Before the NFL season started, football outsider's prediction system had the Bucs winning 9 games this year. Of course all the writers wrote it off as their system being off. With Tampa Bay off to a 3-1 start, it doesn't look so silly after all.
The best times to judge predictions are after the event has passed.
I don't follow baseball at all but has Dave van Dyck of the Chicago Tribune ever written a follow-up to this with a suitable capitulation? I hate when naysayers are never later held acountable for their premature naysaying.
Joined: 30 Dec 2004 Posts: 534 Location: Near Philadelphia, PA
Posted: Fri Oct 05, 2007 5:35 pm Post subject:
The post is mutating a bit and getting into computers vs humans in making predictions (even though John isn't a computer, he plays one in this drama). The Wages of Wins blog actually collected some info on this:
It's worth reading. The Financial Times piece it links to is also good. _________________ Dean Oliver
Author, Basketball on Paper
http://www.basketballonpaper.com
Joined: 18 Feb 2005 Posts: 142 Location: Cambridge, MA
Posted: Fri Oct 05, 2007 8:43 pm Post subject:
Charles wrote:
What strikes me is that teams with top point guards tend to be projected lower than expected (Arenas, Kidd, Ford/Calderon, Andre Miller), while teams with high scoring swing men, top rebounders and other "stat stuffers" (see Atlanta) are ranked higher than most common sense analysis would have them.
Why hasn't Brevin Knight led any of his teams into elite territory? He's been one of the league's better playmakers for some time now. If great-passing point guards bring something unmeasurable to their teams, wouldn't Knight's teams have fared better? Why did the 01-02 Cavs stink so much with Andre Miller dishing 10.9 apg? I'm not saying that there weren't other factors but if otherwordly passing is supposed to elevate otherwise poor teammates, why hasn't this happened in those instances? Why have the 04-07 Mavericks fared better without Steve Nash than with him in seasons prior? Again, I understand there were other factors (defense) but it causes me to question the importance of assists.
My instincts tell me that Jason Kidd's defense has been as responsible as his passing for the improvements his teams have made since aquiring him. It's likely Steve Nash's shooting and scoring have been just as important to the Suns' offense as his passing.
I don't follow baseball at all but has Dave van Dyck of the Chicago Tribune ever written a follow-up to this with a suitable capitulation? I hate when naysayers are never later held acountable for their premature naysaying.
I sent him a pleasant email, seriously, and got no response.
Joined: 14 Jan 2005 Posts: 1521 Location: Delphi, Indiana
Posted: Sat Oct 06, 2007 6:35 am Post subject:
jeffpotts77 wrote:
Why hasn't Brevin Knight led any of his teams into elite territory? ... Why did the 01-02 Cavs stink so much with Andre Miller dishing 10.9 apg?..
Great passers, scorers, rebounders, etc are often stuck with lousy teammates. Oscar Robertson missed the playoffs 3 straight years before they dismantled the Royals. Then they had Nate Archibald. Lenny Wilkens' best years were with poor teams.
Guy Rodgers, Norm Nixon, Isiah, Mark Jackson, Muggsy Bogues, KJ, TBug, etc, played on some good teams and some bad teams. They all made their teams better than they otherwise would have been. They all went to 'elite territory' when they had guys that would score when they got the ball.
Steve Nash doesn't lead the '02 Cavs anywhere, either. Lamond Murray and Jumaine Jones are not Amare and Marion, after all. _________________ 40% of all statistics are wrong.
Had some PG related thoughts that bounce around but bear on prior comments and impact on team results including but not limited to the east.
Following on what Jeff said, Kidd is considered to bring defense. I've thought looking at stats it has slipped some in recent years and may be overrated but 82games counterpart data is rough and misses stuff including help D so compared to some other top passing PGs maybe it is another source of edge for him. Looking at his on/off chart at 82 games the team also rebounds much better and gets to the line much more on over off and compared to league average.
