I built an RPG-style Go training site for kyu players — looking for honest feedback

Hi everyone,

I’m Jason — a 6-dan amateur and a Go teacher with 20+ years in the classroom — and over the past months I’ve been building a training site called Go Odyssey. I’d really value some blunt feedback from this community before I push it further.

It started from a frustration I kept seeing in my own students: people solve tsumego for a week, lose steam, and quietly drift away. They’re rarely short on ability — they’re short on a reason to keep showing up.

So I tried building something a bit different:

  • A 3-minute rank test (no signup) that places you on a provisional range

  • An explorable RPG-style world map — solve problems to lift the fog, unlock regions, beat area “bosses,” level up

  • A “Weakness Radar” that tracks your accuracy by topic (life & death, tesuji, direction…) and feeds you tailored daily problems instead of random ones

It’s still a serious training tool underneath — the game layer is just there to make consistent practice less of a chore. It’s free to use, and aimed mainly at the 10k–1k range.

Try the rank test here (no account needed):

I’m genuinely posting this for feedback, not as polished marketing. I’d love your honest take on:

  1. Does the rank test feel accurate to your actual strength?
  2. Is the English presentation clear, or are there rough spots?
  3. Does the “map + weakness radar” approach feel useful, or gimmicky?

Brutal honesty very welcome — that’s the most useful thing you can give me right now. Thanks!

7 Likes

unusable cause character creation doesn’t even show whole thing even with maximum zoom out.
when sharing ai made products please let up know beforehand so we don’t go in with high expectations.


forgot to mention:
rank test is somewhat close, could be improved more, and having more questions there would help, and I believe rather than regular tsumegos, having full board tsumegos are much better for rank test

4 Likes

Thanks for the brutal honesty and the screenshots! This is super helpful.

  1. UI Bug: Wow, I didn’t catch that character creation issue on that screen resolution (1366x686). You’re completely right, it makes the site unusable. I’ll fix the responsive design for the character creation screen ASAP so it scales properly without needing to zoom out.

  2. AI Art: Apologies for not mentioning it beforehand! Yes, the visual assets/avatars are AI-generated to help me bring the RPG concept to life as a solo developer. I’ll make sure to be more transparent about that.

  3. Rank Test: Glad to hear the rank test felt somewhat close! I love the idea of adding “full board tsumegos” for a more realistic strength assessment. I’ll definitely look into expanding the question pool with those.

Thanks again for taking the time to test it out and giving me such solid feedback to work on!

2 Likes

will wait for an update to do more thorough tests.
btw, I’d suggest looking more into responsive web design with proper viewport breaking, by doing that you can make the site fit across 4 major screensizes, phones, tablets/large phones(or landscape modes there), computer monitors, and monitors larger than hd

3 Likes

Thanks, that’s really helpful. You’re right that this seems to be a broader responsive layout issue, not just a one-time bug. We’ll take another look at the character creation screen across phones, tablets/landscape, normal desktop screens, and larger monitors so everything stays usable without having to zoom way out.

2 Likes

Quick update! :rocket:

I just pushed an update to fix the responsive layout issues based on your excellent feedback. I set up proper viewport breakpoints to cover all 4 major screens:

  • Phones: Switched to a single-column layout with scrollable cards and a sticky bottom button.

  • Tablets / Landscape: Kept the 2-column layout but scaled down the preview, spacing, and paddings.

  • Standard Desktops: Maintained the original layout but added a dedicated max-height breakpoint specifically for shorter viewports (like your 1366x686 resolution) so headers and previews scale down properly, keeping the CTA buttons fully visible without needing to zoom out.

  • Large Monitors: Optimized container widths so elements don’t stretch awkwardly.

I’ve verified these fixes across simulated dimensions from 430x932 up to 1366x640 (short desktop), and everything now stays fully usable and within view.

If you have a chance to try the character creation screen again on your setup, I’d love to know if it works smoothly for you now. Thanks again for helping me catch this!

1 Like

After I logged in I arrived a page that I guess was the one above, but it had just a picture of the apprentice filling the whole screen.

I’m on a mac pro desktop.

It also had two vertical scroll bars.

I scrolled the inner vertical scroll bar down and saw other images like the ones in Soumay’s screenshot, except bigger and arranged vertically.

I clicked on some and got either a short beep or a deep beep, but it was totally unobvious what was happening.

I suspect that I was really only entitled to click on the apprentice boy or girl.

There was no way to get off that page without clicking “maybe later”, which was not visible until scrolling down the outer vertical scroll bar.

I did that, and found it all rich but bewildering and glitchy.

I couldn’t get back to the original page to take screenshots and report.

When I click on “change look” it is much clearer - most looks are clearly not accessible to me.

