Preserve mode

[Antoine]

Yes, but how is that info any way relevant when that is not what happens in real game?

Go easy Timo - don't assume the dev team are always doing something dumb.

They filter the data appropriately to get some impression of what an actual player might collect.

A.

[Timo Pietilä]

Go easy Timo - don't assume the dev team are always doing something dumb.

They filter the data appropriately to get some impression of what an actual player might collect.

A.

That's one smart filter. You have to assume a lot to get anything worth using.

I see a potential usage in balancing monsters, their risk/reward ratio, but not much in anywhere else. Sometimes a good common sense and intuition rules over automation. For game enjoyment differences are a good thing, Azriel way earlier than Carcharoth even that both are near evenly matched for deadliness.

My fear is that you are generating a dull game by trusting automation and spreadsheets too much, much like what happened to Z when it was taken over by committee from Topi Ylinen. TY-era Z was crazy, but fun. Later editions were full of features, but boring. Kind of like Master of Orion vs Master of Orion III.

[Derakon]

The stats collector isn't perfect, but this change (to add monster inventories) is a step in the right direction. More selective monster killing would also be such a step. However, just because the collector isn't perfect doesn't mean it can't generate useful stats already. You can't do any fine tuning with it, of course, but you can compare relative frequencies (e.g. "how much more likely is pConf to show up by 2000' in a standart game vs. a randart game?"), which is very important.

[Magnate]

The main reason for seeking this particular granularity is to try and get a handle on the impacts of pits and vaults on the game. We suspect that they are an extremely significant source of good objects, and it would be nice to test that suspicion with some statistics.

Timo needn't worry though, we are quite a long way from running everything by spreadsheet. There is still plenty of scope for the devteam's "common sense and intuition" to influence the game and then be disagreed with by, er, Timo.

[ChodTheWacko]

Just some random thoughts and Qs.

Do we really want preserve + monsters picking up/holding artifacts?
Won't that make diving incredibly brutal? I mean, it's one thing to beat an OoD monster and get an artifact (maybe). It's another thing to (maybe) know that a really OoD monster is holding an artifact. You won't know, but you'll see 'special feeling' and have cleared the level except for 2-3 very very dangerous monsters.

I understand that 'teleport other' to poof that monster away so you can get the artifact is a little cheesy, but being forced to kill that monster to avoid losing an artifact seems just too much.


side note on statistics:
I agree with Timo, that I don't see the real benefit of creating objects at the same time as monsters to generate statistics versus just storing the information of the object when it is created. That just seems much more useful.

Magnate mentioned:
monster was randomly wandering the corridors, or was summoned, or was generated in a vault, or a pit, etc

Aren't these monster flags, versus object flags? Objects aren't created differently if you kill a spawned orc versus a 'vault orc'. In your examples, all you care about are the origin of the monster that you killed to create an object.
So can't that be information just be stored at monster death?

Don't get me wrong, I think it would be fantasically interesting to know:
1) whether an object came from inside a vault
2) whether an object came from killing a monster inside a vault
3) How many vaults of type X I saw.
4) a histogram of how many kills of monster X I saw at what dlvls, along with which came from a vault.

But for example, since I avoid graveyards like the plague, is it intersting to keep stats of the objects that I would have gotten ahd I decided to kill the
entire graveyard? (which I'll never do?)

It seems to be that you'd be just as well off doing this:
Whenever you generate a monster for a level, generate a set of objects as if you killed it right now, store those statistics, and that's that. When the player actually kills a monster, generate the objects as we do now.

Isn't this really getting you the same thing?

- Frank

[Antoine]

> Do we really want preserve + monsters picking up/holding artifacts? Won't that make diving incredibly brutal?

Only if you set preserve mode (which I understand is unusual).

> I agree with Timo, that I don't see the real benefit of creating objects at the same time as monsters to generate statistics versus just storing the information of the object when it is created.

