This post is for Ottoneu Basketball Categories leagues.
I just spent a little time this evening assembling weekly team totals across all six categories leagues that I could find (leagues 1, 21, 23, 24, 26, 34). There were 72 matchups, including (2) 4-4 ties.
Here are some summary statistics (I excluded one team who didn’t set a lineup in Week 1 and wound up with 0’s for everything):
It’s a small dataset, so there’s limited amount of what can be done with it until we get some more weeks worth of data. But I’m going to be playing around with it some and see if there are any insights to discover. For example, I find that BLK, FG%, and FTM are the strongest predictors (i.e., only statistically significant) of likelihood of winning. PTS and STL are the weakest. This is based on a multivariate probit regression with week fixed effects.
For the next piece of analysis, I wanted to test the relative likelihood of winning a matchup conditional on winning each category. So I created a set of binary variables (1=won the category, 0 if not) and used those as the covariates in a single linear probability model. In addition, I used league fixed effects to account for all unobservable characteristics between leagues. In addition to the 9 categories, I included minutes played as an indicator as well. This was the resulting model output (constant not reported):
As you can see, all of the coefficients are statistically significant and in the expected direction (e.g., turnovers is negative and all others are positive). Most of the coefficients are all around approximately 0.20-0.25, with the exceptions of points (0.323***) and rebounds (0.343***). In other words, winning points and rebounds are stronger predictors of winning the match-up than the other seven categories.
Because it’s a linear probability model, the coefficients are easy to interpret: leading a category increases your likelihood of winning by X% (where X is the coefficient translated from decimal to a percentage). For example, leading in rebounds translates to about a 34% increase in winning matchup, whereas leading in turnovers decreases the probability of winning by 22%.
Before getting into the post week 4 summary statistics, just wanted to report that previous versions of the summary statistics did not include leagues 40 and 41. I discovered that there were additional category leagues after the “Browse League” function was added to the site earlier this week.
So anyway, here are some summary statistics after Week 4. First, all teams (excluding teams that do not record a MP in their entire weekly lineup:
Note that matchup winners are more likely to lose TOV than win it, the only category for which that is true.
In case you were wondering the frequency by which the 192 matchups so far have been decided (includes matchups where one team didn’t set a lineup), here are the counts and frequency of those category matchup totals:
(3) 4-4 Ties (1.6%)
(76) 5-4 Wins (39.6%)
(60) 6-3 Wins (31.3%)
(36) 7-2 Wins (18.8%)
(15) 8-1 Wins (7.8%)
(2) 9-0 Wins (1.0%)
Also, there are 9 remaining undefeated teams across the eight category leagues:
Fresh Gumbo (League 1)
Lifeguard Kelly (League 21)
Mo Money Mo Bamba (League 21)
We Shall All Be Hield (League 23)
Ehlo, Sweet Chariot (League 24)
Inazuma Lightning (League 26)
Roundball Rock (League 34)
Greek Freakonomics (League 40)
Ronen B.C. (League 41)
Note that there are also 10 teams that have yet to win a matchup, but I’m not going to name them.
Finally, in terms of predictors of winning the matchup within a multivariate model with matchup win as the binary outcome (regression results not presented), winning FTM has the strongest association to winning the matchup (followed by PTS and REB), while AST is the lowest (followed by STL and TOV). I’ll just add that this is generally consistent with what I’ve seen so far in other interrogations of these data: teams built around big men are having greater success than those built around guards.
For anyone wondering about the relationships between leading in one category versus another category, here is a raw correlation matrix to start the conversation:
So the reason why I set 600 MP as a threshold is that–with two exceptions in Week 1–every team that has won a matchup accumulated more than 600 MP. Here is the likelihood of winning a matchup based on MP in 50 minute increments (not controlling for quality of opponent):
[0, 600): 2 out of 29 (6.9%)
[600-650): 4 out of 10 (40%)
[650, 700): 3 out of 13 (23%)
[700, 750): 10 out of 33 (30%)
[750, 800): 25 out of 73 (34%)
[800, 850): 72 out of 135 (53%)
[850, 900): 75 out of 122 (61%)
[900, 950): 40 out of 61 (66%)
[950, 1000): 4 out of 4 (100%)
Or, alternatively, what those data look like in graphical form (probability of winning is on the y-axis, MP is on the x-axis):
Or to put it another way, in the 235 matchups where there was a winner (i.e., where the result was not a tie), 169 were won by the team that lead in MP (72%).
Click on File at the top of the screen and select “Make a Copy”;
Enter in the expected or projected values in the Yellow field (Column B); and
If you edit the formula in the Blue field (Column C), then it may not work correctly.
For purposes of illustration, the means for category leaders are already entered. As you’ll notice, the probability of winning each is 50%. If you were to enter 500 points, then the probability of winning the category increases to 75%; if you were to enter 400 points, then the probability of winning the category decreases to 14%. And so forth.
Please note that this doesn’t take into account the quality of your opponent; rather, it assumes an average league winner.
I’ve gone into some detail about why leading in minutes is so critical, but will add a few additional data points. First, leading in minutes is associated with a ~7x increase in the probability of winning one’s matchup. By category, leading in minutes is associated with:
12x more likely to win points
9x more likely to win rebounds
5x more likely to win assists
3x more likely to win steals
2x more likely to win blocks
1.4x more likely to win field goal percentage
5x more likely to win free throws made
1.5x more likely to win three point percentage
6x LESS likely to win turnovers
Here are the teams that have won their category in all 7 of their matchups so far:
Points:
Lifeguard Kelly (League 21)
Only Celtics in the Building (24)
Rebounds:
Mo’ Town Hustlers (1)
Buddy Hield Finals MVP (21)
Giannis Alphabet (23)
Roundball Rock (34)
StrangerAlps (34)
Spirits of St. Louis (40)
Bol? (41)
Assists:
Mo Money Mo Bamba (21)
It’s Darko and Hell is Hot (23)
Sandwich Hoss (24)
Roundball Rock (34)
Steals:
We Shall All Be Hield (23)
Bob Loblaw Lob Ball (24)
Lake Zurich Legends (26)
Roundball Rock (34)
Big Honey (40)
Field Goal Percentage:
Kazaam (24)
StrangerAlps (34)
Free Throws Made:
Buddy Hield Finals MVP (21)
Lifeguard Kelly (21)
Giannis Alphabet (23)
Three Point Percentage: None
Turnovers:
Dunkness on the Edge of Town (23)
Tally the Dubs (26)
Bol? (41)
Brockton (41)
Finally, there were a couple weekly records set this week in a couple of categories (Points and FG%). Here are the global weekly leaders for all nine categories (FG%, 3PT%, and Turnovers subject to minimums):
Points: 610 by Lifeguard Kelly (League 24) in Week 7
Rebounds: 241 by StrangerAlps (League 34) in Week 3
Assists: 165 by Sandwich Hoss (League 24) in Week 6
Steals: 46 by #StephBetter (League 1) in Week 4
Blocks: 42 by Hoopla (League 1) in Week 1
FG% (min 100 FGA): 56.5% by Giannis Alphabet (League 23) in Week 7
FTM: 127 by Steeltown Hookers (League 34) in Week 6
3PT# (min 50 3PA): 55.1% by The Center of Attention (League 41) in Week 1
