regular season win-loss record. This is true whether we measure the absolute number of wins, the win-loss percent, or other ways of measuring regular/preseason performance.
The best way to answer this question is with a statistical technique called regression analysis that basically predicts regular season wins with preseason wins (or win percent). I also added a variable for "net points", the difference between points scored and points against for each team in the preseason. The idea is that perhaps losses themselves are not very predictive, but blowout losses might tip off a losing season (and vice versa).
For those who want the nerd talk, here are the regression results:
Preseason wins:
coefficient: -0.205
p < 0.623
Preseason Net Points:
coefficient: 0.032
p < 0.04
Constant: 8.42
R-square = 0.061
Okay, so what does this nerd stuff mean? First, it means that preseason wins are not statistically significant, meaning that we cannot believe that have any effect whatsoever on post-season wins. Put simply, you're better off rolling dice or flipping coins to predict the regular season than you are using preseason win/loss records.
Second, it means that the "net points" IS statistically significant, meaning there's only a 4% chance that there's not a real relationship. However, the number is really small. A net spread of 160 points (meaning the team scored 160 points MORE than they allowed against them) would only gain the team 3 additional wins during the regular season. So only at the extremes of preseason blowouts can we find any insight into the regular season.
Finally, the "r-square" just tells us how accurately we can predict regular season wins from preseason numbers. And it frankly sucks. Only 6% of the regular season wins can be attributed to preseason wins and net points. In other words, 94% of the regular season outcome is explained by things other than preseason performance.
Based on the Falcons' performance so far, the model would predict that the Falcons would win about 7.5 games in the regular season, plus or minus about 4 games. Obviously, that's not very insightful because it means the Falcons could win between 3.5 to 11.5 games next year.
And that's the whole point of this. Even if the Falcons go 1-3 in the preseason, and as long as we don't get blown out for most of those games, the preseason performance (in terms of wins/losses and points scored) means absolutely, positively nothing.
Update - Since we went 2-2 in the preseason, and we're starting out pretty good so far, it looks like an okay but not great regular season record is likely. But again, the preseason win doesn't predict very well, so it's virtually meaningless.
My best guess right now is somewhere around 9-11 wins.
------------------------------
Nerd Talk - preseason wins do not predict regular season record.
Yeah, I'm sure most people already know this. But it's sometimes good to put hard numbers to questions like these, so I collected data on regular season win-loss records for every team since 2010, as well as preseason win-loss records during that time.
Without going into all of the boring details, preseason win-loss record has virtually no predictive power when it comes to regular season win-loss record. This is true whether we measure the absolute number of wins, the win-loss percent, or other ways of measuring regular/preseason performance.
The best way to answer this question is with a statistical technique called regression analysis that basically predicts regular season wins with preseason wins (or win percent). I also added a variable for "net points", the difference between points scored and points against for each team in the preseason. The idea is that perhaps losses themselves are not very predictive, but blowout losses might tip off a losing season (and vice versa).
For those who want the nerd talk, here are the regression results:
Preseason wins:
coefficient: -0.205
p < 0.623
Preseason Net Points:
coefficient: 0.032
p < 0.04
Constant: 8.42
R-square = 0.061
Okay, so what does this nerd stuff mean? First, it means that preseason wins are not statistically significant, meaning that we cannot believe that have any effect whatsoever on post-season wins. Put simply, you're better off rolling dice or flipping coins to predict the regular season than you are using preseason win/loss records.
Second, it means that the "net points" IS statistically significant, meaning there's only a 4% chance that there's not a real relationship. However, the number is really small. A net spread of 160 points (meaning the team scored 160 points MORE than they allowed against them) would only gain the team 3 additional wins during the regular season. So only at the extremes of preseason blowouts can we find any insight into the regular season.
Finally, the "r-square" just tells us how accurately we can predict regular season wins from preseason numbers. And it frankly sucks. Only 6% of the regular season wins can be attributed to preseason wins and net points. In other words, 94% of the regular season outcome is explained by things other than preseason performance.
Based on the Falcons' performance so far, the model would predict that the Falcons would win about 7.5 games in the regular season, plus or minus about 4 games. Obviously, that's not very insightful because it means the Falcons could win between 3.5 to 11.5 games next year.
And that's the whole point of this. Even if the Falcons go 1-3 in the preseason, and as long as we don't get blown out for most of those games, the preseason performance (in terms of wins/losses and points scored) means absolutely, positively nothing.
Update - Since we went 2-2 in the preseason, and we're starting out pretty good so far, it looks like an okay but not great regular season record is likely. But again, the preseason win doesn't predict very well, so it's virtually meaningless.
