0000031: New DMFlag: "Use Classic PRNG"

> ZD's does 4.24% more damage than ST's. IIRC this has not been fixed in ST (So ST's SSG is the same as ZDoom's). In ZD it is a DMFlag and it's typically only on on OS 1on1 servers, not CTF/FFA. The difference is so minor anyway that most people wont notice it, but it IS there.

> I've heard lots of claims that the ZDoom SSG is less powerful, just from gut feeling. Now, gut feeling isn't scientific, so since it was a possible "compatability problem" for me, I decided to go out and research it. Here is what I found. ZDoom uses a pseudo random number generator, Doom2.exe uses a fixed table of 256 numbers, ranging from 0 to 255, all accounted for. Doom2.exe iterates through this table linearly.

> There are exactly 256 different SSG shots patterns possibile in Doom2.exe. Each shot pattern does the same damage every time. I even wrote a little mod that plays back a demo, takes a screenshot of each ssg shot, and points out the profile number for you to look in another chart I made (which contains all the different SSG spread patterns).

> The ZD SSG uses the doom2.exe mode. Dash also did damage calculations (where the 4.24% comes from), and the ZD SSG does more damage on average.

ZDaemon has something called "classic weapon strength". It's toggled by setting "sv_oldrandom" to 1 or 0.

> "Random" numbers on a computer are normally pseudo-random: not truly random. This applies to practically all computer programs out there. The reason is that computers are deterministic machines and it's pretty hard to find a source of true randomness.

> Every pseudo-random number generator (PRNG) generates some sequence of numbers that look random enough to pass statistical tests. That's usually good enough for games and other simulations (but not for crypto work; you need fancier PRNGs there). Whether that's done via some math algorithm or some table is pretty much irrelevant. What really matters is the mean and std. deviation of the sequence.

> The dmflag (sv_oldrandom) you're asking about determines if ZDaemon uses the PRNG of Doom (OS) or the one of ZDoom (NS). Their means are not very different, but their std. deviations are (OS PRNG has higher std. dev. than the NS one); that is important with weapons such as the SG, SSG, CG where you fire repeated shots at the opponent (vs. the RL for example where the damage per hit is much larger); the increased std. deviation means that wild shots from downtown will bring the other guy down. Of course there is the flip side of the coin: you might hit someone right on the chest and very little damage happens (which also means that you should not blame every SSG miss to netcode, lag, etc)

> Overall, it's fairly clear that the OS PRNG makes the SG, SSG, CG more effective than otherwise. This is just the way things are; I'm not implying that one way is better than the other. Please avoid turning this into the millionth OS/NS debate. There is no point in that.

> Here is my comment from the X-Doom m_random.cpp source tree:

// [Dash|RD] -- Make random number generation use the original random number chart.
// The way we pick numbers from this chart is NOT the way the original
// game does it, but that's hardly relevant. We aren't trying to play back
// demos afterall. More work may need to be done on this.
//
// Why did I do it? With this chart the game appears to pick consistently higher
// numbers. This should address the reports that the SSG (and various other weapons)
// do less damage in ZDaemon.
//
// SSG damage statistics, 20 shot survey.
// PrBoom -- Max: 245 Min: 165 Median: 210 Average: 206.5 Total: 4,130
// ZDaemon -- Max: 215 Min: 155 Median: 195 Average: 194.5 Total: 3,890
// This Code -- Max: 245 Min: 155 Median: 205 Average: 202.75 Total: 4,055
//
// Obviously you want to give a bit of room for randomness, but has is definitely been my experience that
// this code statistically gives off higher `random' numbers. Other shotgun guru's have also claimed that
// this code fixes their issues with weapon damage.