Attempting to use AI to help create a complex schedule local sports league

[ekym]

I tried posting this to some of the popular reddit subs and they just stuff it into a monthly thread and ignore it, so I thought I'd try my luck with a community forum.

I have been working on this with a few different AI options without luck yet, OpenAI, Copilot, ChatGPT and DeepSeek.
So far I have found limited help, for example When I try creating a prompt the results are misunderstood, example, mistakenly pairing team 20 with team 6 and team 20 with team 9 while excluding team 21 in week 2.

I have been reducing the ask, attempting to make it easier for the AI's so my prompt has been changing day to day, I've attached an example of last years manually created schedule and my prompt that has failed to generate anything useful so far as it is today.

I would welcome any ideas, it's possible that AI just is not ready for something like this. It could also be that I'm too new at using AI and don't know what I'm doing. I created last years schedule in about 2 weeks by myself manually. But sometimes we get new teams at the last minute and it would be nice to get AI help generating one in less time.

thanks

 

[Perry]

Hello there! I think you need to divide these instructions into several steps, if possible. Also, have you tried o3 reasoning from openAI?

@Ant what do you think about the prompt below?

I need to create a schedule for a Horseshoe league.

Horseshoe League Schedule Creation Prompt Template:

There must be no duplicate matches throughout the schedule other than the 8 bye teams listed below.

League Name: [Northern Illinois Horseshoe League]

Number of Teams: [24]

Divisions: [Red Division: Teams 1–8, White Division: Teams 9–16, Blue Division: Teams 17–24]

Bye Teams: [Red Division: Teams 7-8, White Division: Teams 14, 15, 16 Blue Division: Teams 22, 23, 24

Schedule Rules:

RED will play 3 matches against White and 3 matches agains Blue every week.

There will be 2 matches between White and BLUE.

There will be 4 matches betwen the bye teams only.

There will be a total of 12 matches each week.

Time Frame:

Season Duration: [10 weeks]

Team names are as follows, Team number Location name Team name

Example: Team 1 - Riverside River Rats

RED teams (1-6) will play against WHITE (9-13) or BLUE (17-21) in a randomized order There will never be a duplicated match for the entire schedule.
Example: if teams 1 and 9 play in week 1, they will not play again in the remaining weeks

Bye Teams (7, 8, 14, 15, 16, 22, 23, 24) only play against each other.

RED Division is as follows:

Team 1 - Riverside River Rats
Team 2 - Hidden Tap Golden Shoes
Team 3 - Kief's Reef Krackens
Team 4 - O’Leary’s Pub Bud's Shooters
Team 5 - Mob-B-Dicks Crew
Team 6 - Bolt On Inn Boomers
Team 7 - Red Division Bye Team 1
Team 8 - Red Division Bye Team 2

White Division is as follows:

Team 9 - Horizontals Hookers
Team 10 - McHenry Moose Repeaters
Team 11 - Whiskey Bungalows Ringers
Team 12 - McHenry Moose MeRyders
Team 13 - Hidden Tap Silver Shoes
Team 14 - White Division Bye Team 1
Team 15 - White Division Bye Team 2
Team 16 - White Division Bye Team 3

Blue Division is as follows:

Team 17 - Kief's Reef Hitmen
Team 18 - Moe-B-Dicks Ringmasters
Team 19 - McHenry Moose Knucklers
Team 20 - Bolt On Inn's Stingers
Team 21 - Horizontals Green Barons
Team 22 - Blue Division Bye Team 1
Team 23 - Blue Division Bye Team 2
Team 24 - Blue Division Bye Team 3

Additional Notes:

I have completed one week.

11 1 13 2 17 3 12 4 5 18 20 6 10 19 9 21 23 7 8 24 14 16 22 15

Bye team matchups can be the same every week, or different as they are fictional teams it will not mater

I will manually adjust for venue constraints.

Note that the sum of all of the team numbers in each week will always equal to 300.

Pause and wait for approval after each week.

 

[Ant]

Hey there! Creating a horseshoe league schedule with AI is definitely a challenging task, but it's also a great opportunity to learn more about AI's capabilities and limitations!

- Clarify the Prompt: The confusion might be due to the complexity of the league rules. Break your prompt into smaller steps, focusing first on pairing teams correctly within divisions and bye teams.

- Use Iterative Adjustments: AI can improve its output with feedback. After each attempt, if the schedule is off, refine your instructions to correct for errors like the team 20 and team 9 pairing mix-up.

- Explore Other Tools: Sometimes, one tool might not be the perfect fit for every task. Have you looked into tools like Google's OR-Tools or Microsoft Solver Foundation? They're designed for operations research problems, which this schedule optimization sounds like.

- Think About Division Structure: The league rules seem to prioritize team numbers instead of the division they're from for matchups. Try emphasizing division or venue constraints early in your prompt to guide the AI better.

- Humor and Lightness: Like AI debugging, getting this schedule right might be like trying to herd cats. It's not impossible, just requiring a bit of finesse.

- Open AI's Reasoning: Although you're already using some of their models, have you tried specifying a step-by-step reasoning process in your prompt? This might help AI to structure the schedule more accurately, addressing each rule as you go.

Your effort to streamline the scheduling with AI is commendable, and it's not just about AI readiness – it's also about honing your skills with AI tools. Let's not give up; sometimes, the solution is hidden behind the next iteration! Keep tweaking and learning – small changes can lead to big successes. If you're willing, share more about how you're adjusting your prompts or other tools you're exploring!

 

[ekym]

Hello there! I think you need to divide these instructions into several steps, if possible. Also, have you tried o3 reasoning from openAI?

Thanks for the advise, I was not aware that the o3 reasoning from openAI was a thing. I checked it out and it seems better at understanding the ask.

 

[Perry]

Thanks for the advise, I was not aware that the o3 reasoning from openAI was a thing. I checked it out and it seems better at understanding the ask.

For complex prompts o3-mini-high is very good. But, as said before you should divide that prompt into smaller batches.