My fixes

Phantasie III keeps track of wounds for every monster, part by part: the head, the arms, the torso and the legs (the game's body parts). Wounds matter. A monster with a broken leg hits less often, and one with a broken head or torso drops out of the fight before its hit points are gone. So you would expect a monster's wounds to belong to that monster.

In the 1987 game they belong to its place in the line instead. A monster group stands in up to five places, and when one of them dies, the survivors close ranks at the end of the round: each moves down one place. The game moves each monster's hit points, magic points and condition down with it, but it leaves the wounds behind in the place they were made. The monster that steps into a badly wounded one's place takes over those wounds, and its own are left behind for whoever moves in next.

What a player would notice: a monster you never touched suddenly has a missing arm, or dies from a scratch, because it is carrying another monster's injuries. A monster you had already hurt badly can come out of the shuffle with none of its injuries at all.

I watched this happen in the running game. With the emulator's debugger I put wounds on the first two monsters of a Gremlin group and killed the first. After the round, the hit points and magic points moved down one place and the wounds stayed where they were, so the Gremlin that moved into the front place carried the dead one's torso wound.

With the fix, every monster takes its own wounds with it when the group closes up, so nobody inherits anybody else's injuries. Nothing else about closing ranks changes: the monsters still move down in the same order at the same moment. It is one of my fixes.

When a hero's blow is big enough to wound a monster, the game rolls which body part it lands on: head, arm, torso or leg. A wound to the head is the most dangerous one. A monster whose head is broken drops out of the fight, even with hit points left.

The game has a rule that some monsters have no head a hero can reach. Fight one of them and no wound ever lands on a head. It feels deliberate, and a good idea: you cannot hit a giant on the head. The list in the 1987 game is the Storm Giant, the Cloud Giant and the Dark Gnome. The Dark Gnome stands out. The Valley Giant, the other giant in the game, is the monster numbered one below the Dark Gnome in the monster list, and it is not on the list.

My reading, which is not something the program can prove: the list was meant to be the three giants, and the Dark Gnome is a slip of one in a list typed out by hand. So my fix puts the Valley Giant on the list in the Dark Gnome's place. With the fix, Valley Giants keep their heads out of reach and Dark Gnomes can be wounded on the head like anyone else.

What the fix does not change is how the rule works, which is a little odd in itself. The game checks once, when the fight begins, which kind of monster is in the front group. If it is one of the listed kinds, no blow in that whole fight lands on any head, including the heads of the other monsters fighting beside them. If the listed kind is only in the back group, the rule does not switch on. My fix keeps all of that as it was, and only changes which three monsters are on the list.

The rule also stops the called shot, Aim Blow, from ever reaching a listed monster's head. The program has a special case that looks meant to let an aimed blow reach the head of a giant, but it can never run, because the no-head rule has already taken over by the time the game looks. My recreation keeps that special case dead, as in 1987, with or without this fix.

I checked the 1987 behaviour in the running game: in a fight with Dark Gnomes in front, the game's no-head marker was switched on (and off in every other fight), and none of the heroes' wounds that fight landed on a head. It is one of my fixes.

Everything in the game tells you heavier armour is better. The armoury prices armour by its point value, the game readies the best armour a hero can wear, and the 1987 manual advises buying "the heaviest armor you can get for fighters in the front ranks".

In the 1987 game, armour does two things, and neither is what you would expect. First, every point of armour makes the hero easier to hit: at the start of each fight the game works out how hard he is to hit (his defence) and takes five off it for every point of armour he wears. Second, the armour soaks up damage from exactly one kind of attack, a monster's breath. Swords, claws and bites do their full damage whatever he wears. So outside a fight with a breathing monster, heavy armour is a handicap: the hero gets hit more often, and each blow hurts just as much.

I tested this in the running game: six heroes with the same numbers, three in Clothing and three in Plate Mail, against three Undead Gnomes for about 90 rounds. The heroes in Plate Mail were hit 87 times in 100, the heroes in Clothing 46 times in 100, and every hit did the same 1 to 9 damage in either armour. The full story is in the finding on armour.

