The Tavern

This may be simple coincidence, but I have received NO results besides "freezing" as an ego, using scrolls of Brand Weapon in 0.27 and Trunk for about a month now.

Interesting... I branded a broad axe in my current game (0.27), the result was freezing.

I pulled up a game in wizmode and read 12 scrolls of brand weapon

Clear case: Siegurt's experience is unquestionably coincidence, whereas OP's and Sprucery's is caused by dark and sinister forces trying to take control of the game in the name of Eternal Frost.

Was this unclear?

this may be a coincidence, but i die to sigmund every time i play for about a month now. curious what is happening to everyone else cuz its straight murder going on here

While Siegurt's test shows this to merely be a coincidence, I feel like people are being a little overly snarky here. Remember, there was once a bug that doubled all melee damage that made it into a stable release. A bug that causes a rarely-used scroll to always give the same reasonable result? I could totally see that flying under the radar for weeks or even months.

So yes, I agree this one is just a coincidence, but it's not absurd on its face that it might not be.

Exactly so; that's why I asked, in fact. I'm aware of the prevalence of confirmation bias, so I wanted to know if it was coincidence or if someone else had noticed anything odd. I didn't think it's the only result, but it does feel a little...unrandom. Unintentional bias in pseudorandom results isn't rare at all.

Frankly, the response smacks of defensiveness.

Oh, you don't even need to go that far. That melee bug was due to something as simple as a developer putting a "+=" instead of a "==" somewhere in the code. NetHack had a bug that persisted for nine years that was due to a programmer forgetting to put a "break;" in a case statement. There are lots of ways for even an experienced programmer to slip up and introduce program-altering bugs - which is a big part of the reason that companies with millions of dollars riding on correctly-working code invest large amounts in automated and human bug detection.

Incidentally, Crawl's PRNG is actually one of its most solid areas. The motivation for this is the fact that it's played competitively at a semi-serious level (it's probably the roguelike that has the nearest thing to a competitive "scene" a la Super Smash Bros or League of Legends), and it turns out that sufficiently determined players can actually manipulate a weaker PRNG by observing large numbers of "random" occurrences. The general concept is that a PRNG is little more than a compact way to generate a long, cyclic list of numbers, with the "seed" just representing where you start on that cycle. Older RNGs (like NetHack's) would generate 2^32 (4 billion) 16-bit integers. It only takes 8GB of hard drive space to store that entire sequence, so by generating enough consecutive "random" integers, you can then search the whole 8GB sequence - once you know where you are, you can predict what numbers will come next. That's why Crawl (and I gather most new applications nowadays) use a PRNG with a period of 2^19937 -1, since the immense size of that sequence makes storing and searching it completely impossible (there aren't nearly enough particles in the universe to do so).