Back to passing, Kidd's close and dunk assists are about 40% of his total. Nash over 50%. Paul about 35%, D Williams maybe 43%, B Davis 56%, T Parker 30%, A Miller in Philly about 50%, Ridnour 40%. Kidd didnt score as high on this as I might have expected. Looking the other direction Kidd's 25% of assists from 3 pt assists. Not sure how that rates or how important it is.
Team eFG% with him on is equivalent to what 5th best team in league puts up but no where near Nash's on court team number. Nets overall eFG% is 7th. Nets slip back to equivalent of 19th on team eFG when Kidd is off court. He matters.
B Knight in Charlotte raised team assisted FGs from 60 to 65% but team eFG% slipped from 48.3% to 47.4%. Knight directed more but ended up with a less efficient product in raw terms, in this one year.
A Miller had both team assisted FG % and team eFG% up when on court over off in both Denver and Philly. TParker pulls team assisted FG % way down from 67% when off to 56% when on but team eFG% rises from 51% to 52.8%. Directing less with the direct pass than with his drives/scoring and perhaps secondary assists.
Iverson in Denver pulls assisted FGs down from 63% to 57% and team eFG% from 51.1% to 49.2%. Directing less with the direct pass and hurting efficiency with the other directing activity too. Arenas pulls assisted FGs way down but he raise team eFG% pretty well. Advantage Arenas.
Team eFG% with point guard on is more significant to me than assists. Dallas saw only a 0.3% drop in team eFG% in year after Nash left, pulled even in year 2 and moved considerably higher last season. Whether Cuban did the right thing or not letting him go can be debated many ways but this is one of the points in favor of their choice. It worked in part (and in line with Jeff's point about Nash's shooting ability being important) because they replaced Nash with Terry, who had a higher individual eFG% than Nash did. It took awhile but last season both Terry and harris had team eFG% while on the court over 52%, up significantly from 04-05. A pretty successful transition.
PG is one reason I tend to think Boston won't reach the full extent of the highest expectations. I guess you could say Rondo doesnt really have to lead, just not screw it up. His his on/off dta shows a small loss assisted FG % and very little positive impact on eFG% so not screw it up may be about what they get out of him this season.
Team eFG with Ford on the court for Raptors is a touch under 50%, not that strong. Calderon's was 51.3%. Calderon's edge on this matters more in my view than Ford's edge on assisted FGs.
But of course the PG test is more than just team's own eFG%. By that measure D Jones should start for Cavs but both Hughes and Gibson had a better raw team +/-.
Lots of considerations. The Cavs had options but not fully satisfactory ones last season. Upgrades are not that available. Though there was Bibby buzz I don't know how close they got.
Last edited by Mountain on Mon Oct 08, 2007 9:53 pm; edited 3 times in total
Sorry, I was away for a couple days, so I didn't have an opportunity to say this at the relevant point in the thread, but I'd just like to chime in and thank John for coming here and defending his predictions. I can only imagine the amount of flak he takes for making these kind of forecasts. It's probably tempting just to stick to the status quo so you don't have to see people ragging on what you say everywhere you go, but instead he made some gutsy predictions, providing all of us some good water cooler talk.
So thanks for sticking to your guns, John.
I think your explanation makes sense, but as Mike G pointed out, a lot of it relies on the projected PER drops for Daniels, Butler, Jamison and Stevenson. As I understand it, (this is a rough explanation) you find similar players to each player and then see how they did in the following year. So it seems that the similar players to each of those had pretty noticeable declines. My question is, can we give a motivation to that? In other words, can we explain why that seemed to happen in a basketball sense (maybe by finding similar players and seeing why they experienced a decline in numbers)? _________________ XOHoops - A New Kind of Fantasy Basketball
Joined: 14 Jan 2005 Posts: 1521 Location: Delphi, Indiana
Posted: Mon Oct 08, 2007 7:43 am Post subject:
Ben F. wrote:
... thanks for sticking to your guns, John.
.... As I understand it, (this is a rough explanation) you find similar players to each player and then see how they did in the following year. ..?
This looks like using a very small sample size to deliberately introduce high uncertainty. This would lead to some 'radical' predictions, as when one team's players are all predicted to get better or worse. Wilder predictions are then more entertaining, to us readers.