With the fix, worn armour also takes damage off every monster blow that lands on a hero: half the armour's value, rounded down, with at least 1 damage always getting through. Plate Mail, worth 10, takes 5 off each blow. Spells that add armour in a fight, such as Protection, now help against blows too. The lost defence stays: armour still makes a hero easier to hit, so the choice becomes a trade, being hit more often but less hard, which is what I think the manual describes.

What the fix leaves alone: breath still takes the full armour value off, as in 1987. Monster spells still ignore armour. And it is for the party only: the monsters' own armour values are still read by nothing.

The size of the soak is my own design, modelled on the way the breath already uses armour. It is not a guess at the numbers the authors had in mind, because the program does not show them. It is one of my fixes.

Every fight opens with a menu, and one choice is Run. Normally Run rolls the dice and can fail, and a failed Run gives the monsters a whole round of free blows. There is one time it never fails: when the party has surprised the monsters, or found them asleep, it may run from that fight without a roll.

The game keeps that free escape as a single marker. A round that the fight survives clears it, and so does the next meeting with monsters out on the map. But a party that runs at once, before the first round is played, ends the fight before anything has cleared the marker. And a fight inside a dungeon never resets the marker when it starts. So the next dungeon fight begins with the free escape still set, and Run works there without a roll, exactly as if the party had surprised those monsters too.

It can carry on from there. Running from that dungeon fight also ends it before the first round, so the free escape rides into the fight after, until the party stays in a fight for a whole round or meets monsters on the map again. I saw it in the running game: a party surprised monsters outside Pendragon, ran at once, and the free escape was still set in the Pendragon Town Archives, where Run worked at once. The full story is in the finding on the free escape.

With the fix, every dungeon fight starts without a free escape, so Run in a dungeon rolls the dice as it should. A party that surprises monsters, or catches them asleep, still runs from that same fight for free, and fights out on the map work exactly as in 1987.

It is not one of my fixes: with my fixes on, this plays as in 1987. You can turn the fix on yourself with Choose my own.

Many fights in Phantasie III are against two monster groups at once, one in front and one behind. When the front group is wiped out, the group behind moves up to take its place. The groups can also change places at the end of a round: when the group behind is the more dangerous, there is a 30 in 100 chance they swap.

When the fight is won, the game pays out gold and experience for every monster killed, each kind at its own price. Because the groups may have changed places, it first has to match each count of kills with the right kind of monster again. It does that by swapping the two counts. The trouble is that it works out the reward twice, once for the gold and once for the experience, using the same two counts, and the second time it swaps them back. The gold comes out right. The experience pays each group's kills at the other kind's price.

What a player would notice, if they did the sums: the wrong experience on the reward screen. In a fight I watched in the running game, the party killed four Wraiths (worth 3605 experience each) and three Wights (worth 1414 each), and by the numbers the groups had changed places during the fight. The game paid the Wraiths at the Wights' price and the Wights at the Wraiths': 16,471 in all, 2745 for each of six heroes, instead of the 18,662 the party earned. A strong group with a weak one behind it pays too little this way; a weak group in front of a strong one pays too much.

It only happens when the two groups are different kinds of monster and have changed places by the end of the fight. Two groups of the same kind pay the same either way, and a fight where the groups never moved is paid correctly.

With the fix, the experience is worked out from a fresh copy of the kill counts, so the matching is done once, as it is for the gold, and every kill pays its own kind's experience. The gold is not changed, because it was already right. It is one of my fixes.

Every blow in a fight gets a line in the fight log, such as "THIEF takes 15 damage!". The game builds that line in a fixed space: the name, then "takes", then the number, and then the words "damage!" placed as if the number could never be longer than two digits. A number of 100 or more has its third digit written over by the word "damage!".

