I'll add juicier tidbits below, but in keeping with the policy of not quoting whole pages so Denny doesn't get sued... http://insider.espn.go.com/nba/trai...=hollinger_john&page=TrailBlazersProfiles1011
Why would you even write that? Didn't you read where Mixum said he wouldn't play at all this season? According to him, expecting a single minute out of Greg is like expecting Trout to ace an IQ test....nah gawna happen
so... of Portland's top 8 rotation, JH has 1 player improving their PER over their 09-10 campaign, 1 staying the same, and 6 regressing??? 2009-10 season - 2010-11 projection Andre 18.1 - 16.1 Roy 21.3 - 20.8 Nic 17.3 - 18.7 LA 18.2 - 17.7 Greg 23.1 - 21.9 Camby 17.0 - 14.1 Wes 12.3- 12.3 JBay 14.3 - 14.1 maybe there is some sort of wet blanket disclaimer I'm not reading as a non-insider, but it seems pretty much like Hollinger is low balling his estimates STOMP
I think its due to the fact that this year if we don't have an insane amount of injuries, there are just to many players that may get time hollinger can't really predict how much of an impact it will have on everyone.
I don't think he really biases his results in any way. He has a formula, he inputs the numbers and it spits something out -- and age definitely plays a role in a couple of the expected drops (Dre and Camby). All-in-all, his system usually seems like it under predicts for youngish players when I've looked at the results in the past, which isn't a bad way to approach these projections as it's usually better to be slightly conservative than to be wildly optimistic.
how about shooting for realistic? Andre has 3 straight seasons in the 18's... now entrenched as the only real PG option in PDX he's suddenly going to drop 2 full points? His Roy projection is below his career average... dude will be 26 next year. His LA projection is below every season save his rookie year... he'll be all of 25 next season. Etc... I understand lowballing to keep egg off your legacy's face, but he's hedging too hard STOMP
Most projection systems, across sports, have some form of regression built in. As good as a player is, you're more likely to be right about his production, going forward, with a number between his weighted average of recent production and the league mean (but closer to the weighted average of recent production). It almost always means you won't come close to getting all the players right, but the more players you combine into the sample, the closer the reality will match the production...the errors cancel out. For any single player, I wouldn't regress his numbers at all. Because if you just want to predict one player, you don't want to use a method that tends to miss on a lot of individuals. But if you want to get a sense for what a team will accomplish, that system seems to work best. Which players end up doing what may not be accurately nailed by the projection, but you'll get a better idea for what the team's production will be.
Good idea, but...It's true that missed games inflated last year's per-game stats, for players who would normally play fewer minutes. But PER is supposedly per minute, so that shouldn't explain this year's deflation discovered by Stomp.
PER is also based upon the number of wins of the player's team. But Hollinger is predicting that our wins will increase, so a win reduction doesn't explain it. I think Rhal's explanation of minutes is right, after all. Hollinger's deflation is because the portion of the wins attributed to the average player will go down from last year, if their are fewer injuries this year. For example, he assumes Oden plays all season and has a 21.91 PER. Oden getting more minutes decreases his teammates' win shares per game. This decreases their PERs, even if their stats per minute remain the same.
It takes into account the number of wins of the player's team. So if you play fewer minutes per game this year due to fewer injuries, and the number of wins stays the same, then you get less credit for those wins in your PER. Hollinger predicts more wins for us, but I'm guessing that that doesn't compensate for the fewer minutes for the average player, due to fewer injuries. Thus, the average player's PER gets a smaller piece of the pie (pie meaning total wins for the season).
No, it doesn't. http://www.basketball-reference.com/about/per.html I think you assume this because most of the advanced stats that make up PER are usually listed together with offensive/defensive/total win shares.
Sure, it's simple! uPER = (1 / MP) * [ 3P + (2/3) * AST + (2 - factor * (team_AST / team_FG)) * FG + (FT *0.5 * (1 + (1 - (team_AST / team_FG)) + (2/3) * (team_AST / team_FG))) - VOP * TOV - VOP * DRB% * (FGA - FG) - VOP * 0.44 * (0.44 + (0.56 * DRB%)) * (FTA - FT) + VOP * (1 - DRB%) * (TRB - ORB) + VOP * DRB% * ORB + VOP * STL + VOP * DRB% * BLK - PF * ((lg_FT / lg_PF) - 0.44 * (lg_FTA / lg_PF) * VOP) ]
Summarized. It is a blending of many different statistics (mostly offensive and unfortunately only two defensive statistics as most are intangible and not tracked) as positives, then subtracts out negative statistics (such as missed shots and turnovers). Then adjusts that overall "score" to a per/minute pace so the entire league as an average will come out to 15.0. So the formula takes all statistics for the entire league and creates a baseline score of 15 per minute for time that players are out there actually playing on the floor. Therefore it not only allows you to see seven different facets of the game in a combined stat score, but it also allows it to only account for the time that player is on the floor as it is clear that if Durant plays 42 mpg and Kobe plays 32 mpg, Kobe would obviously have better overall stats if he played as many minutes as Durant and vice-versa, so this accounts for what they are actually accomplishing while playing on the floor per each minute. Clear as mud?
