Taking a Charge in the NBA
Home | Teams/Players | Commentary | Premium Services | Contact Us

Predicting the NBA's Most Valuable Player

NBA Statistical Analyst Michael 'wiLQ' Wilczynski
by Michael 'wiLQ' Wilczynski
Comments can be directed to [email protected]

The launch of the recent nbatop.com Research Program has brought us a flurry of very interesting ideas for NBA analytical articles, many of which we hope will eventually find their way to publication on the site. The following is the opening salvo, a breakout piece from one of the sharpest NBA analysts in Poland!

"Ingredients of the MVP award"

Every time I read about who should win the MVP award I have a feeling that it is just public guessing. What is even worse is that the more I read about it the less I know for sure, because every writer or publisher has his own candidate and his own justification.

It is rather hard to understand this when there are tons of useful statistics available. Why can we not use some of them to determine the winner? With that mission in my mind I started.

First I examined the winners of the MVP award from the previous twenty years.

SeasonMVP WinnerTeamGames PlayedPlayer's TendexTeam's WinsLast Yr WinsConf. Rank Pts, Ast, Rebs, Leader? Improved Tendex of 3+?Most Wins?
2004/05Steve Nashpho7524,27062291YesYesYes
2003/04Kevin Garnettmin8233,33058511Yes

2002/03Tim Duncansan8129,9916058[2]

2001/02Tim Duncansan8230,9485858[2]
2000/01Allen Iversonphi7123,54956491YesYes
1999/00Shaquille O'Neallal7934,89267511YesYesYes
1998/99Karl Malone (*)uta8025,4546162[2]

1997/98Michael Jordan chi8224,66262691Yes

1996/97Karl Maloneuta8230,54964551

1995/96Michael Jordan chi8230,05272471YesYesYes
1994/95David Robinsonsan8132,59362551

1993/94Hakeem Olajuwonhou8032,4815855[2]

1992/93Charles Barkleypho7632,78662531
1991/92Michael Jordan chi8032,09167611Yes
1990/91Michael Jordan chi8233,54361551Yes

1989/90Magic Johnsonlal7932,92463571Yes
1988/89Magic Johnsonlal7735,54557621
1987/88Michael Jordan chi8235,7805040[3]YesYes
1986/87Magic Johnsonlal8033,93465621YesYesYes
1985/86Larry Birdbos8232,39967631

1984/85Larry Birdbos8035,06363621

(*) - Games played and Wins estimated as if it were full season.

A couple of conclusions can be made within a blink of an eye.

All of the players above:
- were obviously playing very well
(A Tendex Rating of around 20 can be translated into averages of 20pts, 46% (7/15) fg, 75% (6/8) ft, 5 reb, 5 ast, 1 stl, 2 t/o and 3 fouls per game),
- missed few games (Iverson is the lone exception of an MVP who appeared in less than 75),

So these two conclusions can easily become two minimal requirements:

1) MVP candidate must have Tendex better than 20
2) Has to play in at least 70 games.

The Team of the MVP winner:
- won at least 50 games and more often around 60,
- won a few games more than in the previous year,
- was usually best in its conference and no worse than third place.

More than half of the players above:
- won in one main and simple statistical category (points, rebounds or assists per game),
- had a much better year than previous one (tendex improved minimum by 3),

But of course not all of the conclusions above are equally important.

We all know that when a player is having a great year compared to others but his team is at the bottom of standings he will not even be considered for the award. On the other hand when a player is having only a good year but his team is by far the best in league he will draw a lot of attention from MVP voters. So I assume that those two ingredients are equally important.

Continuing my meditation, when a player is having a great year for the best team but he misses a meaningful number of games his chances for MVP are rather weak because the team without him is doing just fine.

Another issue regarding the MVP award is team improvement. We had great examples last year when two players, Nash and Shaq, fueled a huge turnaround in Phoenix and Miami respectively. Because having great improvement in the wins column is often connected with many wins, it is not as important as wins themselves but in expressive situations it can help.

On the other hand, winning fewer games than the previous year has some negative impact on MVP voters even if a player's team is still playing great. The explanation is pretty simple, everybody knows that this team can play better so less praise is kind of the expected experience.

There are also two additional factors, which are not necessary but can definitely help so I called them bonuses. Improved game comparing to previous season and being the best in one main simple category. Because it's always impressive when a great player has become even greater and winning points, rebs, assists per game shows the brilliance in some aspects of player's game.

With those ingredients of MVP in mind I came up with idea:
- Why not try to translate those conclusions into numbers?

Let's name the final result as MVP Points.

