Report From Another Fantasy

[T-hawk]

I was wondering where the thread went, and couldn't imagine one of the mods deleting it as off-topic or something. Fortunately I found both my post and Sullla's reply in browser cache.

Remember that weekend where Ruff and I were complaining about no Civ to play? Well, I started another project on a favorite game that weekend, and finished it this past weekend. Here's the report, in full Realms Beyond style...

http://www.dos486.com/ff/

Enjoy


Sullla's reply:

Wonderful!!! That read just made my morning.

To answer some of your comments, I did in fact play out a significant part of the zerks variant, just never got around to finishing it or typing it up. I was right about to enter the Sea Shrine, and have about 20 screenshots still waiting to be used. We were thinking on the same wavelength in most every other way: I also fully planned (and did) use the class change feature, on the principle that anything that increased offensive output should be employed. I completely agree with your assessment of the Red Mage/Black Mage split too - Red Mage is superior in something like 95% of all situations. Black Mage class is seriously overrated in this game. However, I went with the Black Mage to make the variant more challenging and/or interesting. White magic of any kind had no place on the berserker team!

I am seriously impressed with the low level run using this variant. Not that it can be done, but that you pulled it off with so few party wipes. The run through the Ice Cave was edge-of-the-seat stuff for sure. (Those designers were just evil when they programmed that place!) I encountered the same sort of gameplay you did during my run - characters frequently dying even in normal random battles from either an unlucky critical hit or a gangup by several monsters at one time. My characters were higher levels though, since I was not doing a low-level challenge, which definitely helped a bit. The higher hit points were fairly negligible, but I hit the Black Belt "killzone" level ups sooner (8, 21, and eventually 32) and had more offense on hand earlier. Against bosses, it was the same thing: the path to success was to hurl everything at the opponent and completely ignore defense. I'm extremely pleased to see that my variant goal played out to perfection in your own run.

So congratulations again on your run! I think it's a testament to this game's beautiful design that so many unorthodox challenges can be designed and played successfully. (Although Final Fantasy is actually from 1987 - the American port is dated 1990, but the game itself is even older!) Although I might be the only one, I would love to see more FF content on your website.

[T-hawk]

Coming back late, but to continue with a few points...

My characters were higher levels though, since I was not doing a low-level challenge, which definitely helped a bit. The higher hit points were fairly negligible, but I hit the Black Belt "killzone" level ups sooner (8, 21, and eventually 32) and had more offense on hand earlier.

My problem with level grinding in FF1 is basically where do you draw the line? Some variants do retain challenge all the way up to the level 50 max, like a solo character or four white mages. But for most parties (and especially anything involving a Black Belt), the difficulty largely depends on the player's tolerance for tedium. The only way to take patience out of the equation is to remove grinding entirely.

Compare FF1 level grinding to Diablo 2. In D2, raw leveling stats pale to insignificance among synergies and equipment and the player's twitch skill. And a single pass without grinding earns a character plenty sufficient power; no D2 variant will fail at level 80 but succeed at 99. Not so for FF1, where every single level brings considerable increases to hit points and damage, including some very stark breakpoint jumps.

Although I might be the only one, I would love to see more FF content on your website.

Stay tuned... :neenernee

I never would have thought this, but there's still an amazing wealth of technical information about the game being discovered, even in the past few months. Here's a thread on the Gamefaqs message board describing most of the known formulas. Some of the highlights:

- There's a hidden "Magical Defense" stat, that determines the success chance of status-affecting attacks, like stun and Mute and Rub (death is a status.) With this, the precise chance of a Sorcerer's instant kill is now known. Magical Defense goes up considerably with character levels, providing another upside to experience grinding.

- Did you know that damage spells don't quite behave as advertised? FIR2 deals 30-60 damage, not 30-120. Then there's a chance for the damage to be doubled, depending on the target's Magical Defense. That's why elemental spells always deal even numbers of damage in the upper half of their ranges, and why Fiends never seem to take as much damage as they should (they defend against the doubling.)

- Critical chances are now known exactly for each weapon (highest is Masmune at 20.5%.) An unarmed BB has a crit rate equal to his level (calculated as 2*level/200.) So a level 50 BB crits half the time, or on 5 out of his 10 hits. This forms yet another multiplier to the BB's incredible damage increase with levels: it's not just quadratic (hits * damage), but cubic (hits * damage * crit rate.)

- Turn order is entirely random, not influenced by stats or levels or classes or status conditions.

- Run chance is precisely known now, including a dazzling variety of bugs. Exploiting this knowledge with other characters renders a Thief's running ability basically irrelevant.

- Hit% and Evade mechanics are precisely known. It's similar to Diablo 1, with the formula essentially being 168+Hit%-Evade% out of 200.

I'm still digesting all this info and figuring out what to do with it...