So a blow of 150 is printed as "takes 15 damage!", and a blow of 176 as "takes 17 damage!". Only the text is wrong. The full 150 is taken off the hit points, so a hero can die from a blow the log says was only 15.

It takes a big blow to get there. A monster's strongest fire spell, Fireflash IV, can do up to 120 to one hero, and a veteran hero striking several times in one turn can pile up a total of 100 or more, because each blow of a turn adds to a running total. It works the same way on both sides, for blows on heroes and on monsters.

I saw it in the running game, with the debugger giving a monster a much bigger damage roll: a blow of 155 printed "Mage takes 15 damage!", and the hero's hit points fell by the full 155.

With the fix, the word "damage!" goes after the last digit, so the line shows the whole number. Nothing else changes: the damage itself was always right. It is one of my fixes.

The stronger monsters strike more than once a round: a Dragon twice, Boo Boo Kitty three times, the Light Fairy five times, Nikademus nine times. In the 1987 game, the moment one of those blows misses, the monster's turn is over. "Nikademus misses", and he is done for the round.

Why it happens is a small programming slip. The game counts a monster's blows with a counter, and the part of the program that prints "NAME misses" uses the same counter to step through the letters of the monster's name. After printing, the counter is left at the length of the name. So after a miss, the game carries on from blow number "letters in the name". Every monster in the game has at least as many letters in its name as blows a round (Nikademus: nine blows, nine letters), so in the shipped game every miss ends the attack.

A player would notice it only as monsters with several blows landing fewer of them than they should: the more blows a monster has, the more it loses. A monster with more blows than letters in its name would behave differently, swinging on after a miss and even going back to swing again, but no monster in the game is built like that. I tried it in the running game with the debugger giving monsters more blows: a Viper Moth, ten letters, stopped at its first miss every time, and an Orc, three letters, swung on.

With the fix, every blow is rolled, hit or miss, and the turn ends after the last one. Nothing else about a monster's attack changes: how many blows it has, how often each one lands and how hard it hits stay as in 1987. Because the strongest monsters gain the most, this fix makes the hardest fights harder. It is one of my fixes.

Four monsters have a stunning blow: the Small Devil, the Wight, the Wraith and the Vampire. When it lands, the hero is hurt and then put to sleep (stunned) until he wakes up again later in the fight.

The 1987 game stuns the hero after the damage is done, without asking whether the blow killed him. It only checks that the hero is not already out of the fight. So when a stunning blow kills, the fight log reads "THIEF dies!" and then "THIEF is stunned!", and the dead hero is marked as stunned instead of dead. It does not last. The first time he passes the roll to wake up, the stun wears off and he is dead again.

I saw it in the running game, with the debugger giving a monster the stunning blow: a hero was killed by it, the log said he was stunned, his hit points read 0 and his condition read stunned, and it went back to dead at his first successful wake roll.

With the fix, a hero the stunning blow has killed is left dead and is not stunned, so the log no longer says "is stunned!" over a dead hero. A hero who survives the blow is stunned exactly as in 1987. It is one of my fixes.

A monster whose leg is broken is meant to fight worse. When it attacks, the game takes 40 off its chance to hit, a big penalty: a blow that would land 90 times in 100 lands only 50 times. It is one of the ways wounds to a monster's body parts pay off for the party.

In the 1987 game the check looks at the wrong monster. A monster group stands in a line of up to five, and when any monster of the group attacks, the game asks whether the first monster in that line is lame, not whether the attacker is. So break the leg of the monster at the front, and every monster in its group hits you less often. Break the leg of one further back, and it fights as well as ever.

What a player would notice is a group that suddenly misses a lot after one blow to its leader, while other lame monsters seem unbothered. I saw both halves in the running game with a group of Gremlins: with the first one lame, every Gremlin's chance to hit read 50 (90 less 40); with the second one lame and the first one sound, every Gremlin's read 90, the lame one included.

With the fix, each monster's own legs count. A lame monster anywhere in the line attacks at 40 less, and the rest of its group attacks as normal. The size of the penalty is the game's own and is not changed. It is one of my fixes.

