Steps ------ Don't let all these function addresses throw you; Variable $00BE is the thing to watch, as that decides where we read from our Random Number Table when picking an encounter type. 1) The Enemy group has already been picked. 2) C2/23ED initializes $00BE to Variable $021E (the frame counter, ranged 1-60) * 4. Hence it's always a multiple of 4 between 4 and 240. :P Skip to Step #7 if you don't feel like wading through crap, and would rather take my uncertain word for things. 3) C2/30E8 calls random number function (C2/4B5A) to further pick the formation, and $00BE is incremented by the random # function. (NOTE: this step only applies to the Floating Continent, where you can randomly get 4 formations for each entry in a 4-pack.) 4) C2/23ED calls random number function (C2/4B5A) 10 times; each call increments $00BE by 1. 5) In function C2/2C30, the Vigor for every monster (active or not) is randomly calculated using C2/4B5A; each call increments $00BE by 1. 6) C2/083F calls C2/09B4 for each target onscreen, though it apparently skips the call for dormant enemies. (Actually, it checks a mystery bit whose meaning I'm not quite sure of, so I'm going largely on observation here.) - C2/09B4 calls random number function C2/4B65, which increments $00BE. 7) Now it's time to randomly pick an encounter type. C2/5247 is responsible for that, and it calls C2/4B65, which increments $00BE before giving us our random number. 8) So, assuming we're not on the Floating Isle: Variable $00BE = (Value of $00BE After Step 2) + 10 + num_total_enemies + num_chars + num_active_enemies + 1 If we're on the FC, just add another 1 to that. *** Steps 9 through 11 just apply to my epic hunt for a lone Latimeria that pincers the party. Skip them if you don't care about that. *** 9) If Front, Pincer, and Side (but no Back) attacks are allowed, we need a random number between 1Fh and 26h to get Pincer. In this case, 0 - 1Eh give Side, 1F - 26h give Pincer, and 27h - F6h give Front. As you'll see in the Algorithms FAQ, the chances of each formation are given as X/247 rather than X/255 when all but Back is allowed (we've effectively cut out Back's 8/255 probability). Now C2/4B65 gives us ( Input * Random Number Table Value ) DIV 256 as a result. Thus, with an input of 247, we need a Random Number Table value from 21h to 28h to get Pincered. Note the 256-byte random number table can be found at file address 0FF00. 10) Those table values are found at positions 37h, A4h, 84h, 64h, 80h, 24h, 4Ah, and AFh, respectively. But note that for a 5-target brawl (my 4 party members plus Latimeria), $00BE will always (??) be 17d (11h) greater than it was in Step #2. This rules out A4, 84h, 64h, 80h, and 24h right off the bat, as they're all multiples of 4, and 17d is obviously not. Now to examine the remaining three: 4Ah - 11h = 39h. That's not divisible by 4, so we never could've had it in step #2. 37h - 11h = 26h. That's not divisible by 4, so we never could've had it in step #2. AFh - 11h = 9Eh. That's not divisible by 4, so we never could've had it in step #2. 11) From those numbers, it appears lone Latimeria (or L.90 Magic) will be able to pincer a party of 1, 2, or 3 characters just fine. I got that to happen with Lati and 2 characters in a matter of 8 minutes (not hours! garghh!!@), so my steps are still holding up. I can't see Front or Side attack ever being thwarted by this oddity, as they have such a wide range of numbers that will yield them (see the chances in Algorithms FAQ). However, Pincer and Back attacks can be affected, sometimes in less "convenient" situations (nobody's losing sleep over one enemy not being able to pincer your party). Obviously, none of this applies to forced encounter types (dual Tyranosaurs, the IAF, etc). Also, keep off the Back Guard if you want to reproduce these results. Holy shit; this is like ZED's Slot Guide, except of zero significance to the player!! :D Now if people can invent ways to *avoid* nasty pincers without using a relic slot, we'd almost have something marginally useful.. Thanks to SeR (aka SprintGod) for deciphering the frame counter variable, and Jyzero for bringing it to my attention.