Prismatic Trials and Determinism: Some crude experiments

[Zeddy]

Out of curiosity, and because I haven't seen anyone else do it, I decided to do some tests with the Prismatic Trial. All tests were done on the trial that's ending in a few hours from the moment this post was created.

Test 1: Enemy behavior

Experiment: Start up Pristmatic Trial as Commando, touch no input, observe if the same events take place.

Result: Enemies spawned in different locations, took different actions, got hit at different time. Seven second difference in time to death (49 and 56 seconds), killed by two wildly different enemies.

Conclusion: Two different runs with the same input can have two different results.

Test 2: Enemy spawn

Experiment: Start up Prismatic Trial as Commando, observe times and order enemies spawn, avoid killing anything to minimize randomness.

Result: I kept a sketchy log over two runs (plus the previous two runs), and the same enemy spawned at the same time every time as far as I could tell. I'd also notice this being the case in later experiments. I.E. in this prismatic the first Golem spawns at 52 seconds.

Conclusion: Enemy spawn times and order is deterministic.

Test 4: Crits

Experiment: As Huntress, attack enemies until a critical hit was achieved, then restart run and try to reproduce it.

Result: First run got a crit at the Lemurian spawning around 1:40. Killed all enemies before it, including a Golem. Second run I did the same, no crit.

Conclusion: One shot missed in the first run, but not in the second, and the crit order could be influenced by enemy attacks as well, so results are ultimately inconclusive but even if crits are deterministic they seem at the very least unreliable.

Unplanned test: Lunar Coin

Experiment: Got a Lunar Coin from a beetle spawning at 3:40 while doing other experiment. Ran again twice to see if reproducible.

Result: Saw no more lunar coins in any future runs, from the 3:40 beetle or otherwise.

Conclusion: Lunar Coins are not determinstic, and thank goodness for that.

Test 5: Shrine of Chance

Experiment: Exhaust the two Shrine of Chance in the canyon to the right, note results. Try doing them in different orders.

Result: Bottom shrine gave Fuel Cell then Lensmaker's before dying. Top shrine gave three nothings, then Monster Tooth, then Disposable Missile Launcher before dying. Tried shrines in different order, with and without opening a chest in between. Consistent results.

Conclusion: Shrin of Chance is totes deterministic.

Test 6: Chests

Experiment: Open as many chests as possible around the first map of trial, note the drops, try different orders.

Result: Same drops from the same boxes every time, including the myster entries on booths.

Conclusion: Chests are very deterministic.

Bosses?

As far as I can tell, the boss is the same and spawns in the same location every time, but I remember Trial 2 had two different bosses that could spawn in the teleporter. Will have to wait until a Trial rolls around where that happens again before I can determine how it happens. My initial hunch is the order in which crystals are destroyed, and my second guess would be time used.

Hope this was interesting to anyone, and if any of you have more data or knowledge I'd love to know about it!

[fluffy buny]

it would be interesting to know how exactly the 57 Leaf Clover affects level generation if it does at all. at this point i'm not even sure it impacts anything unrelated to combat

i first assumed the clover would just roll a second item the moment a chest is opened and give it to you if it's a higher tier than the chest's original item. but then i considered that there's a chance it works something like the rusted key: on level generation, each chest might be rolled one extra time for each clover the party has, and their contents are replaced where applicable. it'd be pretty hard to test any of this before a prismatic trial with a clover shows up to begin with, though

[Zeddy]

From guesswork and hearsay, I think Clover only works in conjunction with your items, and even then specifically not the Teddy. I dont have the item yet, but a friend who does say there's an audible ding every time the Clover works. If you have one, try listening for it during level generation.

[Sotanaht]

Originally posted by JILL♥PILL:

it would be interesting to know how exactly the 57 Leaf Clover affects level generation if it does at all. at this point i'm not even sure it impacts anything unrelated to combat

i first assumed the clover would just roll a second item the moment a chest is opened and give it to you if it's a higher tier than the chest's original item. but then i considered that there's a chance it works something like the rusted key: on level generation, each chest might be rolled one extra time for each clover the party has, and their contents are replaced where applicable. it'd be pretty hard to test any of this before a prismatic trial with a clover shows up to begin with, though

I'm pretty sure it only impacts combat. More specifically I think it only affects chance on hit activations like the two rings or the ukulele, and maybe chance on kill activations like the bandoleer. I pretty much know for certain it has no impact on chance shrines, there's no way you'd get multiple (3+) consecutive fails with 3 clovers if it did.

[Paranoia]

♥♥♥♥♥♥♥ SCIENCE
A+

[Dallagen]

All the prismatic trial is, is a run with a set seed.

You can view this seed by opening the console (CTRL + ALT + `), and typing "run_print_seed".

Crits, on-hits, lunar coins, etc are all rolled off of a randomized seed from the unity rng implementation and not the world generation seed.

Shrine bosses are static across the seed, but change every difficulty increase on the top right.

World gen, chance shrines, chests, and enemies spawns are also seeded.

[13]

Well written and concise information, Dallagen.

[Lunacy]

On the 2nd trial where you start on snow map, the boss spawned elite greater wisps.
Every other time, spawned elite imp squad. I was confused by the initial inconsistency.

[Zeddy]

Originally posted by Dallagen:

You can view this seed by opening the console (CTRL + ALT + `), and typing "run_print_seed".

Thanks for this! Do you know if there's a console command or something for using a specific seed in a run?

[Dallagen]

Originally posted by Zeddy:

Originally posted by Dallagen:

You can view this seed by opening the console (CTRL + ALT + `), and typing "run_print_seed".

Thanks for this! Do you know if there's a console command or something for using a specific seed in a run?

There was, but the devs removed it this patch due to them blanket removing all the cheat commands.

normally it'd be "pregame_set_seed" then "pregame_start_run".

The only way to get that functionality working again is to open Assembly-CSharp.dll in dnspy and edit console.cs to readd the command names and disable the cheat check for the pregame seed set command.

[Zeddy]

Originally posted by Dallagen:

There was, but the devs removed it this patch due to them blanket removing all the cheat commands.

Oh. Because of the prismatics? Seems like disqualifying people opening the console from prismatic times would be simple.

Guess I'm not seeing that run with 2 alien heads in the first 2 levels again.

[Dallagen]

Originally posted by Zeddy:

Originally posted by Dallagen:

There was, but the devs removed it this patch due to them blanket removing all the cheat commands.

Oh. Because of the prismatics? Seems like disqualifying people opening the console from prismatic times would be simple.

Guess I'm not seeing that run with 2 alien heads in the first 2 levels again.

Partly prismatics, but they could just check the integrity of the dll before uploading a leaderboard score.