Don't mean to be rude, but I'm not sure if it's super useful for people to comment adversely on the object simulator if they haven't been involved in it up til now - there is a lot of context about how it works and how it is used. If the devs think they get some mileage out of it then that should be sufficient...?

A.

[ChodTheWacko]

Don't mean to be rude, but I'm not sure if it's super useful for people to comment adversely on the object simulator if they haven't been involved in it up til now - there is a lot of context about how it works and how it is used. If the devs think they get some mileage out of it then that should be sufficient...?

A.

I will admit, I've been a bit out of the loop.

For any non-trivial project, it's always useful to weigh value vs cost.
I didn't say it was a bad idea, I was just pointing out that for the examples given, It doesn't seem to be necessary. I was wondering if I was missing something. This is a forum after all - a place for discussion.

In retrospect, I suppose there is a thread somewhere discussing all this.
Anyone have a link? I guess I missed it.

Anyway, I didn't mean to drag this thread too off topic (though I did originally want to note the difficulty of preserve+carry+diving)

- Frank

[takkaria]

Magnate mentioned:
monster was randomly wandering the corridors, or was summoned, or was generated in a vault, or a pit, etc

Aren't these monster flags, versus object flags? Objects aren't created differently if you kill a spawned orc versus a 'vault orc'. In your examples, all you care about are the origin of the monster that you killed to create an object.
So can't that be information just be stored at monster death?

Before this change, that information was not stored anywhere. Now it is. We could have implemented it as setting a flag on the monster which was then passed along to object generation when it died, but there's no real advantage to doing that over what we've done, especially with view to future changes (e.g. maybe stealing things from monsters, monster using arrows as ammo rather than having an unlimited supply, a 'chief' monster of a pack be the one carrying the best item).

[Timo Pietilä]

Before this change, that information was not stored anywhere. Now it is.
....
future changes (e.g. maybe stealing things from monsters, monster using arrows as ammo rather than having an unlimited supply, a 'chief' monster of a pack be the one carrying the best item).

You could as well code then a data collection code that stores that info in some file and then ask people to send you those files to see what people are really killing in what depths. I have got an impression that most devs aren't very good at playing the game, and those that are are extreme cases playing game in way different way than most of us (like Eddie). This casts shadow over the trust for devs being able to generate useful filtering mechanism for what people really get in which depths.

For that "pack leader has best object", I realize you say there "maybe", but in case that was serious thought it would reduce the randomness of the game which is very bad. You could make it more probable for pack leader to have better items, but forcing that is bad. Part of the fun in angband is that you can have an epic fight with major demons getting zillion drops and get just few useful things, nothing major and then go around a corner, kill an orc mage and get Ringil. You never know where you get that wow! -item, though probability is higher for deep monsters with good drops and vaults.

[takkaria]

You could as well code then a data collection code that stores that info in some file and then ask people to send you those files to see what people are really killing in what depths. I have got an impression that most devs aren't very good at playing the game, and those that are are extreme cases playing game in way different way than most of us (like Eddie). This casts shadow over the trust for devs being able to generate useful filtering mechanism for what people really get in which depths.

All I was really trying to do was to explain that the change from generate-on-drop to generate-on-creation is largely a technical change and was done for good technical reasons, as well as reasons of possible future extendability in the face of people questioning it. As to stats, stats are useful when you're experimenting, even if they're not massively representative of actual play (for example, looking at item frequencies across levels, or when you're thinking of tweaking artifact generation and you want roughly the same results as you had before but with a different algorithm, or when you're changing dungeon generation and you want to see how it affects item distribution). I don't think there's any suggestion of using stats as an automated balancing tool without oversight, but I could be wrong...

For that "pack leader has best object", I realize you say there "maybe", but in case that was serious thought it would reduce the randomness of the game which is very bad. You could make it more probable for pack leader to have better items, but forcing that is bad.

It's something that's crossed my mind before, but since the game has no concept of 'pack leader' now anyway it's largely academic. I think you're right, it would reduce randomness if it were guaranteed, and I'm not sure that would be a great idea.