(Many have broken image icons also)

I did the rank appraisal. The first time I was disoriented about what was expected, and got a low rank.

The second time, I got a reasonable indication of my OGS rank.

I’m not wild about this puzzle, which was presented as “direction of play” during rank-test:

From a direction of play perspective, this is wierd - a 4 stone handicap game where no-one played in the bottom.

Surely a bottom corner is a correct answer, but the only accepted one is to leap into the top side.

Which seems to also contradict the hint …

I used the “report alt answer” function. It left the stone that I had played on the board (leaping in at the expected answer point).

2 Likes

Thanks so much for the incredibly detailed feedback, GreenAsJade! This is gold for me.

You ran into some massive UI/UX issues, and I really appreciate you pushing through it to test the site on your Mac Pro desktop.

Here is what I’m taking away from your report and will fix next:

  1. The Double Scrollbar Nightmare: Having two vertical scrollbars and hiding the “Maybe Later” button at the very bottom is a terrible experience. I’ll fix the layout container so it fits the viewport properly without creating nested scrolls.

  2. Confusing Clicks & Beeps: The short/deep beeps were meant to indicate locked skins, but clearly, the visual feedback is missing. I need to add clear “Locked” overlay UI or tooltips so it’s obvious why nothing is happening.

  3. Broken Images: I’ll inspect the “Change Look” assets immediately to see why those image links are broken.

  4. Rank Assessment Guidance: I’ll work on making the instructions for the first-time rank test clearer so players don’t feel disoriented when starting.

Glad to hear the second attempt gave you a reasonable OGS rank, though! Thank you again for your patience with these glitches—I’ll get to work on smoothing out this onboarding flow!

I found my way back to the “Create Character Initial Page”. You already fixed some layout, but looking more closely, I see that one reason for confusion is that there are text colour problems:

Also, when I scroll down and click on lower images, I get the low beep, which of course means that I can’t select that one … but why do you present it if I can’t select it?

Clicking on this not-available “look” just beeps. Since I can’t see what’s above, I have no idea that this means “this image was not selected”, nor why it was not.

Clicking on any of these “black text with white background” areas beeps, but does nothing.

There is no obvious way to get off this page (other than maybe later).

1 Like

+1
plus, still a lot of ui/ux issues remain.
first step should be to fix ui issues depending on viewport, that still exists in various pages, then fix the theming color palette, then the ux issues.
then would be localisation, because most things are still in Chinese.
and keep the premium off during testing phase so everything is testable

1 Like

Thanks — these screenshots help a lot.

They make it clear that the problem is not just layout, but also visual clarity and feedback:

  • Some text contrast is too weak.

  • Unavailable looks are still being presented in a way that makes them feel selectable.

  • The low beep alone does not explain what happened or why.

  • Some text areas also look clickable even though they do nothing useful.

  • The way out of the page is still not obvious enough.

I’ve already started fixing the onboarding flow, but these screenshots show that I still need to simplify it further and make the page much clearer about what is available, what is locked, what was selected, and how to leave.

This is very useful feedback, and I’m treating it as a real UX bug, not user error.

By the way, as a huge thank you for your continuous testing and taking the time to capture these helpful screenshots, I’ve just upgraded your account to 1 Month of VIP for free! :wrapped_gift: (I recognized your account from your active puzzle sessions!)

I’ll be working hard to iron out these UX bugs over the next day or two. I hope you enjoy exploring the VIP features once the dust settles. Thanks again for helping make Go Odyssey better!

Thanks for the +1 and for continuously following the updates, @SoumyaK4!

Seeing you here reminds me—I really want to thank you for catching that initial critical layout bug on laptop screens. It was a huge help.

As a huge thank you to my early testers, I’d love to upgrade your account to 1 Month of VIP for free as well! :wrapped_gift:

However, I couldn’t find an exact match for “SoumyaK4” in my user database (you might have used a different username or email to register).

Could you please shoot me a private message (PM) or reply here with your website username or email? I’ll get your VIP set up right away!

Thanks again for your support and for helping me catch these blind spots early on!

Even with a new layout, the site remains basically unusable on mobile. I’m looking at a puzzle screen and I can’t scroll down, as soon as I scroll it snaps me back up to the top. There is discrepancy between the explanatory text and the actual button text (it tells me to press the “Answer :play_button:” button, but the actual button says “Show Answer” with no emoji, and I encountered a ladder situation where your GnuGo instance insisted that my answer was invalid when it was in fact equivalent to the “correct” solution. I went to replay the problem, but couldn’t complete it because I got an error your AI timed out.

I would encourage you to do a bunch more testing on your own before soliciting outside feedback. These are bugs that you should be able to catch yourself; you are “wasting” the opportunity to get more useful qualitative feedback about people’s opinions by having your site be so buggy that all you’re getting is basic bug reports.

