Share
Created By
amitc

amitc

Home » Challenges » Appian World 2021 Hackathon » DUNK (Smash Squad Stats)

DUNK (Smash Squad Stats)

A site that empowers Super Smash Bros. Ultimate players by providing game statistics and intelligent prediction tools powered by Appian and Google Auto ML.

Inspiration
With everyone remote, I began co-hosting a happy hour on Fridays where friends and I play each other in the game Super Smash Bros. Ultimate.
I began tracking salient statistics from our matches, such as winners and knockouts. Using Excel, I aggregated these stats and provided a weekly summary of how everyone did. Many players enjoyed seeing their week-by-week performance, and this inspired me to begin tracking the data in Appian.

How I Built DUNK

My initial goal was to simply migrate the data into a relational database and display the data in Appian. Using the ease and power of SAIL, I quickly built various dashboards that provided similar value to the Excel reports.

The next step was fleshing out leaderboards to rank players against their peers. I was motivated by two key factors. The first factor was seeing how games such as chess and competitive multiplayer games ranked players. The second motivating factor was reading articles on gamification. I modeled a ranking system based on Elo in Appian and automated those calculations. I also built an achievements system that allowed me to highlight key milestones that players achieved. All this combined gives the players awareness into how they are doing in Smash Bros. Ultimate and incentivizes them to train their favorite characters if they wish to do so.

The final step in building this application was figuring out how I could use all this data to help players decide who they want to use. With over 80 playable characters, choosing which ones to use is a daunting task. The reports only provide a birds-eye view of the trend of the Smash metagame and doesn't take into account factors such as players' playstyles. I decided this use case would be a good candidate for machine learning.

I trained two models using the match data I recorded. These models can predict which character a player would have success against other opponents and predict the outcome of matches. This data, plus the records and reports in DUNK, allows players to learn character matchups and take action by deciding which characters to train.

Technologies I Used I built the site using the latest version of Appian. The machine learning models were built and trained using Google Cloud Auto ML. I used a connected system to integrate Google with Appian.

Challenges
The biggest challenge I faced was deciding how to rank players. Showing simply the win-loss stats of each player didn't tell the whole story. I needed a rating method that encourages players to improve their skills without heavily penalizing them in the early stages of training. As discussed previously, I ultimately decided to build a rating method based on Elo. Appian's low code toolkit allowed me to create this system reasonably quickly and iterate on it.
The second challenge was deciding how to present all the data to the players. I wanted to avoid overwhelming players that wanted to use DUNK. I chose to use dashboards that provided rankings and other leaderboards to give users an initial idea of where they stood. If they decided to get more information, I leveraged records to provide additional knowledge and awareness on individual players and their characters.

Future of DUNK
I'm proud of this application, and I have received high praise from my friends who play Super Smash Bros. Ultimate with me. Future enhancements can include RPA integrations and exploring how I can use Appian to import data I collect (for MVP, I import directly into the database). I also plan to explore using OCR to automate data capture.

"

Anonymous

© 2024 Appian. All rights reserved.