Monsters can lose their nerve. At the end of every round the game asks whether the monsters' strength is now below half of what it was when the fight began, and if it is, a 50-50 roll decides whether "The monsters flee!". The strength is meant to add up the monsters still standing, each weighted by how bold that kind of monster is. Nothing from the Dark Gnome onward in the monster list ever flees, whatever the count says.

The 1987 game counts the wrong monsters. Instead of counting the monsters that are still alive, it counts only the ones whose hit points are an odd number. All monsters of one kind start with the same hit points, so a kind whose hit points start even, such as the Sniverling with 10, starts at a strength of nothing, and nothing can fall below half of nothing. Those monsters never run away. That is 53 of the 80 kinds of monster.

The kinds that start odd waver instead. A blow that leaves a Skeleton on an even number of hit points takes it out of the count, so a scratch can make a group run while a heavy blow that leaves it on an odd number does not. I saw both in the running game: the starting strength read 0 for every even kind I met (Viper Moths, Orcs, Wild Cats, Dark Gnomes), and two Skeletons I changed with the debugger from 3 hit points to 4 ran away by the second round, unhurt, from a party that had not struck a blow.

With the fix, every living monster counts, whatever its hit points. Monsters of every kind can then lose their nerve as the fight turns against them, and the odd-numbered kinds no longer run from a scratch. The rest of the rule stays: the half-strength test, the 50-50 roll, and the monsters that never flee.

It is not one of my fixes: with my fixes on, this plays as in 1987. The fix rests on my reading that the odd-number test was meant to be an "is it alive" test, and that is a guess about what the authors meant, not something the program shows. So I chose to keep the 1987 behaviour. You can turn the fix on yourself with Choose my own.

Binding is a spell a hero can cast in a fight, in four strengths, Binding I to IV. The 1987 manual says Binding spells "inhibit the movement of monsters, making them easier to hit", and that they are "particularly effective against flying creatures and monsters that are difficult to see". In the 1987 game, the magic points go and nothing else happens. The fight carries on exactly as it would have.

Binding does change something: it cuts a number in both monster groups by a percentage. But the number it cuts is the monsters' armour, and the game never reads a monster's armour at all. Every monster has an armour value, it is loaded into every fight, and nothing in the program ever looks at it. I tested this by running the original code: 500 short fights with the monsters' armour forced to every kind of value came out the same down to the last line of the fight log. I also watched Binding cut the armour in the running game, and no other number moved.

The number beside the armour is the monsters' defence, the value a hero's blow has to beat to land. The monsters the manual names have high defences: the Death Fly 35, the Illusion 60, the Light Fairy 90. A percentage cut does most against the biggest values. My reading, from the manual and from where the two numbers sit, is that Binding was meant to cut the defence and the program points one number too early.

With the fix, Binding I, II, III and IV cut both monster groups' defence to 85, 70, 55 or 45 percent of what it was, so every hero's blow and arrow lands more often. Each cast cuts again from the new value, as each cast cut the armour again in 1987. A monster whose defence is already below 0, which is 43 of the 80 kinds, is left as it is: those monsters are already easy to hit, and a percentage cut would move their defence up toward 0 and make them harder to hit, the opposite of what the spell is for.

The spell's cost and everything else about casting it are as in 1987. It is one of my fixes.

The game's list of monsters includes a Wolf, but nobody playing the 1987 game ever met one. When the dice chose a pack of Wolves out in the wilds, Small Dragons came instead, and nothing on the screen suggested that anything was missing.

Every monster is a short list of numbers in one file: its hit points, its magic points, which picture to draw, and so on. In the Wolf's list one number, the size of its damage roll, went missing, and every number after it moved up one place. Read like that, the Wolf has no hit points and a picture that does not exist. The program then has a check, right at the end of picking a monster for a fight in the wilds, that turns a Wolf into a Small Dragon, which looks like someone kept the broken Wolf out of the game on purpose rather than mend its numbers. The other way a Wolf could appear, as a companion of the High Devil, fails too: the broken picture number makes the game leave those companions behind, so the High Devil fights alone. The full story is in the finding on the Wolf.

