Front Office Football Central - View Single Post

molson · 04-06-2011, 03:30 PM

It's amusing to read the brainy Red Sox fans at Sonsofsamhorn put together statistical analysis of their crappy start and how meaningful it is.

"Just because starting 0-4 doesn't follow normal distribution doesn't make it non-random, any more than a 4-game losing streak to end the season is non-random, or any isolated 4-game stretch*. All we know for certain is that there appears to be a mild correlation between starting badly and ending the season out of contention. History tells us that's because these two factors are likely caused by an 0-4 starting team being, on average, worse than playoff teams. But that likelihood is far from certain, and there is enough variance that drawing conclusions at this stage is making a mountain out of a molehill. This type of reasoning is precisely why people here lament the use of small sample sizes to draw conclusions that are disproportionate to that sample.

*EDIT - This is to say: flip a coin 4 times in a row and tell me that the result is non-random. I just flipped a quarter and got 3 heads and 1 tails. The tails side clearly has a problem facing up. Have 30 people flip coins 4 times in a row, and there's a good chance results will still not follow a normal distribution in terms of <.500 vs. .500 vs >.500"

04-06-2011, 03:30 PM	#228
molson General Manager Join Date: Oct 2002 Location: The Mountains	It's amusing to read the brainy Red Sox fans at Sonsofsamhorn put together statistical analysis of their crappy start and how meaningful it is. "Just because starting 0-4 doesn't follow normal distribution doesn't make it non-random, any more than a 4-game losing streak to end the season is non-random, or any isolated 4-game stretch. All we know for certain is that there appears to be a mild correlation between starting badly and ending the season out of contention. History tells us that's because these two factors are likely caused by an 0-4 starting team being, on average, worse than playoff teams. But that likelihood is far from certain, and there is enough variance that drawing conclusions at this stage is making a mountain out of a molehill. This type of reasoning is precisely why people here lament the use of small sample sizes to draw conclusions that are disproportionate to that sample. EDIT - This is to say: flip a coin 4 times in a row and tell me that the result is non-random. I just flipped a quarter and got 3 heads and 1 tails. The tails side clearly has a problem facing up. Have 30 people flip coins 4 times in a row, and there's a good chance results will still not follow a normal distribution in terms of <.500 vs. .500 vs >.500" Last edited by molson : 04-06-2011 at 03:33 PM.