Hopefully John isn't going the Berri route, figuring the more there is to scoff at, the more attention he gets. A player at prime age (26-29) generally shouldn't be expected to lose 10-15% of his productivity -- amounting to 20-40% of his 'value over replacement' -- from one year to the next. _________________ 40% of all statistics are wrong.
Joined: 05 Jan 2005 Posts: 509 Location: Columbus, OH
Posted: Mon Oct 08, 2007 10:04 am Post subject:
Mike G wrote:
This looks like using a very small sample size to deliberately introduce high uncertainty. This would lead to some 'radical' predictions, as when one team's players are all predicted to get better or worse. Wilder predictions are then more entertaining, to us readers.
Hopefully John isn't going the Berri route, figuring the more there is to scoff at, the more attention he gets. A player at prime age (26-29) generally shouldn't be expected to lose 10-15% of his productivity -- amounting to 20-40% of his 'value over replacement' -- from one year to the next.
Mike, why don't you put your method where your mouth is? John's predicted PERs for the upcoming season are out there. Why don't you publish predicted T-Rates (or whatever you call them) for the 2007-08 season. At the end of the season we can compare the results. _________________ Regards,
Justin Kubatko
Basketball Stats!
Last edited by jkubatko on Mon Oct 08, 2007 10:41 am; edited 1 time in total
On John's projected Wizards PER drops I don't have a problem accepting Daniels or Thomas. Jamison I would tend to expect to be pretty stable but a small drop wouldn't surprise me. I havent looked into Stevenson but he is minor.
Butler I had to think about more as it was the most suprising prediction. But looking at his top 10 comparables at same age I found the rest of their careers after 25/26 is pretty checkered and could be considered disappointing in a number of cases (e.g. Jim Jackson, Mashburn, Augmon) and probably would not have been predicted on an individual basis.
Last edited by Mountain on Mon Oct 08, 2007 1:22 pm; edited 1 time in total
Joined: 13 Oct 2005 Posts: 374 Location: Atlanta, GA
Posted: Mon Oct 08, 2007 11:16 am Post subject:
jkubatko wrote:
Mike, why don't you put your method where your mouth is? John's predicted PERs for the upcoming season are out there. Why don't you publish predicted T-Rates (or whatever you call them) for the 2007-08 season. At the end of the season we can compare the results.
Mike, you can use eWins to project future team records, right? Both John and myself have posted predicted standings based on our method of choice; I think it'd be cool to see what eWins predicts about the upcoming season, too. I'm also hoping Berri will predict records using WP, as well as anyone else out there who has a projection method. It would be fun to see who comes closest to getting it right.
Joined: 14 Jan 2005 Posts: 1521 Location: Delphi, Indiana
Posted: Mon Oct 08, 2007 12:40 pm Post subject:
I guess my feeling is that players don't predictably change for better or worse from one season to the next. The only trends I've noticed is that rebounding diminishes as a player gets older (by maybe 1 or 2% per year), FT% and Assist rates get better. Not much else, and nothing has ever indicated to me that a 'type' of player will suddenly, in mid-career, get dramatically worse at age 27 or 29.
So it would be a time-consuming project, and I'd just manipulate lineups to account for player movements. I can't imagine (suddenly) starting to evaluate rookies. Lineup chemistry predictions are only done by clairvoyants. So in brief, I'd just predict whatever happened last year, unless there has been talent in or talent out.
T-Rate (unlike PER or WS or PW%) is designed to be context-independent. A player moving to another team, coming off the bench, getting more or fewer minutes, should not change. This is another reason I don't expect one to change. Some flukishly high PER's can be expected to drop, since the conditions that propped it up are unlikely to persist.
The real unpredictability from one year to the next is in injuries, chemistry dilemmas, and such (essentially) unpredictable stuff. Whoever predicts better is more likely to have just been luckier than righter. _________________ 40% of all statistics are wrong.
All times are GMT - 5 Hours Goto page Previous1, 2, 3, 4Next
Page 3 of 4
You cannot post new topics in this forum You cannot reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot vote in polls in this forum