The missing number is known because the Commodore 64 version of the game keeps the same monster list, and there the Wolf is whole. A fan on the English Amiga Board first pointed the way. With the fix, the lost number is put back from that version and every number after it goes back to its place: the Wolf comes in packs of up to five, with 14 hit points, 7 magic points and the Wild Dog's picture. The check that swapped it for a Small Dragon now lets it through, so Wolves appear in the wilds and behind High Devils.

This changes some fights quite a lot. Where a Small Dragon would have come in its place, a much tougher monster with a breath attack and more than twice the Wolf's hit points, the party now meets a pack of Wolves, and is paid the Wolf's experience for them, not the Dragon's.

What the fix leaves alone: the same check also turns the Mud Elemental into a Small Dragon, and that stays as in 1987. Small Dragons still come wherever the dice choose a Small Dragon. The dungeons are not changed, because none of them ever calls up a Wolf.

It is one of my fixes. Until October 2026 I kept it as a curiosity, a thing recorded and left alone, and then I chose to bring the Wolf back.

Flame Arrow is a spell that makes a hero's bow better for one fight, in two strengths. Party Flame Arrow does the same for every hero. The 1987 manual, which calls it Arrow Flame, says Flame Arrow I "increases accuracy by +50 percent and Bow Point Values by +15", and Flame Arrow II by +100 percent and +25.

The accuracy part works as promised. The damage part does not. At the start of a fight the game turns each hero's bow into a damage roll: twice the bow's point value, plus 4. A Long Bow, worth 14, hits for up to 32. Flame Arrow I does not add to that roll. It throws it away and puts the bow's point value plus 15 in its place, without doubling it. For a Long Bow that is 29, three less than before the spell.

For a weak bow the spell still helps: a Short Bow, worth 4, goes from 12 to 19. The two come out equal at a point value of 11. Above that, Flame Arrow I makes the bow worse, and the better the bow, the more it loses. Of the strong bows a party can find in Scandor's dungeons, the Gnome Bow drops from 30 to 28, the Long Bow from 32 to 29, and the Old Bow, a quest reward, from 40 to 33. I saw it in the running game: Flame Arrow I took a Long Bow from 32 to 29, and Party Flame Arrow took a God Bow from 44 to 35. Flame Arrow II, at plus 25, never makes a bow in the game worse, because no bow is worth more than 20.

With the fix, the spell adds 15 (or 25 for Flame Arrow II) to whatever the bow already does, as the manual promises: the Long Bow goes from 32 to 47, the God Bow from 44 to 59 (69 with Flame Arrow II). The accuracy bonus is the same as in 1987. Casting it twice still does not add up: a second Flame Arrow replaces the first, as in 1987 and as the manual's spell list says for this kind of spell.

Which addition the authors meant cannot be told from the program, only that the spell was meant to increase a bow's damage. Adding the bonus to the bow's usual roll is my reading of the manual's promise. It is one of my fixes.

When a trap goes off in a dungeon, a line of text at the bottom of the screen says what happened. In the 1987 game that line is usually the wrong one. In the Pendragon Town Archives an arrow trap tells you "Molten lead is poured from the roof." The arrow line the designers wrote, "Arrows fire from a slit in the roof.", never appears at all. Only three dungeons have traps of this kind, and the same slip happens in the other two: in the Giant Caves the spears line stands in for the falling slab, and in the Gnome Caves the falling rocks stand in for the poison gas. A player who did not know what each trap was meant to be would simply believe the text.

It is an off-by-one. Each dungeon keeps its own numbered list of messages, and there are five kinds of trap. The designers put the text for each kind of trap in order: trap kind 2 at message 5, trap kind 3 at message 6, and so on. The program asks for the message one further along. So every trap shows the text written for the next kind of trap, the first trap text of each dungeon is never shown, and the highest kind of trap in each of the three dungeons asks for a message that is not there.