Hi @marinakai,

Thank you for the blunt and honest feedback. You are absolutely right, and I appreciate you taking the time to write this up despite the highly frustrating experience.

I completely agree with your point about internal testing—I was too eager to get the site out for feedback and ended up letting basic bugs get in the way of the actual user experience. You are right that it wastes the opportunity for qualitative feedback, and that’s a valuable lesson learned on my end.

Regarding the specific issues you caught, I am logging them as high-priority bugs right now:

  • Mobile scrolling: The snap-to-top issue makes it unplayable. I am making this my top priority to fix.

  • Button text discrepancy: A sloppy mistake on my part. I will align the explanatory text with the actual UI immediately.

  • GnuGo ladder logic: You make a great point. Alternative valid solutions definitely need to be recognized. I will look into how the GnuGo instance handles these equivalent variations.

  • AI timeout: I’ll check the server logs to investigate why the AI instance hung up during your replay and work on stabilizing it.

Thank you again for the “tough love” and for holding the project to a proper standard. As a token of my genuine appreciation for you catching these (and for putting up with the bugs!), I’d love to upgrade your account to 1 Month of VIP for free, just as I’ve done for other early bug-catchers.

I’ll be heads-down fixing these issues over the next few days. Thanks again for the reality check!

I suspect you will find the solution to more robust puzzle variations isn’t “use GnuGo better” but “use hand-authored puzzles”. This isn’t a task that Go AIs are currently particularly well-suited for.

Hi @marinakai,

I think you’re right. For strict tsumego-style validation, relying on live engine judgment is not robust enough. Go engines are useful for analysis, but they are not a perfect source of truth for local puzzle variations, especially when the goal is to teach exact sequences.

The direction I’m moving toward is a hybrid system: hand-authored SGF variation trees as the authoritative source for important puzzles, with engine analysis used only as a support tool for reviewing, suggesting, or calibrating content.

That does mean more work on the content side, but it should make the training experience much more reliable and avoid frustrating false negatives. Thanks for pointing this out so clearly. It helps shape the roadmap in the right direction.

Hi Jason,

I started using your new website today (got a pass to test it properly). So far I like it, and I noticed a few things that could be improved.

  1. When answering a problem correctly, the UI displays some message on the right about the correct answer, however it immediately automatically moves on to the next problem, so I don’t have time to read it. It would be helpful if moving on to the next problem could be prompted by the user. Also, even when I get a problem right, I’d like the opportunity to put down stones and explore, to double check if I read all paths correctly.

  2. I can click on Prev to go to the previous problem, and then “solve” it again - this shouldn’t really do anything, but instead it deals damage to the monster. So it is easy to cheat this way :slight_smile:

  3. Colour scheme sometimes hides text, eg in the screenshot. The Go - Orc Arena - Capturing Race is barely visible because the text has the same colour as the background.

  1. The name listed under the avatar is not the nickname I chose, but instead some weird g_firstname_lastname, probably grabbed from google (I logged in with my google account).

Anyways, that’s it for now. Will send more issues if I spot any.

1 Like

Hi Imd11,

Thank you so much for testing it out and providing such detailed and honest feedback! This is exactly the kind of input I need right now.

To address your points:

  1. Auto-skipping on correct answers: You are completely right. Not being able to explore the variations or read the text is a major UX flaw right now. I’ve added this to the issue tracker. I plan to separate the “solved” view from the “next problem” trigger so you can freely place stones and explore the board after solving a problem.

  2. The “Prev” button monster damage exploit: Haha, excellent catch! You found a classic client-server sync loophole. The good news is that I am currently in the middle of completely rewriting the backend problem-judging engine (moving to a strict server-side SGF authority). Once that’s deployed, the server will strictly validate all moves, and this cheat will naturally be fixed. Feel free to abuse it while it lasts! :wink:

  3. Color contrast: Thanks for the screenshot, that is indeed barely readable. I’ve logged the CSS issue and will adjust the contrast for the arena UI.

  4. Google nickname: Ah, that’s a bug with how the Google OAuth pulls profile data on the backend. I’ll fix the logic so it respects your chosen nickname instead of overriding it with the Google name string.

Thanks again for taking the time to write this up. I’ve logged all four items into my backlog. Please do let me know if you spot anything else!

Cheers, Jason

Uh this response feels vaguely AI generated - is it? I’d like to know if I’m actually communicating with the creator here, or an AI agent.

1 Like

Haha, you caught me! Yes, it was.

My English isn’t very fluent, so I use an AI to help polish my writing and make sure it sounds natural. But please rest assured, the actual content, the bug analysis, and the roadmap are 100% from me (the creator).

I really appreciate your detailed feedback, and I read every word of it myself. Thanks for understanding!