APBRmetrics

Scott S · Joined: 10 Feb 2008 Posts: 46 Location: East Rutherford, NJ

I did a quick study on the strength of Olympic competition since 1960, the first Olympics in which the US sent a significant number of future NBA players to the games. I also included World Championships that featured NBA players from the US (1994, 2002 and 2006) and did a rough estimate of European NBA players in the 1996, 2000, 2002 and 2004 games.

There are a number of factors which make my measurements far from perfect, but I did the best I could with the data I could obtain and the amount of time I considered reasonable. First, I estimated adjusted plus minus based on the statistics of the NBA players (for all NBA seasons) based on the most recent coefficients provided by Dan Rosenbaum’s analysis. I then estimated the possessions of each player in the International event for the US teams (I could not find minutes for most years.) Based on these ratios in the next year of NBA competition (or previous in Bird and Magic’s case), I came up with an expected scoring differential of that year’s team against an average NBA team. I then took the ratio of that team’s expected results versus the NBA versus the actual ratio. I assumed 200 points in an NBA game. Thus, if an Olympic team combines for an expected plus 16 against the NBA one year, I assume they would win in an NBA game by an average of 108-92. If they outscore their opponents in Olympic competition by 70-50, I rate the Olympic competition as (108/200)/(70/120)=92.6% of the NBA level.

I will point out a few of the assumptions that I used. The first few years, there were a couple players who did not play in the NBA. Typically, these players were used sparingly (although not always) and I assumed their adjusted plus minus was 0. I could make an improvement on this by doing some replacement analysis, but I didn’t think it was worth the time. I did not adjust anything for age, which is usually reasonable since the players will often play in the NBA immediately following the Olympics, but not always. This factor is more important to note when I analyzed the European NBA players. I also did not adjust for ABA strength or league strength over time. There did not seem to be any oddities based on low minutes that I felt the need to address since most players had plenty of minutes. I also did not adjust for team fit such as a player being more valuable on one team than another based on his specific skill set. Finally, I didn’t adjust for strength of schedule. These would all be interesting to look into, but most would take an entirely different study to complete. Now that you have an idea what this data means and doesn’t mean, here are my results:

_________NBA Expected Pts Pts
Year Strength Margain Scored Allowed
2006 1.12 32.88 103.6 83.1
2004 1.34 34.07 88.10 83.5
2002 1.07 27.01 92.30 75.4
2000 1.02 27.49 95.00 73.4
1996 0.98 34.73 102.0 70.3
1994 0.88 25.30 120.1 82.4
1992 0.94 40.05 117.3 73.5
1988 0.78 15.09 91.60 61.3
1984 0.77 14.91 95.40 63.3
1976 0.96 11.18 97.30 83.3
1972 0.62 06.44 77.30 44.6
1968 0.69 00.41 82.10 56.1
1964 0.64 03.73 78.20 48.2
1960 0.76 26.32 101.9 59.5

(I apologize if it is difficult to read, I don't know how to format tables on these forums yet.)

First of all, the 2004 Olympics appears to be a very bad, somewhat unlucky appearance for the US. Also, the World Championships, on a whole seem to be less competitive, which is reasonable due to the decreased emphasis and increase in the number of competing teams.

As you notice, the Olympic competition improves at a reasonable pace for the first 30 years, as do most of the USA Olympic teams, with the exception of the astounding 1960 team. Then both improved dramatically in 1992. Although the US team’s improvement can be explained easily based on the addition of NBA players, the other teams improvement would not be as easily explained. Although there were a few NBA players on other teams, there seems to be something else going on. For one, there is quite possibly a diminishing return being experienced among the US team since I didn’t perform replacement oriented analysis. Notice that the international strength was also high in 1960. Also, I imagine that the NBA players were already more adjusted to the NBA game in 1992 than their former college representatives. I would guess that even the college players would be more acclimated to the NBA game than the Olympic game. In order to look at this quickly, I also compared the productivity of foreign players in International competition versus NBA competition. This effect would likely be reversed of what was experienced by the US teams. Since there weren’t many teams with significant NBA players, I estimated the adjusted plus minus’ of the players in the Olympic games and compared them to their career averages in the NBA. This is not ideal, but hopefully the results will be somewhat telling since the average ages of these players in international competition and their NBA careers was similar. Obviously, the coefficient would be different in a different style game, so I ratioed the results so the average would be the same. I also had to estimate turnovers and fouls and, in some years, blocks. I took the ratios of 20 (expected points produced) plus the estimated adjusted plus minus’. The results were that international players performed in the NBA at a rate of 92% of their international competition in these years. Note that I only found data for 57 player/competitions for this figure. As with all of my figures, with few data points, the results are quite variable, but I don’t think they are useless. It does seem surprising that the results are indicating that the Olympics could be stronger than the NBA, but I guess when you consider that my figures are weighed based on actual games played (The US and Argentina play more games than Angola, for example), these assumptions could be reasonable.

