Venue: Littlejohn Coliseum, Clemson, South Carolina
Conference: ACC
Recent Form
The 13-13 (6-7) Seminoles visit the 20-7 (10-4) Tigers in Clemson, South Carolina for a early afternoon tip-off. Before we get into the picks, let’s observe the head-to-head stats comparison:
Warning: package 'tidyverse' was built under R version 4.5.2
Warning: package 'lubridate' was built under R version 4.5.2
── Attaching core tidyverse packages ──────────────────────── tidyverse 2.0.0 ──
✔ dplyr 1.1.4 ✔ readr 2.1.5
✔ forcats 1.0.1 ✔ stringr 1.5.2
✔ ggplot2 4.0.0 ✔ tibble 3.3.0
✔ lubridate 1.9.4 ✔ tidyr 1.3.1
✔ purrr 1.1.0
── Conflicts ────────────────────────────────────────── tidyverse_conflicts() ──
✖ dplyr::filter() masks stats::filter()
✖ dplyr::lag() masks stats::lag()
ℹ Use the conflicted package (<http://conflicted.r-lib.org/>) to force all conflicts to become errors
Warning: package 'gt' was built under R version 4.5.2
Rows: 8 Columns: 4
── Column specification ────────────────────────────────────────────────────────
Delimiter: ","
chr (2): Metric, Advantage
dbl (2): Home, Away
ℹ Use `spec()` to retrieve the full column specification for this data.
ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.
Last 5 Games Statistical Comparison
Florida State Seminoles @ Clemson Tigers
Metric
Home
Away
Advantage
eFG%
0.5
0.5
Florida State Seminoles
TOV%
0.1
0.1
Florida State Seminoles
ORB%
0.3
0.2
Clemson Tigers
FTR
0.2
0.3
Florida State Seminoles
ORtg
107.3
119.5
Florida State Seminoles
DRtg
109.5
107.5
Florida State Seminoles
Net Rating
−2.2
12.0
Florida State Seminoles
Pace
63.3
67.2
Neutral
Source: hoopR | ProPlotFits Analysis
Breakdown
Let’s skip the first four metrics since the two teams are basically even. The story is in the Offensive and Defensive Efficiencies and thus ultimately in the Net Rating. Looking at just descriptive statistics, Florida State on average scores 12 more points than they allow, while Clemson allows 2.2 points more than they score, creating a difference of 14.2. Clemson is also allowing 12 more points per game than Florida State.
Model Output
Model Predictions for Florida State @ Clemson
Model
Prediction
Model MAE
Win Probability
Clemson 50.6%
69%
Predicted Margin
Clemson by 0.3
9.5 points
Predicted Total
139.5 points
13.5 points
Our models give Clemson a 50.6% chance to win at home. The final score should either be 70-69 or 71-70.
Recognizing Model Uncertainty
It would be irresponsible to look at this output and take it for its word without contextualizing these outputs with what our model error (Mean Absolute Error, MAE) is saying.
Our spread model has an MAE of 9.5, meaning on average across all games in our test set, the predictions were off by about 9-10 points from the true margin. When say “Clemson by 0.3” we’re actually saying “In the midst of intense noise, there is the tiniest bit of signal indicating that Clemson will win but just barely”. We aren’t claiming precision here.
Why does the model favor Clemson despite FSU’s better stats? Let’s be honest and say the models aren’t actually favoring one team or the other. “This is as close of a match-up you’re going to get” is essentially what they are saying.
This doesn’t mean we don’t have an edge.
Comparing to DraftKings
Let’s look at the odds that DraftKings gives for this match-up and compare to our models.
Code
vegas_comparison <-tibble(Metric =c("Spread", "Total"),`Our Model`=c("Clem -0.3", "139.5"),Vegas =c("Clem -8.5", "143.5"),Gap =c("8.2 points", "4 points"),`% of MAE`=c("86.3%", "29.6%"),Confidence =c("Very High", "Low"))vegas_comparison |>gt() |>tab_caption("DraftKings Lines vs Model Outputs")
DraftKings Lines vs Model Outputs
Metric
Our Model
Vegas
Gap
% of MAE
Confidence
Spread
Clem -0.3
Clem -8.5
8.2 points
86.3%
Very High
Total
139.5
143.5
4 points
29.6%
Low
Edge Analysis
Before I go into what this table is showing, let’s breakdown how I’d classify the confidence of my edges (comparing the difference between Vegas and our model with our model MAE).
Elite (>100% of MAE): Highest conviction but doesn’t happen that often. This shows that Vegas is way off
Very High (80%-100%): Don’t miss these high premium picks
High (60%-80%): Strong and worth betting
Moderate (40%-60%): Definitely the place to incorporate additional information
Low (20%-40%): Maybe there is value, but it’s definitely not premium
Noise (0%-20%): Model and Vegas agree, so we ignore
Now here’s how our match-up stacks against DraftKings.
Totals: The difference between our model and Vegas is 4 points, which is less than 30% of our MAE. It is an edge, but not that great. I’d personally skip.
Spread: Now we got something interesting. Our model is showing that these two teams are dead even, and it even predicts that Clemson will win by less than a point. DraftKings thinks Clemson will win by at least 9, and all indications from our analysis conclude that that margin is unlikely to happen. I will definitely take this one.
The Picks
Total: Take the UNDER
According to each team’s pace, this game should average 65.3 possessions, and at 1.07 points per possession (average for college basketball), that’s 65.3*1.07*2 = 139.7 projected total just based on raw stats. As I said, I think I will pass.
Spread: Take Florida St +8.5
I just don’t see how Clemson has enough skill to beat a scrappy FSU squad by 9 points. FSU beats them in head-to-head stats match-up except for offensive rebounding.
Conclusion
I hope you enjoyed this breakdown. This is a drawn-out version of what I’m doing to discover edges analytically, but I wanted to show y’all how to think statistically and probabilistically rather than just going with gut feeling.
Data-driven betting looks like this. We “run the numbers”, then take a second to sit back and engage with the information before going of and placing bets. And I’m showing you my work, which is the difference between guessing and analysis.
Want to see these predictions daily with full model outputs and edge analysis? Subscribe for $25/month (reach out to peter@solplots.com) and lock in that rate forever (first 50 subscribers only)