This is my first time participating in ludum dare jam. Creating art, music, and sound effects is challenging for me, but I found the process really enjoyable! My game idea was inspired by this YouTube video (https://www.youtube.com/watch?v=1QGTpdzDE7Y) and I thought it would be cool to let the player control Tetris pieces to explore a world while keeping going down.
Hi everyone,
I hope you're all doing well after the Ludum Dare Jam! I'm looking for a bit of help. Does anyone happen to have a screenshot of the Ludum Dare main page from the recent Jam, showing the "Submission Time Ends" timer in the top-right corner?
I’m writing a blog to inspire my classmates to join future Ludum Dare events, and I’d love to include a screenshot to explain the extra hour for Submission clearly. Any screenshot of the main page with that timer would be perfect, even if it’s partial!
I’d really appreciate your kindness and support! I’ll be checking back here, so feel free to reach out anytime.
Thanks so much, and great job to everyone on their submissions! 🚀
Can't wait to join Ludum Dare 58!
This is the first time I've tried making "real" art myself for a game jam. I was surprised I could actually pull off some simple drawings, and they turned out pretty cute! Although it's a bit of a shame that I didn't have time to adjust the high difficulty, I really enjoyed this jam!
Hope you enjoy!
Let's GAMEJAM!!!
Recently, there have been discussions (discussion1, discussion2) about the use of AI coding tools in Compo. The posts mainly focus on compo spirits, but there are also some quantizable discussions for example whether AI coding tools are silver bullet or not, so I made an analysis. My statistics only apply to quantifiable metrics (such as scores or participation frequency) and cannot easily analyze aspects like spirit or honesty. Feel free to point out any errors.
Among the 290 entries, at least 32 used AI coding tools, 174 didn't, and 84 are unknown. Here's the the data.
Before we get into the data, I'll firstly explain how I collect the data. I categorized every LD59 Compo entry into three classes: AI (coding tools) used, No AI, and Unknown. An entry was considered to have AI used if keywords of AI tools appeared in the description or on the itch.io page, or if there were traces of AI agent usage on GitHub. If the description or the itch page explicitly stated that AI was not used, it was tagged as No AI. The remaining games were classified as Unknown.
I also manually checked the games after keyword matching. But the actual number of entries in the AI group could be higher than reported here considering the Unknown category.
Now let's dive into the data.
First, do AI group had a higher or lower historical score before AI tools became widespread? Let's look at the data from LD38 to LD56, since generative AI started gaining popularity after the 57th event.
| Group | No History Before 56th Ratio | Mean Historical Score | Mean Historical Entry Count | | --- | ---: | ---: | ---: | | AI used | 28.1% | 3.3652 | 3.9688 | | No AI | 35.1% | 3.4609 | 4.5575 |
| Comparison | Metric | Difference | Welch p | MWU p | | --- | --- | ---: | ---: | ---: | | AI used vs No AI | mean historical score | -0.0958 | 0.3326 | 0.5557 | | AI used vs No AI | mean historical entry count | -0.5887 | 0.5678 | 0.8874 |
The p values are all greatly above 0.05, which means there's no significant evidence that the skill and participation counts of the two groups are different historically.
Next, let's look at how AI usage is associated with the quality of Compo entries in LD59. I analyze scores rather than rankings here, as I believe scores reflect the true quality of a game better than rankings do.
| Metric | AI Mean | No AI Mean | Difference (AI - No AI) | Welch p | MWU p | | --- | ---: | ---: | ---: | ---: | ---: | | Overall | 3.55 | 3.46 | 0.09 | 0.3717 | 0.3125 | | Fun | 3.37 | 3.29 | 0.08 | 0.4651 | 0.3379 | | Innovation | 3.58 | 3.34 | 0.24 | 0.0154 | 0.0099 | | Theme | 3.78 | 3.61 | 0.17 | 0.0928 | 0.1512 | | Graphics | 3.52 | 3.53 | -0.01 | 0.9842 | 0.9645 | | Audio | 3.40 | 3.35 | 0.05 | 0.6929 | 0.8228 | | Humour | 3.12 | 2.87 | 0.25 | 0.2283 | 0.2080 | | Mood | 3.45 | 3.43 | 0.02 | 0.9188 | 0.8039 |
We can state that compo games that use AI seems to have higher innovation score than others.
| Comparison | SD Difference | Variance Ratio | Brown-Forsythe p | | --- | ---: | ---: | ---: | | AI used vs No AI | +0.1366 | 1.7857 | 0.0329 |
Also, compo games that use AI have a larger variance than others.
Now let's look at AI usage by rank. The curve starts at the 30th entry to avoid being too sensitive to single entries.
Entries that used AI seem to appear near the top more often, but the sample is too small to draw a firm conclusion.
In conclusion, among the 290 entries, at least 32 used AI coding tools, 174 didn't. There is no significant difference in historical skill level or participation counts between the AI group and the non-AI group.
However, entries of AI group have a higher innovation score and larger variance than others.
I am not against the use of AI coding tools. I believe that as long as AI is used to enhance the experience rather than being abused, it is perfectly fine. As for compo spirit, I view compo as a challenge to make a game solo in 48h with any tools, including AI tools. As long as the game is freshly created, it's fine to use AI. But I can understand that many people would disagree with me since every earthman are arguing about AI usage in art (game is art!) nowadays. If AI coding tools are banned in Compo, I'll manually code again happily.
AI will only get stronger in the future, and I think we need to reach a consensus on its usage. According to the current rules, AI is allowed. But this rule is quite old, and generative AI was not widespread back then. Therefore, I think it is necessary for Ludum Dare to update its rules to specify its attitude on Gen AI. If Gen AI should be banned, the rules should be updated to reflect that; if it is allowed, it would be best to require developers to disclose whether they used Gen AI, just like itch.io does.
Besides all this, I am not an expert in data analysis, so everyone is welcome to do their own analysis using this data (it's basically the same as the data at the beginning of this post).