[martyup]

[quote author=martyup link=topic=13664.msg143391#msg143391 date=1448347112]
How could WIM's prediction of 41 to 28 be better than three of us that predicted 38 to 28?  Also, PalmTree's prediction of 41-27 should not be better than 38 to 28.  Also, how does Garvin's prediction of 42-28 tie those who predicted 38-28?

The "modified error" uses the sum of the difference in spread and the average score.  In WIM's case, he nailed the 13 point spread, whereas you underestimated the spread by 3 points.  On the other hand, WIM's average score of 34.5 was off 6 from the actual average score of 28.5.  Your average score of 33 was closer (off by 4.5), but didn't make up for the missed spread.  That's how your absolute error is lower, but your modified error was higher.

Similar situation with PalmTree: his spread (14) was off by 1, and his mean score was off by 5.5 and Gavin's spread (14) was again off by 1 but his mean was off by 6.5.

The reason I switched to the modified error is that it gives credit for getting the spread right.  Under the old system, a guess of 38-19 would have gotten the same number of points (6) as a guess of 38-25 even though the second score nailed the spread while the first one was off by 6.

BC 
[/quote]

I understand.  Thank you for explaining.

[martyup]

[quote author=martyup link=topic=13664.msg143391#msg143391 date=1448347112]
How could WIM's prediction of 41 to 28 be better than three of us that predicted 38 to 28?  Also, PalmTree's prediction of 41-27 should not be better than 38 to 28.  Also, how does Garvin's prediction of 42-28 tie those who predicted 38-28?

The "modified error" uses the sum of the difference in spread and the average score.  In WIM's case, he nailed the 13 point spread, whereas you underestimated the spread by 3 points.  On the other hand, WIM's average score of 34.5 was off 6 from the actual average score of 28.5.  Your average score of 33 was closer (off by 4.5), but didn't make up for the missed spread.  That's how your absolute error is lower, but your modified error was higher.

Similar situation with PalmTree: his spread (14) was off by 1, and his mean score was off by 5.5 and Gavin's spread (14) was again off by 1 but his mean was off by 6.5.

The reason I switched to the modified error is that it gives credit for getting the spread right.  Under the old system, a guess of 38-19 would have gotten the same number of points (6) as a guess of 38-25 even though the second score nailed the spread while the first one was off by 6.

BC 
[/quote]

I understand.  Thank you for explaining.