I started with tendex as an index for a player's game. The best players in the NBA usually have tendex around 30 so to make team's wins equally important we can simply divide number of wins by two. Because 60/2=30 and best teams usually have around 60 wins. Then I added another conclusion to mix, games played divided by four (80/4=20 so it is important but not as much as tendex or wins) and the same with team's improvement (Wins minus Wins from previous season divided by four).

But unfortunately not all of the conclusions can be translated as simply as that because for example the best team in the conference can have a different number of wins each year.

So I compared every MVP winner from last 20 years with all other players on any given year and determined how to set these bonuses.

I found a perfect match with:

  • + 2 MVP Points for most wins in conference (with simple explanation: home-court advantage in playoffs),
  • + 5 MVPP for most wins in conference having at least 4 game advantage.
  • (home-court advantage with high expectations that this team will make it through to the finals),
  • - 5 MVPP when a player's team has lower than third most wins in the conference (MVP is all about winners not good players on bad teams),
  • - 1 MVPP when a player's team has less wins than previous year,
  • + 2 MVPP for having best average on one of three main simple statistical categories (points, assists or rebounds per game)
  • + 4 MVPP when second player on that list is at least 10% worse (predominant bonus)
  • + 1 MVPP if a player is having by far better year than previous one (+ 3 tendex)
So the final version of formula goes like this:

MVP Points = Tendex + Wins/2 + GamesPlayed/4 + (Wins-Wins_year_before)/4 or (-1) when team won less games than previous year) + bonuses.

Example using stats from 2004/05 season:
Let's consider just a few important players to shorten and simplify that example.

Kevin Garnett had the best tendex in the league (32,4), he played in 82 games. Wolves were 9th in West and they won 44 games. In previous season they had 58 wins. In addition Garnett was the best rebounder in the league. MVPP = 32,4 (tendex) + 22 (44/2 Wins) + 20,5 (82/4 games played) - 1 (less wins than previous year) + 2 (bonus for the best average of rebounds per game) - 5 (his team was worse than at third place in its conference) = 70,9 MVP Points.

Let's check another candidate, Dirk Nowitzki. He played 78 games with third best tendex in league (28,2). Mavs won 58 games, 6 more than previous year, and they had third most number of wins in West. MVPP = 28,2 + 58/2 + 78/4 + 6/4 (improvement in wins category) = 78,2 MVPP

Many saw Shaq as a leading candidate for MVP award so using the same formula: MVPP = 25,5 + 59/2 + 73/4 + 17/4 (improvement in wins category) + 5 (bonus for having the most wins in East with huge advantage) = 82,5 MVP Points.

But the MVP was won by Nash. Did the formula suggest the same thing? MVPP = 24,3 + 62/2 + 75/4 + 33/4 (improvement in wins category) + 4 (by far had the best assist average per game) + 2 (Suns won its conference but with close margin) + 1 (his own tendex has improved by 4!) = 89,3 MVP Points.

Many of experts argued that Nash wasn't even the best in his own team. But formula works even in that case: Marion - 87,7 MVPP and Amare - 88,5 MVPP so less than Nash.

Of course many doubters will think that it is just coincidence. Wrong. Because using that formula we would predict correctly the winner of the MVP award from the previous full twenty seasons!
[See the attached file historychecked.html]

Bold prediction for 2005/06:
(Statistics and standings through games of 2/22/06)

Of course all players from actual season don't have enough games played to qualify (70) but we can just assume that all their averages will not change until the end of season so tendex will also be the same.

To compare team's wins to previous season we have to estimate the number of wins using actual percentage of wins. For example Pistons have 43 wins and 9 losses. Their winning percentage is 0,827. Assuming they continue this pace they will have 82*0,827=67 (67,8 to be exact) wins at the end of season.

Finally using my formula let's predict top MVP vote-getters:

1) Chauncey Billups - 85,9 MVP Points
2) Dirk Nowitzki - 81,7 MVPP
3) LeBron James - 77,8 MVPP
4) Steve Nash - 77,3 MVPP
5) Tim Duncan - 75,7 MVPP

So providing the Pistons continue their pace for around 67 wins, historically Billups will be an easy choice. But the race for second place is very close with four possible players - Duncan, Nowitzki, LeBron and Nash. Nowitzki's only advantage right now is that his team is atop West but who knows for how long?

Editor's Note: Outstanding work Michael! We'll check back with 'wiLQ' after the regular season to get his final predictions.

Rate this Feature
Poor   Fair   Good   Excellent

Enter your comments in the box

Email (optional)

We want your feedback! Tell us your thoughts

Copyright ?2006 by www.nbatop.com, All Rights Reserved