Additive Multipliers

[Derakon]

(More off-topicness! Yay!)


Just wondering... has the RNG ever actually been subjected to statistical tests?

Years ago on RGRA there was some discussion about replacing the Angband PRNG with the Mersenne Twister or another standardized PRNG. Someone did some tests and showed that the Angband PRNG is basically as random as the Twister is in all meaningful definitions of the word "random".

Of course, now I can't help but wonder if the RNG knew to fake its test results...

[fizzix]

It's also been my perception that a subsequent attempt is more likely to fail. It's gotten to the point that if I fail on my first attempt in a critical situation, I'll attempt something else or retype the command rather than use the 'n' key.

If there was a problem with the RNG for this specific situation it could only matter in the situations when the RNG was called several times. If you sustain a fire breath, for example, the RNG gets called many times to determine what gear was destroyed. So, in pretty much any game critical situation, there are multiple calls to the RNG between casting attempts, and any correlation between successive failures should not exist.

A correlation of bad attempts might show up if you were doing what I was doing and phasing around a destruction zone with no monsters around. Still though, the RNG gets called between successive attempts some to determine whether monsters spawn on the level and how much SP you regen. So I'm not sure how there can be any issues here. It doesn't seem feasible.

[buzzkill]

... However, I did notice that I was slightly more likely to fail if I had failed the previous attempt, and slightly more likely to succeed if I had succeeded the previous attempt....

It's also been my perception that a subsequent attempt is more likely to fail. It's gotten to the point that if I fail on my first attempt in a critical situation, I'll attempt something else or retype the command rather than use the 'n' key.

[fizzix]

(More off-topicness! Yay!)


Just wondering... has the RNG ever actually been subjected to statistical tests?

I once did a haphazard statistical test to see if the fail % was right. I cast 1000 phases with a rogue and found that I failed just about 5% of the time, as expected. However, I did notice that I was slightly more likely to fail if I had failed the previous attempt, and slightly more likely to succeed if I had succeeded the previous attempt. This was not done to any statistical meaningfulness though, and I bet if I was to extend the test to 10000 casts, that would have regressed to the expected value as well.

[LostTemplar]

RNG must be unpredictable. For a true random probability of "something strange happens" is many times higher then probability of "expected scenario happens", the reason for this is a human way of thinking, basically one divides all infinite set of events into two subsets, one of them contains only one expected event, and another contains all remainig infinity-1 events. Probability of the first subset is obviously lesser then probability of the second one. So complains about RNG will never end.

[AnonymousHero]

(More off-topicness! Yay!)

No, the RNG hasn't changed in aeons.

Just wondering... has the RNG ever actually been subjected to statistical tests?

[fyonn]

Also RNG seem to behave a bit strangely in my games. Things seem to appear in clusters. First no speed whatsoever for a long time then single drop contains three of them. RoRPoison doesn't appear before killing Sauron? Has there been changes in RNG code?

heh, my level 31 HE ranger currently has 48% failure rate on the identify spell and it succeeds about 1 time in 5 or less... I can literally sit then and fail 6 or 7 times in a row, and I never seem to get that happening on the success side...

dave

[Magnate]

I find that actually only healing potions are a problem. All other endgame-gear are easy to get. With help of big stack of CCW it isn't that much a problem now that CCW cures are percentile instead of fixed.

Also RNG seem to behave a bit strangely in my games. Things seem to appear in clusters. First no speed whatsoever for a long time then single drop contains three of them. RoRPoison doesn't appear before killing Sauron? Has there been changes in RNG code?

Lack of consumables is a problem but not enormous game stopping problem. Game is fine. It could be better but it isn't entirely broken even with reduced drops.

No, the RNG hasn't changed in aeons. I agree with you that the game isn't broken - but tweaking drops is always an interesting topic. I think weapons and armour could be toned down more, and I *really* think a bunch of consumables should be tagged as {good} or {great} drops. But yes, it works fine. I just need to find a damn staff of healing ....

