b(AI)ndScores Predicts the 2024 DCI Season

It's officially the first week of June so you know what that means.....the new drum corps season is only weeks away.

As of writing, we are less than two weeks from the Season Premiere event once again being held in Rockford, MI on June 26th. That show kicks off the 2024 DCI season consisting of 89 competitive events in 34 states across a period of just over six weeks. The season will conclude as always with the DCI World Championships at Lucas Oil Stadium on August 8-10.

The beginning of a new season always brings a lot of intrigue into competitive storylines, many of them a continuation from the previous season. 2024 will be no different with a number of competitive groupings that will be fun to follow over the course of the season. Some of these storylines include:

• Can Blue Devils continue their dynastic run and claim DCI's first ever streak of four consecutive championships?

• Will Mandarins remain among the top echelon of corps following a strong 2023 that catapulted them into their first ever top six finish?

• How will Santa Clara Vanguard fare in their first season back after being inactive for the 2023 season?

• Can Troopers and Colts continue to climb the competitive ladder after both corps achieved consecutive appearances in finals for the first time in over 25 years?

• The battle for the 12th and final spot in finals looks to be a bloodbath yet again. Who will come out on top by the end of the season? Will Blue Knights defend their spot in the top 12 following their return in 2023, or will Crossmen, Pacific Crest, Sprit of Atlanta, or Madison Scouts challenge for an opportunity to perform on the final night of the season?

Since we won't receive definite answers on a lot of these questions until mid-August, we thought it could be fun to take a stab ourselves and predict final scores based entirely on historical data. The result is an extensive machine learning model with predicted scores and placements, for both overall scores and the five caption awards. The model was built and trained using over 1,000 round-by-round scores from DCI World Championship events dating back to the 2008 season. These scores were then used to find final metrics for the 475 corps that competed at World Championships in this time frame. A total of 28 metrics and variables were computed and used such as average scores and placements in each caption; each corps' score and placement at the previous two World Championships; and each corps average score, placement, and number of finalist appearances in the previous five World Championships. It is also important to acknowledge that this model is strictly built on data and does not consider intangible factors such as show design, staff changes, and changes in corps membership and talent.

In training, iterations of the model that predicted final score performed with an accuracy in the range of 90-95% , while the model that predicted final placement performed with an accuracy in the rage of 91-96.5%. This high accuracy shouldn't be terribly surprising to anyone who has followed DCI scores throughout the years. A lot of a corps' success in a particular season can be closely tied to successes in recent seasons, as corps rarely jump or drop a significant number of placements between seasons. Case in point, in the last 30 years, only one corps has won a championship after placing lower than fourth the season prior: The Cadets who won in 2011 after placing fifth in 2010. Similarly, since 2008, only six corps have improved by five or more placements from the prior season to finish in the top 12. All that said, a combination of caption and historical scores alone has high predictive accuracy for the models.

The predicted results will be released over a series of three posts on our Instagram page from June 12-14 and updated here accordingly. After all results have been released, we will add some analysis and explanation at the end of this piece. Note that we will only be sharing predictions for the top 15 corps - the reason being is that predictive accuracy is much lower for lower tier World Class and Open Class corps since there is much more placement variability among these groups year to year. It would not be fair to rank these corps when their final 2024 placements may vary widely from the model’s predictions.

So without further ado, here are the results...