Sorry I wasn’t able to get the data for this past year to estimate this year’s US team’s NBA expected scoring margin. I will try to do this when I get the chance. I am certain that they should be the favorite again, but can probably be expected to lose about one game in the tournament since the scoring differential variability is much higher than in the NBA. In Olympic history, I estimate the Pythagorean exponential factor to be more like 6 in the Olympics opposed to the 14 used for the NBA.

I would appreciate any feedback. I am not sure what studies would be most appropriate to produce. I would like to have elaborate analysis to post, but it is difficult to finish them at a relevant time with my schedule as it is. Hopefully my quick study was helpful, obviously there are a ton of things I can do to improve on it, but then I will never finish an analysis. I guess it would be best to “master” specific and small analysis’s until I have something of larger scale completed. Let me know what you think.

Mountain · Joined: 13 Mar 2007 Posts: 1527

Scott S · Joined: 10 Feb 2008 Posts: 46 Location: East Rutherford, NJ

Yes, estimated adjusted plus minus is statistical +/-.

For all of the things I did not take into account, usage level was accounted for at some level in international competition. Ideally, I would have created estimated plus minus' per individual possession on offense and per team possession on defense and then converted these to the percentage of possessions used on offense and the percentage of minutes played on defense for that person in international play. However, I could not find individual minutes for many years, so I created the best estimate of player possessions (usage) that I could and used this for both offense and defense. If you are referring to adjusting for usage level based on NBA statistical +/-, I definately agree with you. I would imagine that this could exaggerate the expected margains and, thus, the strengths. I plan on making adjustments to my NBA data for usage and other things in the future, but I haven't done that yet.

You are right in that there are certain aspects to defense that are not directly considered in statistical +/- and, to me, defense is not as accurately measured in this estimated stat, and therefore, my study. However, this does make me feel better about weighing my statistical+/- toward the offensive end. (This is predominantly coincidental.)

I really appreciate your comments and welcome any informative criticism you may have as well.

Mountain · Joined: 13 Mar 2007 Posts: 1527

Sounds like we are on same wavelength.

If you take it further I (and I assume others) will appreciate the look. New data coming soon to add to any methodological adjustments. The story will add a chapter and perhaps some of the old chapters will be worth another look and might change and become sharper and more tied together or more blurry. Always appreciate as much narrative as the researcher feels they can or want to give to help digest the data and spur discussion.

dquinn1575 · Joined: 03 Jun 2008 Posts: 12

Good work, I like this topic a lot.

Can you explain the 1st column - maybe walk me through one year's calculation - I understand your example, but I would think the number for the 1960 team would be .897 -113.16/86.84 nba spread versus 101.59 - 59.5 olympic spread is .566/.631 or .897 - What am I missing?

Scott S · Joined: 10 Feb 2008 Posts: 46 Location: East Rutherford, NJ

Thanks and good point. I actually explained my strength different than I calculated it. The formula for 1960 is actually (113.16/86.4)/(101.9/59.5) = 1.30/1.71=.76. I don't know if one is better than the other and I didn't have any specific reason I chose that formula, but I think any reasonable measure would be useful. I appreciate the feedback.

dquinn1575 · Joined: 03 Jun 2008 Posts: 12

I like the way you actually did it (.76) better. Two things:

I think rather than using 200 points, you should use the ppg for the league - 236.2 for 1960-61 instead of 200.

