[Homer]

[ QUOTE ]
[ QUOTE ]
[ QUOTE ]
does anyone know of any studies which track how accurate projected team wins (per PECOTA) are? I saw back in 2004 Nate did a study on individual PECOTA projections vs. other projection systems, but I haven't been able to find anything like this on team wins.

I'm looking at these in contrast w/ Vegas lines, and there are some big differences. for instance...

The over/under on Tampa Bay is 67, over -115. Thats 11 games less than the projected PECOTA wins.

The over/under on Toronto is 86.5, while PECOTA projects 80, and the under is +105.

[/ QUOTE ]

crockpot said this in a post in the SB forum:

"the pecotas have a standard deviation of about 8.7 games from actual results over the past four years. i believe nate claimed a standard deviation of 9.5 games. the system has been refined over that time, so in theory the number should be a little lower, but we can't really conclude this is the case. considering a perfect system has a standard deviation of 6.3 games, that's still pretty good."

[/ QUOTE ]

tx. is a 'perfect' system off by 6.3 games b/c of luck, injuries, etc?

[/ QUOTE ]

It's just random luck, like flipping a coin. A team that is expected to go 81-81 has a standard deviation of sqrt(162*.5*.5) = 6.3. For a team with a different expected record, it would be sqrt(162*p*(1-p)), where p is probability of winning a game.