2024 Placement Predictions

1. Blue Devils - 98.264

2. Bluecoats - 97.300

3. Boston Crusaders - 97.225

4. Carolina Crown - 96.613

5. Santa Clara Vanguard - 95.694

6. Mandarins - 93.463

7. The Cavaliers - 93.189

8. Phantom Regiment - 92.61

9. Blue Stars - 90.427

10. Troopers - 89.963

11. Colts - 89.748

12. Blue Knights - 86.679

13. Crossmen - 86.041

14. Spirit of Atlanta - 84.963

15. Pacific Crest - 83.817

General Effect Caption Award

1. Blue Devils - 38.957

2. Bluecoats - 38.719

3. Boston Crusaders - 38.671

4. Carolina Crown - 38.403

5. Santa Clara Vanguard - 38.101

Brass Caption Award

1. Blue Devils - 19.612

2. Carolina Crown - 19.354

3. Boston Crusaders - 19.319

4. Bluecoats - 19.312

5. Santa Clara Vanguard - 19.018

Color Guard Caption Award

1. Boston Crusaders - 19.596

2. Blue Devils - 19.576

3. Bluecoats - 19.533

4. Carolina Crown - 19.082

5. Santa Clara Vanguard - 18.963

Percussion Caption Award

1. Blue Devils - 19.422

2. Boston Crusaders - 19.312

3. Santa Clara Vanguard - 19.247

4. Carolina Crown - 19.171

5. Bluecoats - 19.132

Visual Caption Award

1. Blue Devils - 19.674

2. Boston Crusaders - 19.417

3. Bluecoats - 19.415

4. Carolina Crown - 19.266

5. Santa Clara Vanguard - 19.101

Summary

We know what everyone is probably thinking after posting these results - aren’t these placements nearly identical to 2023’s finals results? Yes, it’s true, the model was fairly conservative with its selections, but it only reinforces the point that a corps’ prior year score and placement has high predictive value. When it sees that Blue Devils and Bluecoats have placed first and second respectively in three consecutive seasons, it feels like a fairly safe assumption that the same will likely happen again in 2024.

The model picking Boston Crusaders to place back ahead of Crown in 2024 despite falling to fourth last year was an interesting selection. We imagine this was largely influenced by Boston’s higher visual scores, which we determined has a strong correlation with a corps’ final score. Regardless, it will be fun to watch these two corps go back and forth this summer and perhaps even challenge Bluecoats or Blue Devils for one of the top two spots.

As mentioned earlier, Santa Clara Vanguard’s return is one storyline everyone is looking forward to this summer. That said, their absence in 2023 makes their season hard to predict. We used their 2022 scores for the purpose of this model while keeping in mind that many historic corps have struggled to immediately find their same competitive success after a return from being inactive. However, SCV was able to bring back a number of key staff and designers from the 2022 season, so expectations are high that they should at least return to the top 12. A fifth place finish may seem high for the 2018 champions, but also not totally unrealistic.

Moving to the 6th through 12th sees a bit more shake up with The Cavaliers and Phantom Regiment swapping places in 7th and 8th and Blue Stars moving back ahead of Troopers and Colts to 9th. In both cases, the model likely took historical success and average scores and placements from the past five competitive seasons into account. 2018 and 2019 saw The Cavaliers finishing in the top six for the third and fourth consecutive years, while Phantom fell to 11th and 12th. Meanwhile for Colts and Troopers, both corps were only second year consecutive finalists in 2023 while Blue Stars have been a mainstay in the top 12 since 2013. Taking these factors into consideration, the model gave the edge to The Cavaliers and Blue Stars while also noting how close these competitive groupings were in the 2023 season.

Everyone would likely agree that the competition for the last spot in finals will once again be a top storyline in the 2024 season. The model currently gives the edge to Blue Knights to remain in 12th for the second consecutive year, but with only a 2.8 spread between them and Pacific Crest in 15th. All four corps in the 12th to 15th placements were in contention for a spot in finals at one point or another last season, and given this predicted close spread, we can likely expect an up and down battle again this season.

Before we put a bow on this piece, there’s one final interesting thing to note about these predictions. As mentioned earlier, two models were made and tested - one that predicted scores and one that predicted placements. The predictions above were generated with the score model since it had the higher accuracy, but for the most part, the model that predicted placements ended with nearly identical rankings as the score model. Two cases of it differing though was it having The Cavaliers above Mandarins and Colts above Troopers. Given these disagreements and the close predicted spreads, we can perhaps expect that these two pairs of corps will also be an exciting battle to follow this year.

Only time will tell how accurate these predictions will be for the 2024 season. One thing is for certain - we all love the competitiveness, close races, and shake-ups in any season. With these rather safe predictions in mind, it’s safe to say that we’ll be rooting against the grain and for the model to be wrong in a number of places to make for a fun and exciting summer.