With the college basketball season officially completed it’s now a time for reflection.
Players and coaches will think about plays that could have gone their way, and what could have been changed to change the outcome of their season.
For me, it means reflecting on a prediction I made before the season.
Really, things started two seasons ago during the redshirt years of transfers Tyree Appleby and Anthony Duruji. Using statistics of past transfers from their respective leagues (the Horizon League and Conference USA) to the high major ranks, I came up with an algorithm that would predict their statistical production for the Gators. It ended up being quite a success, in my eyes at least, as you can see here (you may need to clink the link to see all the images that hold the information):
In the preseason I made formulas based on the performance of past Horizon and C-USA transfers to predict the stats of Tyree Appleby and Anthony Duruji. Pretty stoked with how close they were. With so many transfers happening these kinds of predictions could be very useful. pic.twitter.com/K7Yx69bxLj
— Eric Fawcett (@Efawcett7) March 25, 2021
For Tyree Appleby, the numbers were off by 0.1 points, 0.3 assists, and 0.7 rebounds. For Anthony Duruji, the numbers were off by 2.1 points, 0.0 assists, and 0.2 rebounds.
With the predictions being pretty close, it was clear what I had to do when the Gators landed Brandon McKissic, Myreon Jones, CJ Felder, and Phlandrous Fleming in the transfer portal. I would need to fire up the old algorithm and try to make predictions once again.
If you want to read more about how I did it you can look at the prediction article here. I (hesitantly) changed up the method a bit from what I did the year prior, hoping it wouldn’t burn me. You see, the Appleby and Duruji prediction was all about looking at similar players in the past and how their production changed in order to make an accurate prediction. However, due to the free one-time transfer rule and the extra COVID year, we couldn’t exactly rely on any past data to make a one-for-one comparison with Florida’s incoming transfers.
I ended up weighing different play type data differently for each player, predicting what their role might be with this team and how it would change their production. It was a lot more manipulation by me, a curveball that had me a bit nervous regarding how close these predictions could be.
Without any more backstory, let’s get to the results and see how close the predictions were.
Phlandrous Fleming
Prediction:
10.5 Points
1.9 Assists
4.1 Rebounds
1.4 Steals
0.6 Blocks
33% Three-Point
25.8 Minutes
Actual Stats:
11.1 Points
2.1 Assists
4.4 Rebounds
1.5 Steals
0.7 Blocks
29.4% Three-Point
27.5 Minutes
Fleming was one of the pleasant surprises of Florida’s season, ending up second on the team in scoring and rebounding. Looking at the prediction–I’m pretty happy with how this one ended up. He outscored what the algorithm expected, and perhaps that’s because he played almost two more minutes per game than what was predicted. Assists, rebounds, steals, and blocks were pretty close too. Three-point percentage–not so much, and spoiler–that will be the case for a couple of these predictions.
Myreon Jones
Prediction:
11.5 Points
1.9 Assists
2.9 Rebounds
0.9 Steals
0.1 Blocks
38.7% Three-point
29.5 Minutes
Actual Stats:
8.5 Points
1.6 Assists
2.8 Rebounds
1.3 Steals
0.2 Blocks
32.1% Three-Point
27.9 Minutes
Myreon Jones’ season ended up rather disappointing, but I will say that the algorithm predicted a dip from the 15 points per game he scored at Penn State so maybe we should have seen things coming a bit. The big number that jumps out here is the 38.7% three-point percentage that was predicted (a dip from his 40% mark at Penn State) and then the 32.1% stroke he actually had. If he was able to get closer to the predicted percentage, I bet his points would have been very close to what the algorithm thought. But, we’ll never know. Once again–the prediction did pretty well on assists, rebounds, steals, and blocks.
Brandon McKissic
Prediction:
7.2 Points
2.3 Assists
3.0 Rebounds
1.2 Steals
0.1 Blocks
38.2% Three-Point
21.2 Minutes
Actual Stats:
5.8 Points
1.8 Assists
2.5 Rebounds
1.1 Steals
0.1 Blocks
24.3% Three-Point
24.2 Minutes
Let’s get right to it, that three-point percentage prediction was WAY off. However, once again the prediction was pretty close on a lot of the other numbers. McKissic was also able to play more minutes than predicted, though it didn’t result in him blowing any of the statistical categories out of the water.
CJ Felder
Prediction:
6.3 Points
0.7 Assists
5.2 Rebounds
1.1 Steals
1.6 Blocks
32.1% Three-Point
20.8 Minutes
Actual Stats:
3.5 Points
0.4 Assists
2.3 Rebounds
0.4 Steals
0.6 Blocks
38.9% Three-Point
12.1 Minutes
Injuries really disrupted Felder’s season, rendering the predictions pretty useless here. Could I prorate both numbers to a per-40 minute number and prove that from a per minute standpoint the predictions were actually pretty good? Maybe, if I really wanted to, but I’ll just let the raw numbers speak for themselves and allow you to draw the conclusions you’d like. When Florida landed Felder I thought they landed a very good piece and should he stick around and be fully healthy next year I think fans will be extremely happy with his play.
There it is, folks!
Overall, I’m pretty impressed with how using numbers can make for pretty accurate predictions when it comes to transfers. Considering Todd Golden and his staff and analytically driven and savvy, you can see pretty quickly how they can use numbers to their benefit when it comes to the transfer portal where they are currently very active.
I will call this year’s predictions a success, and therefore with two years of successful predictions I will unquestionably have to do predictions again for the next wave of transfers.