I think you somehow are overrating how these teams would do in the NBA - the 1960 team would rank as virtually undefeated -in the NBA -

If you took the 1964 Royals (Lucas,Oscar,Boozer,Smith,Arnette) Embrt/Twyman instead of Bellamy/Dischenger and add West (1st year in NBA) I dont think you get a 65 win team, let alone 80 wins (in 82 games)

Scott S · Joined: 10 Feb 2008 Posts: 46 Location: East Rutherford, NJ

I agree that I shouldn't have converted the margin to 200 and that the league average would be best, I will try to revise my results later this week. (I typically convert scores to an average of 200 to be consistent, but I don't think I did that this time (I calculated the adjusted plus minus' a while ago). As far as overestimating the NBA performances, I imagine that would be possible, since I did not look into the effect of diminishing returns (meaning that adding 2 estimated adjusted plus minus' of 9 might not reach a total of 18 points added) and regression factors Rosenbaum used were not designed for the 1960 game, but the first would at least be consistent from competition to competition and the second would be expected to at least be a reasonable approximation. Also, diminishing returns should not be a major factor in the comparison of European NBA players in their NBA performances compared to their international performances.

dquinn1575 · Joined: 03 Jun 2008 Posts: 12

34 points better than league average team seems way too high - so heres what I did for the 2004 team

I allocated the minutes based on their Olympic minutes.
I allocated the fga based on Olympic fga.
I calculated nba reb/min to get rebounds
I allocated fta and 3 pt fga based on nba ratios
I assumed nba average turnovers, and average assists to fga

I got a team that scored 102.9 pts on 92.8 poss - and one that got 55% of the rebounds which I think would be the best rebounding team ever in the NBA.

For defense, no player was all-defense in 2003-2004, so I assumed average defense other than rebounding.

It gave me a defense of 94.6 ppg on 92.8 poss -1.019 der- the league best was .988 or 1.1 points better

That gives me 8.3 pts better than average, or 62 wins -

That appears much more likely than 34 pts better

Statman · Joined: 20 Feb 2005 Posts: 217 Location: Arlington, Texas

Scott S · Joined: 10 Feb 2008 Posts: 46 Location: East Rutherford, NJ

It seems like you are estimating some Offensive Rating equivalent, or at least a per possession basis. I think that, in extremes, this would likely be much more accurate than my method, but probably less accurate in "normal" margins. As I mentioned before, I didn't have a great deal of olympic stats available for all years, so I didn't think it would be worth it to allocate all of those stats. As far as comparing an Offensive Rating to an offensive estimated adjusted plus-minus, I had some reservations with each. I have already stated the problems with the plus minus method. I was concerned that the Olympic teams had recently been teams who were more of shot creators. The coefficients for the est adj pm indicate a player can be a well below efficient shooter, percentage-wise, and still produce value taking shots. I think that ideally circumstantial shooting percentages should be compared and incorporated into an Offensive Rating, (I plan on doing this study in time). A good shot late in the shot clock could be a bad shot early in the shot clock. I would have to agree that I suspect your figure would be more accurate, but I wanted to be as consistent as I could and had limited data. Out of curiosity, what weights did you give to offensive and defensive rebounds, steals, blocks and assists? And thanks again for the input.

dquinn1575 · Joined: 03 Jun 2008 Posts: 12

I was working up the 04 team, which Dean Oliver had as just a little better than an average team.

http://82games.com/comm52.htm

I really dont think the 04 team is 20 points better than the 08 Celtics

With a 40 minute game (less need for bench) I dont think the 08 team is 20 points better than the Celtics.

I took an excellent NBA defense, made a team 60% total rebound % (34 0ff/80 def). With 8 3 pointers a game, in order for the team to be 30 points better than average the team had to shoot 57%

Scott S · Joined: 10 Feb 2008 Posts: 46 Location: East Rutherford, NJ

dquinn1575 - I don't know exactly how the US teams would have done in the Olympics, but I am almost certain that the 92 Dream Team is good enough to consistently beat the 96 Bulls when you compare player to player and skill to skill. If you look at Oliver's analysis, it indicates otherwise. There are definately some factors that can undervalue some players when analyzing them on offensive efficiency alone, although it may be a good estimator in general and generally moves in the right directions (positive and negative values). That said, I am not sure as of yet which methods are best in this environment.
Statman - what do your ratings mean and what are they based on? Rank-wise they seem reasonable I would suppose. Does Lebron's rating of 163 mean that if he plays the whole game on an average team the team will, on average, win by 63/5 = 12.6 points (ignoring other factors)?

Statman · Joined: 20 Feb 2005 Posts: 217 Location: Arlington, Texas

Statman · Joined: 20 Feb 2005 Posts: 217 Location: Arlington, Texas