[Sullla]

That's... some really useful info there. I thought I knew everything there was to know about FF after downloading and reading through one of the massive (330KB) FAQs that one team put together, but evidently not. That info explains a LOT though; I've wondered for years how the critical hit chance was calculated, and why certain enemies (like the fiends) always seemed to take less damage than they should from spells.

Also, can I link to your zerks report from my own website? I figure anyone who stumbles across my unfinished concept deserves to read how it played out.

[KingOfPain]

Hey guys, you may want to add your future "Other Games" HERE
Or dig up your old ones and add them too.

KoP

[timmy827]

- Turn order is entirely random, not influenced by stats or levels or classes or status conditions.

Looking at their actual formula, the nerd in me is immediately curious whether that method actually produces a totally random order, or if there is some residual skewing. Can't recall anything offhand on the subject, time to read up to see if there is an analytical approach, tracking 91^17 events into 13! buckets might take some time for my computer to chew through...

[dathon]

@timmy827

Well, doing it analytically would probably be a major headache, but if you could easily code something to run, say 1 billion trials and get pretty good empirical statistics. I'm actually tempted to do that in Matlab, as the code would literally take 5 minutes to write, and Matlab is so good with handling large chunks of data. The real question is, exactly HOW are the random numbers determined? That will be the lion's share of any residual skew. So basically what we need is a formula for how the NES calculates random numbers. Or, to put a finer point on it, how the emulators do it (if that's even uniform across emulators. The answer could be platform-specific!)

dathon

[T-hawk]

There's definitely a skew in that algorithm. The classic shuffle-swap algorithm loops through the array, swapping each element once with a randomly chosen element. FF1's algorithm (as reported in the Gamefaqs thread) does not loop, but swaps randomly-chosen elements. So there's a fair chance that any particular element will never swap at all, so each monster or character is more likely to remain in its original position than to swap to any other particular position.

Let's quantify it. FF1 generates a total of 34 random numbers to apply to an array of 13 elements. We can simply apply the binomial theorem. 34 trials, 1/13 chance of success, means that each actor has a 6.6% chance of going completely untouched by the shuffle algorithm. Dividing the remaining 93.4% among the 13 slots evenly (an element can be swapped to itself) gives 7.2% chance for each other slot, but a total 7.2% + 6.6% = 13.8% chance to remain in the original slot.

The Light Warriors begin at the end of the list. So if it always seemed that the Light Warriors got shafted and would act last disproportionately often... that's correct!


To answer dathon... The random numbers are generated by a pseudorandom algorithm in assembler code, as they are for all console games. For the NES, it's 6502 assembler. Emulators can't detect and emulate high-level concepts like a PRNG algorithm, all they do is execute the same CPU instructions as the game ROM, so they'd all behave the same way. (Modern RNG can use sophisticated random concepts like electrical noise or quantum effects, but 8-bit NES hardware is a wee bit shy of that. )

If Timmy really wants to analyze the game's actual results, he'd have to replicate FF1's PRNG logic as well as the shuffle algorithm. Actually, a better way is just to run FF1 itself in an emulator, get into a battle, and examine the memory locations that hold the turn order. (One battle can run unlimited iterations of the turn order shuffle, if the characters keep healing themselves and don't kill the monsters.)

[Kylearan]

Hi,

There's definitely a skew in that algorithm. The classic shuffle-swap algorithm loops through the array, swapping each element once with a randomly chosen element.

A nerdy clarification: You have to loop through the array and swap each element once with a randomly chosen element with an index not smaller than the current loop index. If you swap with any randomly chosen element, the distribution of possible outcomes will not be uniform.

-Kylearan

P.S.: I've never played FF, but I've always be a fan of oldschool RPGs (guess where my name comes from). Reading through the report really whets my appetite...maybe I should try it out.

[Deceptus]

Gah. I feel almost ashamed as a member of the newer generation, who prefers Baldur's Gate and Fallout to the older titles. Nobody discusses those titles here, what a shame.

Well, not really ashamed... I am still playing ADOM. I'm still trying to overcome the really crazy interface of Nethack. Final Fantasy... I tried it months ago and the game is quite fun. Which is a surprise to me, because I don't know what's so much fun with that, at the first glance, boring game.

I once sent an e-mail to Sullla, talking about what can be found fun in any game, inspired by the "the best cRPG made in years". No idea what did I wrote (I never got the answer), but I asked about the other games of the genre (like BG, Fallout, Arcanum, etc.) compared to FF, but found out that while each of these games aren't perfect, they all have that crucial element that makes the game fun, so it's really hard to compare them.

[FoxBat]

While we're sidetracking on old-school RPGs....

AGD has just released a VGA remake for Quest for Glory II: Trial By Fire. If you missed out on this one due to the archaic text interface or amazing 16-color graphics, now is the time to check it out, or take a trip down memory lane.