First Win!!!

[Derakon]

Hm, maybe my memory is faulty then. I mostly just remember not finding the really big artifacts at all often, and all three of those are what I'd call "big artifacts". Oh well.

[fizzix]

Even in 3.0 I tended to find the equivalent in most winning games, and artifacts are substantially more prevalent in 3.1+. Now that you can't help tripping over powerful vaults, that set is quite ordinary. If I play a char to CL50, I am disappointed if I do not find 2 elven rings, and I half expect to find all 3.

I guess a diving game in 3.0- with autoscum off might produce less in the way of artifacts. Could that be what you are remembering?

I clicked on Angband dumps, set to show winners, went to #500 and looked for an old one. http://angband.oook.cz/ladder-show.php?id=129 That's 2.9.3, and better stuff than the dump in this thread.

Kind of hard to get a good comparison unless you have kill counts. I have no idea how long that player spent scumming for those items. They could have spent 4-5 million turns on dlevel 99 alone.

[myshkin]

Nothing out of ordinary there except that ring. It must have got a lot of supercharges to get that high value. Someone familiar with the code should calculate how rare that actually is.

This is not a simple calculation. The value for a speed ring is 5 + MB(5) + supercharges. MB(5) is a pseudo-normal distribution capped at four standard deviations. At dlvl 91, where his ring dropped, its mean is 3 with probability 71/128 and 4 with probability 57/128, while its standard deviation is 1 with probability 3/4 and 2 with probability 1/4. The code then supercharges the value by one every time it passes a one-in-two test.

I am presently too lazy to run an exact calculation, but some Monte Carlo tests indicate that the probability that a speed ring dropped on dlvl 91 has pval >= 21 is about 0.0002. (By comparison, P(pval >= 15 | dlvl 91) is about 0.0136.)

[Timo Pietilä]

I am presently too lazy to run an exact calculation, but some Monte Carlo tests indicate that the probability that a speed ring dropped on dlvl 91 has pval >= 21 is about 0.0002. (By comparison, P(pval >= 15 | dlvl 91) is about 0.0136.)

So one in 50000 speed rings get that value. I'd say that's pretty rare :-)

[PowerDiver]

So one in 50000 speed rings get that value. I'd say that's pretty rare :-)

I think you miscounted the 0s and it is 1 in 5000. Still pretty rare.

[camlost]

This is not a simple calculation. The value for a speed ring is 5 + MB(5) + supercharges. MB(5) is a pseudo-normal distribution capped at four standard deviations. At dlvl 91, where his ring dropped, its mean is 3 with probability 71/128 and 4 with probability 57/128, while its standard deviation is 1 with probability 3/4 and 2 with probability 1/4. The code then supercharges the value by one every time it passes a one-in-two test.

I am presently too lazy to run an exact calculation, but some Monte Carlo tests indicate that the probability that a speed ring dropped on dlvl 91 has pval >= 21 is about 0.0002. (By comparison, P(pval >= 15 | dlvl 91) is about 0.0136.)

Assuming random number generation. IIRC, an old version of Angband had some trouble with getting trapped in a low-probability loop.

[Timo Pietilä]

I think you miscounted the 0s and it is 1 in 5000. Still pretty rare.

Oops, so I did. So not that rare, but still rare. I think The One Ring had once rarity of 1/1000000. Literally. It basically existed in game, but nobody found it. Since then artifact generation has been revised several times.

[myshkin]

Assuming random number generation. IIRC, an old version of Angband had some trouble with getting trapped in a low-probability loop.

My simulation used the actual random number code that the latest Angband has, although an actual game would have less predictable numbers of RNG calls between speed rings. A few months ago, d_m converted z-rand.c to the WELL1024a RNG (see this paper for a description), and it seems pretty solid.