Originally Posted by
Badly Ironed Shirt
Let's go back . You said "only once in 50 years has a team come back from being 5 points adrift by winning their last 6 games"
Those are the constraints you set yourself, of course the pool of data will be small. Your statement has as much validity as me claiming that a man named Claudio has never won the English League.
First rule of statistical analysis (trust me, I do this sort of thing for a living) is to be clear on what pool of data you are working from. Your pool of data is flawed, here is why
1) over the last 50 years teams have had to play 42 or 38 games in a season. Your pool of data is inconsistent, and you have made no allowances for that. If you have, you need to say what those allowances are
2) before 1982 there were only 2 points for a win. Your data becomes even more inconsistent
3) you add the constraint FIVE points adrift, yet you have openly admitted you have only gone as far back as 1992. Furthermore, your data sample includes teams more than 5 points adrift having games in hand! Surely you can see why your sample is looking more ridiculous
4) you add even more constraints "by winning their last 6 games". Point 1 should be factored in again (i.e. 6 games over 38 or 6 over 42). But you also failed to expand - e.g. Had any team overturned a Five point deficit by not winning the final 6.
You, basically, use stats to back up your own answers to your own questions . You are guilty of this regularly. You are more than welcome to shadow me on a day at the office so that you can get a better understanding of just how statistical analysis is performed to both answer questions (always raised by an outside source) and to use those statistics to forecast (not predict) the future.
You are a weekend staristician, and your stats are one dimensional.
"All grass is blue" - my sample is one garden in Tokyo that has blue grass. That's, essentially, what you are doing here. It's deeply flawed and I don't need to produce my own statistics to prove why.