Categories and Correlations - Part 3

...Continuing from Categories and Correlations - Part 2 (Please read that post first)

There are 126 different groups of 5 categories. Each group of 5 is defined by the 10 relationships between its variables. For example, if we pick categories A,B,C,D, and E, then we need to look at the correlations: 1. A vs. B, 2. A vs. C, 3. A vs. D, 4. A vs. E, 5. B vs. C, 6. B vs. D, 7. B vs. E, 8. C vs. D, 9. C vs. E, and 10. D vs. E

The strength of how well 5 categories work together is defined as the average of these 10 correlations. I ranked the 126 possible groups of 5 categories on this average correlation. The average correlation of the 126 possible groups ranges from -.15 to .49. Any average below 0 means a player who's good at any one of the 5 categories is likely to be bad at the remaining 4. The group of 5 at the top of the list, the one with an average of .49 is quite special, b/c not only does it have the best average correlation (best by a large margin, #2 has an average of .41), but it is the only 1 in 126 where none of the 10 correlations between the 5 categories are negative. All 10 are nicely positive, the minimum of the 10 correlations is .27. The 5 best categories to go for are: FT%, 3 Pointers Made, Points, Assists, and Steals.

So that's it for categories and correlations. You now know which 5 categories around which to build a team.

Categories and Correlations - Part 2

...Continuing from Categories and Correlations - Part 1 (Please read that post first)

First, a necessary note on turnovers: From now on turnovers will refer to the opposite of turnovers. In the last post we saw that points and turnovers had a very strong correlation. if we use the opposite of turnovers, then points and turnovers become strongly negatively correlated. For the purposes of determining which categories to go for, strong correlations will now always be a good thing and negative correlations will now always be a bad thing.

There are 9 categories and there are 36 unique pairs of categories. We have 36 category to category correlations to analyze. The 2 scatter plots in the previous post show the category to category relationships for 2 of these 36 combinations. I won't show any more scatter plots, but here are the correlation coefficients for all 36 possible pairs of categories:

The negative relationships are in red and the positive correlations are in black. The 5 highest correlations are (These aren't the strongest correlations, just the strongest positive ones):

• Steals vs. Assists - .70
• Blocks vs. Rebounds - .67
• Steals vs. Points - .64
• Points vs. Asssits - .59
• Points vs. 3 Pointers Made - .53

You might be thinking, "Great let's use those 5 to pick which categories to go after". The above 5 strong correlations involve 6 categories (everything except FG%, FT%, and TO) and as a group of 6, there are many negative correlations between them. (Blocks vs. Assists, Rebounds vs. 3 Pointers, Etc.)

So back to the question of determining which 5 categories to go after.. We can't just look at the correlations between pairs of categories. We need to see how 5 categories can work together.

There are 126 ways to pick 5 categories from the 9. We need to pick the best group of 5 categories from the 126 different possibilities.

To be Continued...

Categories and Correlations - Part 1

Continuing from Fantasy Basketball Introduction... (Please read that post first)

There are 9 categories and you need to pick which 5 to go for. Take a look at this graph showing the points and turnovers from last season for the top 150 fantasy players:

With the exception of Dirk Nowitzki, every player over 2,000 points averaged at least 2 turnovers per game. As you can see from the above graph (and already know from common basketball sense), the players that score a lot of points are also the players that turn the ball over the most. So for starters, if you decide to go for points, you can't go for turnovers. And if you want to make your team good at turnovers, then you can't also try to be good in points.

By the way, points and turnovers have a 0.84 correlation coefficient. If you do not know know what a correlation coefficient is, here's a link to the wikipedia article about them: http://en.wikipedia.org/wiki/Correlation_coefficient

Unlike the strong correlation between points and turnovers, some pairs of the 9 fantasy categories have a strong relationship that can be used to our advantage. For example, here's a scatter plot showing the steals and assists for the top 150 fantasy players:

As you can see, there is a fairly strong positive correlation (As you already knew from basketball common sense). Players that get the most assists also tend to get the most steals.

To be continued...

Fantasy Basketball Introduction

The 2007-2008 NBA season starts on October 30, which means it's time to start thinking about your fantasy draft strategy! My posts will assume you are competing in a head to head league with the standard 9 statistical categories (FG%, FT%, Points, 3-Pointers Made, Rebounds, Assists, Steals, Blocks, and Turnovers) and standard 10 starting position slots (PG, SG, G, SF, PF, F, 2 C, and 2 Util).

Let's assume that you agree with your league's "projected" rankings. The big question is then, on draft day, besides picking the best rated player with each of your picks (and making sure you fill your starting positions), what else can you do gain an edge that will benefit you all season long (and throughout your league's playoffs)?

Some players have a good overall rating because they excel in a single category. Andrei Kirelenko is one such player (He's a blocks machine). If you always pick the best overall player available, you might end up with a team full of players that are only good at a few categories. You'll be destined to lose all your head to head matchups despite dominating your opponents in a few categories.

At the other end of the spectrum, some players have a good overall rating because they are decent in almost all 9 categories. For example, look at Manu Ginobli, Leandro Barbosa, and Teyshaun Prince's stats from 2006-2007; these guys were good in 8 or all 9 categories. If you always pick the best overall players, you might end up good talent in all 9 categories. You might have a decent regular season, but in the playoffs, you might meet a team that excels in 5 categories and is lousy in the other 4. Your team was good all season long in all 9 categories, but this team will likely beat you 5-4 and advance in the playoffs.

You want to be the that kind of team, the team that excels in 5 categories. You might not have the best regular season (and you must accept this). One team might draft to be good at all 9 categories. Against some of the worst teams in the league, that "all around" team might win 9-0 or 8-1. Your team won't be able to beat up on this lousy team, you'll just win 5-4 because unfortunately, b/c you're so good at 5 categories, you're going to be pretty bad at the other 4.

But, when the playoffs come around, and you go to play that "all around" team that won the regular season, you're going to beat them 5-4 and advance in the playoffs.

In my next posts (over the course of the next month). I will show you how to become a team that excels in 5 categories. I'll show you how to decide which 5 of the 9 categories you should go for and then I'll show you how to determine which players will suit your new strategy the best.

Since you will only be aiming to be good at 5 of the categories, the players that fit your team the best will not be the same as the highest overall rated players that every other team is trying to draft. Players that will fit well into your system may be available throughout every round of the draft.

Enjoy, I hope you like it!