Durvasa, are you able to list the top five, 5-man rotations for net +/- which include both Yao & Tmac?
Hughes is a natural SG. Hubie Brown said that was one of Cleveland's problems was they were using Hughes out of position and that he couldn't break presses or distribute as well as a natural PG. He is also injured an awful lot, between him and TMac we would miss a whole season of games between 2 players. I would rather start Brooks and see what he could do.
There aren't many lineups that have played more than handful minutes with those two, so it's hard to rank them. Here's how they rank in terms of raw net points: Code: linueup min pts_for pts_opp net/poss net 1 Alston,Battier,McGrady, 76 178 147 +22.0 +31 Ming,Wells 2 Alston,Battier,McGrady, 82 175 148 +14.0 +27 Ming,Scola 3 Alston,Battier,Hayes, 338 646 631 +3.5 +15 McGrady,Ming 4 Alston,Hayes,McGrady, 14 40 27 +44.8 +13 Ming,Wells 5 Alston,Battier,Francis, 7 19 9 +62.4 +10 McGrady,Ming 6 Alston,Battier,Landry, 7 19 10 +64.3 +9 McGrady,Ming
If your hypothesis is that the coin will land on heads 50% of the time, your maximum error is 0.01 and your confidence interval is two standard deviations (which from what I remember is the standard cutoff for disqualifying random chance in testing a hypothesis) then you actually need to flip the coin 10000 times. I'm not sure as to your point here. For one thing there is no such thing as true "randomness" in a computer without some really cool hardware for generating what the comp sci guys call "maximum entropy". I think it's fairly safe to say that you should stick with the formulas in the stats book. That is indeed correct. Once you reach a certain threshold in sample size increasing the sampling size doesn't gain you anything. However most statistical tests are most effective at rejecting a false null hypothesis (the so-called statistical power of the test) when the test group and the control group are roughly the same size. I'm also going to guess that the sample sizes for James, Francis and Brooks are way too small to begin with, especially in the case of Brooks. Small sample sizes mean large random errors. At the very least I'm guessing that when you calculate the error propagation that it's going to be so high as to really make the usability of your results questionable. Put simply, the sample sizes don't look large enough to me to be able to generate useful results, even for ballpark estimates. In other words, saying "The door is six feet tall, plus or minus three feet" is useless to a carpenter.
I have this based on efficiency data from bbvalue... 1. Alston, Rafer - Battier, Shane - Landry, Carl - McGrady, Tracy - Ming, Yao (unit): (135.71 OFF, 71.43 DEF, +64.29 net, only 7 minutes) 2. Alston, Rafer - Battier, Shane - Francis, Steve - McGrady, Tracy - Ming, Yao (unit): (126.67 OFF, 64.29 DEF, +62.38 net, only 7.5 minutes) 3. Alston, Rafer - Hayes, Chuck - McGrady, Tracy - Ming, Yao - Wells, Bonzi (unit): (137.93 OFF, 93.1 DEF, +44.83 net, 13.97 minutes) 4. Alston, Rafer - Battier, Shane - McGrady, Tracy - Ming, Yao - Wells, Bonzi (unit) (118.67 OFF, 96.71 DEF, +21.96 net, 75.73 minutes) 5. Alston, Rafer - Battier, Shane - McGrady, Tracy - Ming, Yao - Scola, Luis (unit) (111.47 OFF, 97.37 DEF, +14.1 net, 82.13 minutes) 6. Battier, Shane - Hayes, Chuck - James, Mike - McGrady, Tracy - Ming, Yao (unit): (136.46 OFF, 125 DEF (YIKES), +13.46 net, only 7.32 minutes) 7. Alston, Rafer - Battier, Shane - McGrady, Tracy - Ming, Yao - Scola, Luis (unit) (102.87 OFF, 99.37 DEF, +3.5 net, 337.67 minutes) I wouldn't consider any of the lineups with less than 7 minutes to be highly statistically significant. Also ... there have been so many lineup changes this season that it's hard to separate good data from junk.
Looks like your data agrees with mine and what I saw at 82games.com -- the small ball lineup with Battier at the 4 and Wells at the 3 was working extremely well for us earlier on. Now that McGrady is back from injury, it would be nice to see Adelman start going to that again, as nice as it is to give rookies playing time. Having Battier, Wells and McGrady on the floor together gives us a nice combination of mismatch opportunities and versatile, high IQ players who can both stretch the floor on offense and switch on defense as needed.
Hmm, maybe I could have been a little clearer here. You need to get a certain sized sample to get a representative picture of your total population. Once you get sample size then increasing it gives you only negligable gains in terms of an accurate picture of your total population. But once you get that sample you still need to compare it to the control group in your experiment, right? And the mathematical tests for comparing two groups work best when those two groups are roughly the same size.
Thanks druvasa and JeopardE. I didn't really have any hard and fast expectations from these observations, nor do I see any after the fact. I was just curious if anything stuck out. While durvasa's stats on Alston w/ Tmac & Yao (etc.) were interesting, I think there are too many factors that could influence those stats to be very meaningful. That's why I asked about 5-man rotations. Looking at both top 5 lists, you do notice that Alston, Battier, and Wells are all represented. I was looking to eliminate or confirm that Alston's success in durvasa's comparisons could also be attributed to the other 2 guys on the court besides Tmac & Yao.
On closer look, when you consider the lineups that have significant minutes, Wells has only been a factor in the specific small ball lineup I mentioned earlier with Battier at the 4. The more common thread there is Battier. My inference is that if you were to rank which players have played the best alongside McGrady and Yao together, you'd have to go with Alston first, then Battier, and then possibly Scola (whose data seems a bit more significant than Wells on a wider scope).
There were these two guys who were supposed to help. I think their names are James and Francis. Whatever happened to them?