[Timo Pietilä]

@Timo: I wasn't asking about turncount for judgemental reasons, simply for balance - if he'd said 2M turns then I could have understood him not thinking consumables were an issue. But at the speeds he quotes, he plays roughly the same speed I do, and I find them a problem!

I find that actually only healing potions are a problem. All other endgame-gear are easy to get. With help of big stack of CCW it isn't that much a problem now that CCW cures are percentile instead of fixed.

Also RNG seem to behave a bit strangely in my games. Things seem to appear in clusters. First no speed whatsoever for a long time then single drop contains three of them. RoRPoison doesn't appear before killing Sauron? Has there been changes in RNG code?

Lack of consumables is a problem but not enormous game stopping problem. Game is fine. It could be better but it isn't entirely broken even with reduced drops.

[Magnate]

http://angband.oook.cz/ladder-show.php?id=10455

A recent ladder post of YA1TW. Note he has 2 stacks of holy might and 4 stacks of slay evil. He also maxed all but INT. And he's under 1M turns.

I don't have a great sense for what different turn counts represent, if you care to evaluate winners that way. From my vantage point under 1 million is achievable but still above-average, under 500K is impressive, under 300K is epic maniac style.

If you can for all intents and purposes max stats, stow crazy endgame ammo, and win in under 1M turns drops are not a big problem. Again, I suspect I am a minority of 1 here. I'm jut saying if it were up to me I would worry about other stuff first. In other news I am way off topic.

Your judgement is tougher than mine - I consider under 500k to be awesome, and under 1M to be very good. But everybody keeps telling me the game has got easier, so perhaps you're right.

Given your three sub-1M turncounts to dl98, I'm surprised that you weren't hurting for any consumables. I must be quite unlucky - I've never got to dl98 without a serious lack of something (speed, resists, healing, whatever). In fact ammo is just about the only thing I do seem to find enough of - which I guess is grist to Timo's mill.

@Timo: I wasn't asking about turncount for judgemental reasons, simply for balance - if he'd said 2M turns then I could have understood him not thinking consumables were an issue. But at the speeds he quotes, he plays roughly the same speed I do, and I find them a problem!

[AnonymousHero]

Yeah, my bad. That's what I meant .

[LostTemplar]

probably he just sold some artifacts, got tensers, and then bought artifacts back

[fizzix]

I did as it happens; there was a Tenser's for sale in the BM and I didn't have the cash for it -- with a little buying, branding and reselling I was just able to scrounge up the cash. (I usually don't bother going back to sell stuff during the game; 1 or 2 ego items at most per trip. Trips for me also tend to last a lot longer these days than they used to.)

IIRC, I also bought a few stat potions from the BM with this trick.

wait wait wait.

How did you brand without tensers????

[AnonymousHero]

Really? This is what you do? Brand ammo and sell it back to the stores? You still need money on dl70?

I did as it happens; there was a Tenser's for sale in the BM and I didn't have the cash for it -- with a little buying, branding and reselling I was just able to scrounge up the cash. (I usually don't bother going back to sell stuff during the game; 1 or 2 ego items at most per trip. Trips for me also tend to last a lot longer these days than they used to.)

IIRC, I also bought a few stat potions from the BM with this trick.

[nullfame]

http://angband.oook.cz/ladder-show.php?id=10455

A recent ladder post of YA1TW. Note he has 2 stacks of holy might and 4 stacks of slay evil. He also maxed all but INT. And he's under 1M turns.

I don't have a great sense for what different turn counts represent, if you care to evaluate winners that way. From my vantage point under 1 million is achievable but still above-average, under 500K is impressive, under 300K is epic maniac style.

If you can for all intents and purposes max stats, stow crazy endgame ammo, and win in under 1M turns drops are not a big problem. Again, I suspect I am a minority of 1 here. I'm jut saying if it were up to me I would worry about other stuff first. In other news I am way off topic.