Someone asked earlier about the extrapolation of portability vs. cross-era. There's a huge difference.
-Portability IS indeed asking for some extrapolation. Part of that extrapolation is using an
exemplar model of player analysis, where you look at players with similar skillets in different situations. Some of it is pure extrapolation (this is the hardest and fuzziest), and some of it is doing stuff like changing the sample size or mildly increasing an effect. This is why I constantly harp on lineups -- you want to see how a player does with different lineups. Sometimes, those lineups can't play for very long, guys get injured, other players are inconsistent in another area and the coach doesn't go the lineup exclusively, etc.
But for the most part, you aren't just looking at a "team" result but a result of many lineups (perhaps across multiple years), combined with player
exemplars to paint a picture of how a player's skill (and his decision-making of how to implement that skill) apply to different situations. You don't really have to examine crappy teams, because even the best player on bad teams will move the title odds needle by only a few percent.
-Cross-era takes the complexity and heavy analysis required for a portability examination, and exponentially increases complexity. I'm not saying that as a figure of speech, I'm saying the number of variable interactions you have to handle skyrockets...AND all the while your confidence is radically decreasing because many of the variables are very fuzzy. (e.g. how would he handle rule changes? how would he be used/viewed growing up? technology? diet, nutrition, lifestyle? etc.)
DQuinn1575 wrote:drza wrote:[eventually ElGee published which showed Wilt's in/out to be towards the lower end of the spectrum:
Player Years Games MOV Net SIO
Walton 77-78 41 9.3 13.0 11.2
.......
Wilt 65, 65, 70 156 -0.3 0.8 0.3
Paul 07, 10 55 -1.6 1.2 -0.2
This merely quantifies the phenomenon that we were noting basketball-wise in the RPoY. But it's clearly a (very) counterintuitive result, so we spent a lot of time trying to figure out what might have been happening and how important it was to our evaluations.
The difference in this chart is that Wilt was traded, and the team supposedly received equal value. Virtually everyone else was either injured or a free agent. In those cases it is an ADD, not an EVEN situation.
If you are injured, there is a good chance your team doesn't have an adequate short term replacement.
If you are a free agent, then you are being added to the team with no replacement.
That's only true in 1965 and the situations of both teams are explicitly stated, as well as the annotation of a trade circumstance. Furthermore, the trade gives us more data (!) you can view the micro-trends during 1965:
-Wilt plays 9 games without Hal Greer in Philly and the 76ers are -5.6 SRS team without Greer. This should be alarming, even in a 9-game sample. The 9-game sample doesn't mean it's 99% likely Philly was a -5.6 SRS team with that lineup, but it means it's
highly likely they weren't a good team.
-Wilt plays 26 games with Greer. Philly is +2.8 in those games, basically swapping Wilt for Dierking (18 mpg) and Neumann (28 mpg).
-Back in San Francisco, the Warriors played at a 28-win pace with Wilt before the trade. (Better than without him.)
These are three pieces of information that strongly suggest Wilt Chamberlain wasn't having a big impact on the game in 1965. Doesn't mean he was terrible. Or even average. It means it's incredibly unlikely he was fantastic. (No, I didn't control for Costello.)
Also a general reminders about WOWY as I used to have this breakdown on my blog:
-it's a measure of situational value
-it's not exactly the same as on/off bc coaches can't hide/cheat lineups -- in science it's more like the "observed outcome"
-confidence of sample is a problem until 15-20 game samples
That said, I think there are a few SUPER valuable things about this statistic, which is while you'll notice me using it even post-1997 when we have lineup data:
1. It CORRECTS -- in a big way -- the general perception of "how a guy's team performed"
2. It fairly accurately demonstrates situational value, which is really important
3. Even in smaller samples (e.g. 10g, 14g) it suggests trends that are likely
This is valuable information, and like all information should be married with other analysis.
). While this obv takes a little longer, I agree it's valid and fair. 
?