[ChodTheWacko]

For that "pack leader has best object", I realize you say there "maybe", but in case that was serious thought it would reduce the randomness of the game which is very bad. You could make it more probable for pack leader to have better items, but forcing that is bad.

The randomness doesn't need to go away, but I do like the concept of an 'alpha male' of the pack. Perhaps even bump up their hit points/damage slightly and give them a chance of a slightly better object. It would make it slightly more interesting to fight through the pack, and try to kill the leader.

- Frank

[Timo Pietilä]

All I was really trying to do was to explain that the change from generate-on-drop to generate-on-creation is largely a technical change and was done for good technical reasons, as well as reasons of possible future extendability in the face of people questioning it. As to stats, stats are useful when you're experimenting, even if they're not massively representative of actual play (for example, looking at item frequencies across levels, or when you're thinking of tweaking artifact generation and you want roughly the same results as you had before but with a different algorithm, or when you're changing dungeon generation and you want to see how it affects item distribution). I don't think there's any suggestion of using stats as an automated balancing tool without oversight, but I could be wrong...

Raw data is mostly good, but when that raw data doesn't represent the real life it is misleading, and that is dangerous. Exactly that "frequencies across the levels" is the case where that isn't showing real data. You don't actually get all the items in all the levels or kill all the monsters in those levels. Especially data from mid-game way OoD vault monsters would give extremely wrong result, because you can't handle most of them so you also don't kill them. It can be useful, but only with thought put behind it.

For the record, I have nothing against the monster inventories, I think they are great (once the preserve off is fixed). I just got scared about the one aspect to what it is used for.

[EDIT: typo fixed]

[ChodTheWacko]

All I was really trying to do was to explain that the change from generate-on-drop to generate-on-creation is largely a technical change and was done for good technical reasons, as well as reasons of possible future extendability in the face of people questioning it. As to stats, stats are useful when you're experimenting, even if they're not massively representative of actual play (for example, looking at item frequencies across levels, or when you're thinking of tweaking artifact generation and you want roughly the same results as you had before but with a different algorithm, or when you're changing dungeon generation and you want to see how it affects item distribution).

By the way, thanks for the additional examples - those were helpful.

From a player perspective, I would eventually like to see stats which show you what is happening based on your playstyle. It would be interesting to see how your style is affecting how you acquire objects.

From a general debugging standpoint, there is definitely a use for 'across the board' object stats. We've often seen people ask questions like 'Is object XXX being generated too frequently?'
Across the board stats would help answer these questions.

It would be quite interesting, I think, to generate say 100 levels of dungeon levels 1-100, and see how frequently objects get created. This could be compared with later builds to see if a particular object (like say, leather boots) is suddenly being generated signifigantly more.

- Frank

[takkaria]

It would be quite interesting, I think, to generate say 100 levels of dungeon levels 1-100, and see how frequently objects get created. This could be compared with later builds to see if a particular object (like say, leather boots) is suddenly being generated signifigantly more.

Aye, this is in fact what it will be used for.

[takkaria]

Raw data is mostly good, but when that raw data doesn't represent the real life it is misleading, and that is dangerous. Exactly that "frequencies across the levels" is the case where that isn't showing real data. You don't actually get all the items in all the levels or kill all the monsters in those levels. Especially data from mid-game way OoD vault monsters would give extremely wrong result, because you can't handle most of them so you also don't kill them.

But that's the point of getting more granular data- now we'll know from a stats run which items came from vaults / vault monsters and which didn't, which means that if we make changes, we can tell how, broadly speaking, those changes affect various things. If you get 40 dlev OoD monster on dlev35 then instead of its items being counted as part of standard dlev35 drop, they'll be counted as from vaults on dlev35, and this should help separate out the cases where you can't handle monsters from when you can. Sure, it won't catch every change that could go wrong, but that's what nightlies are for...