From the FanPosts -Joel
(If you don't get the title, here's a hint. And this is a little longer than I thought it'd be, but I would highly recommend reading it all. It will be well worth it or your money back.)
I have been reading Arrowhead Pride for about a year now, give or take a month or so. With one year under my belt, I now know that this website goes in a predictable cycle with three main stages: the regular season, the off-season before the draft, and the off-season after the draft. And, like the changing of the seasons of the climate, each stage of the AP Cycle has it's own distinctive traits. In the case of the AP Cycle, a person could tell which season it is using only the AP comment and a minimal understanding of the English language.
During the regular season, most comments will contain a paragraph pointing out all of the next opponent's weaknesses and all the Chiefs' strengths to explain why a Chiefs win is plausible, followed by a prediction that the Chiefs will win. Most of the rest are telling all of those who predicted a Chiefs loss how stupid they are.
As the regular season winds down, the game predictions are replaced by lists of numbers, with each number followed by a name. Yes my friends, mock drafts, more divisive than religion, politics, and Tim Tebow combined. Whether you love them or hate them, but you can't escape them during this wondrous/terrible season.
And then comes the third season, the season we're currently in; the post-draft off-season. During this season the numbered lists of the pre-draft off-season are replaced by a numbered list of another kind. And it is these lists that I am here to warn you about. The terrible, horrible season predictions.
To be clear, season record predictions are fine in my book. It's the predictions of individual games throughout the season that I am talking about. That's because the latter come from humanity's general lack of understanding of how the world works. People fail to realize that everything, from how long you wait in line at McDonald's to how long a tennis match will last, is a random process. And when talking about random processes, you have to speak in terms of probabilities. You have to. Football is no exception. If I correctly predict the score of our Monday night opener (23-17 Chiefs) while you didn't, it's not because I knew something you didn't, it's because I just happened to pick the right numbers.
When looking at games individually, you have to look at it like ESPN's Accuscore or whatever it's called. It says, basically, that the Chiefs score an average of 20.2 points with a standard deviation of 3.6 points while the Chargers will score an average of 18.9 points with a standard deviation of 2.9 (although it doesn't give you the standard deviation, which is a shame), and then runs a simulation with those numbers to say the percent chance of each side winning.
For example, in America's tragic loss to Ghana today, the better team probably lost. But if they played each other 100 times, the USA would probably win most of the time. This one just happened to be one of the ones where Ghana won.
To understand what really happens, check this fancy graph I made, plotting the point differential of a game against the opponent for that game:
It's a little hard to see (I'm not the best at making good looking graphs), but, to put it simply, any point above 0 is a win, and any point below zero is a loss, and the x-axis (the horizontal one) is the number of wins our opponent had at the end of the season. As expected, the better our opponent is, the more we would tend to lose by. But, and this is important, it is not a perfectly straight line.
You see that number next to R^2? that tells you what percentage of the variation in the data can be described by the best fit line. In English, that means that only 37% of the outcome of a game is determined by who's playing in it. That means that chance alone determines about 63% of the outcome of a game!* So, for those who didn't hear it the first time...
HOW GOOD THE TEAMS ARE ARE ONLY THE SECOND LARGEST FACTOR IN WHO WINS A GAME BEHIND CHANCE!
So, when somebody tries to predict how every game in the season will go, they are simply wasting their time. Every year, for every team, there will be losses against teams they should have beaten, and wins against teams they should have lost to. Last year, the Chiefs should have lost to Pittsburgh, but should have beaten Oakland at home. But they'd still be 4-12.
And this is why it's ok to predict records. I think the Chiefs will win about 8 games next year, given how they will play on average. But, whenever making a record prediction, even the most knowledgeable person should include an error of about 2 games in either direction to account for chance. So I expect the Chiefs to win between 6 and 10 games next year, and will be surprised if they win more than 10 or less than 6.
And as a little P.S., for those who are optimists like me, that graph I made offers a little more encouragement that chance was not kind to the Chiefs last year, and thus we are closer to being respectable than thought. According to the data, the Chiefs should expect to beat any team with less than 5.6 wins, so the 2009 Chiefs played at the level of a 5-6 win team. In fact, the schedule might have been the killer, since there were only 4 opponents below the 5.6 win point last year, so our expected number of wins was actually less than the average win total of team as good as us.
*For those who want to know, that's not completely right, but I checked the only other thing that I could think of that would effect the numbers (improvement of the team) and it was only about 3%, so unknown causes are still about 60% of any game.