Automated generation of Manic Miner speedrun/walkthrough

[crem]

Ok, the level 4 generation has been done again.

Although it's not exactly the same as yesterday (but the general route is the same), it's still one frame longer than the current human record. Guess I'll leave it as it is for now.

Score=1853

lvl4_20210309-0539.rzx

[IRF]

I believe that if the jump at 22 seconds was replaced by walking off the platform instead, then it might save one time-frame (equating to one more point)? Because jumping forwards, when you can walk continuously forwards instead, slightly slows you down.  (I'm not certain though, because that jump ends in a fall to a lower level which has to be executed anyway, even if you walk that bit.)

[crem]

20 hours ago, IRF said:

For Eugene, you might need to adjust your algorithm so that it collects the nearest item to the portal last - even if that means passing close by to it without grabbing it straight away, but coming back for it once all other items have been bagged.

And also, if Eugene is too close to the portal at the point when Willy is about to finally collect that item, you might have to get him to wait around [which I know you've said the algorithm doesn't like to do!] until Eugene has ascended to a height that allows Willy to sneak into the portal once it starts flashing (whereupon Eugene becomes angry and starts descending again).

I don't want to give it too much hints, that way there is more hope that it finds something that humans didn't think about (extremely unlikely though, but that way the algorithm still feels more "pure").

Basically currently there are only two components in scoring:

• Prefer new places on the screen to the places already visited (encourages exploration rather than staying in one place).
• Slightly prefer more collected items to less collected items (to have some feel or progress).

I think I'll do the following two changes:

• If Eugene sits on portal, that's loss.
• Encourage more exploration not only by looking at Willy's location, but also at Eugene's location (when it's falling).

Only if that won't work, I'll also add penalty to scoring, preferring Eugene at higher positions.

[crem]

1 hour ago, IRF said:

I believe that if the jump at 22 seconds was replaced by walking off the platform instead, then it might save one time-frame (equating to one more point)? Because jumping forwards, when you can walk continuously forwards instead, slightly slows you down.  (I'm not certain though, because that jump ends in a fall to a lower level which has to be executed anyway, even if you walk that bit.)

I've just checked that, and it doesn't help.

If I replace that jump with the same amount of frames walking right, it still works, but if it walks one frame fewer, the second jump is too early.

In general I expect that the tool should find such local optimizations fine; usually it's much harder to discover strategies where cause and effect are further in time.

If someone happens to have .rzx file for the 1854 score walkthrough (or a video, which is slightly worse but doable with frame-by-frame manual comparison), that would be really helpful to debug what's wrong there.

[IRF]

Attached is JetSetDanny's (Daniel Gromann's) RZX recording of Abandoned Uranium Workings, in which he achieved a score of 1854 points.

04 Abandoned Uranium Workings.rzx

[crem]

9 minutes ago, IRF said:

Attached is JetSetDanny's (Daniel Gromann's) RZX recording of Abandoned Uranium Workings, in which he achieved a score of 1854 points.

04 Abandoned Uranium Workings.rzx 20.82 kB · 1 download

Thanks, will take a look where the difference is.

Do you know which emulator can open it? Fuse complains at it for some reason.

Meanwhile, here is the level 6. This time it did achieve the highest known score of 1821.

lvl6_20210309-0450.rzx

[IRF]

I watched Danny's recording in SPIN.

[jetsetdanny]

It will also work in Spectaculator.

Thanks for uploading it, Ian! 👍

The complete set of my RZX recordings showing my personal bests in each MM cavern be downloaded here.

Thanks for the automated walkthrough of "The Processing Plant", crem! It's good to know that the algorithm confirms that my highest score there cannot be improved 🙂.

I take it you are still working on "Eugene's Lair"?

[IRF]

4 minutes ago, jetsetdanny said:

The complete set of my RZX recordings showing my personal bests in each MM cavern be downloaded here.

Danny, I was going to provide that link to the full set of recordings, but I decided against it as it will presumably soon need updating - once you have managed to match (or perhaps even beat?) the current record score for the Cold Room? 😉

[jetsetdanny]

Yes, I remember about the Cold Room and I will have to record a "proper" solution to it (matching the current human- and logarithm-established high score). I will do it soon, when I will also update >>cryptic reference here to my current activities<< some other MM recordings of mine... 😎