That missing message leads to the strangest part. With nothing to point at, the game reads whatever lies just before the dungeon in memory, which is the space the scrolls are read into. Until somebody in the party has looked at the map of Scandor (Scroll IV) since the game was started, that space is empty and the line stays blank. The map is a picture, though, and fills most of that space, so after it has been read the same trap prints seven odd symbols: a scrap of the packed map picture shown as letters. Both the blank line and the seven symbols were seen in the running 1987 game. More about the map in The map of Scandor is Scroll IV.

The fix makes each trap show the text the designers wrote for it, so no shipped trap reaches the missing message and the stray symbols cannot appear. It changes only the words. What each trap does to the party, how often it springs and who spots it are all as in 1987.

This one is not among my fixes: played that way, the traps keep their 1987 text. A player who wants the right words can turn it on with Choose my own.

In a dungeon, some squares hold a guardian: step onto one and a fight starts. Win, and the party moves onto the square. Run, and the party normally stays where it was, with the guardian still waiting. Random fights on ordinary floor squares work the same way. But in the 1987 game, if the party was standing in an open doorway when it stepped into the fight, running away still moves it forward onto the monsters' square. The guardian is not beaten: it is still there, and stepping off and back on starts the fight again. The same happens when the party stands on a trap that has already gone off.

There is a second effect. If the party was standing on a floor square where random fights happen, a guardian's fight can be followed at once by a random fight, as if the party had just stepped onto that floor square again.

Both come from one slip. After a dungeon fight, the game goes on to finish the step the party was taking, and it looks at a square to decide what happens next: whether it is an open door or a spent trap (which always lets the party through), or a floor where random fights happen. It is meant to look at the square being stepped onto. But the fight redraws the dungeon, and the last square drawn is the one the party stands on, so the game ends up looking at the party's own square instead. It was seen in the running 1987 game: from an open doorway in the Pendragon Town Archives, the party ran from a random fight on the square beyond, twice, and both times it ended up on that square.

The fix makes the game look at the square the fight was for, as it means to. A Run then always leaves the party where it stood, and a guardian is never followed by a surprise second fight. Winning a fight, the guardian staying on its square after a Run, and how often random fights happen are all as in 1987.

The correction is my own design: the part of the program that runs here is the one for the square being stepped onto, so that is the square it now tests. It is not among my fixes; a player can turn it on with Choose my own.

Almost nobody will ever meet this one. But if it happened, the whole party would lose four skills at once: nobody would hear monsters coming (in the open or at a dungeon door), find a trap, find treasure after a battle or pick a lock, however good their skills. It would last until the party arrived in a town whose armoury builds a shorter list.

Every one of those skill rolls goes through one part of the program, and it is meant to skip a member who is out cold. It checks the wrong place: instead of the member's own record, it reads the word just in front of the party in memory, as if there were a member before the first. If that word reads "out cold", nobody gets a roll at all. That word is not part of the party. It is the 73rd entry of a list the game builds every time the party arrives in a town: one entry for each kind of item the town's armoury has in stock, holding the strength that item needs. A town has to stock at least 73 kinds of item for the list to reach that far, and the 73rd has to be an item whose number happens to read "out cold". Only one item fits: the Dagger +1, with 72 kinds of the items before it in stock as well.

That is why you would never see it in a normal game. Every armoury in the game as it ships holds 15 kinds of item or fewer, so the list never comes close. Selling an item adds its kind to that town's stock, so a very patient collector could in principle build such an armoury. It has been tried in the running 1987 game, with an armoury edited to those 73 kinds: Listen, Find Trap and Pick Lock all set to 250, as high as a skill goes, and still nobody heard the monsters, found the trap, found treasure or picked the lock.

The fix keeps the armoury's list away from the party's skills: the rolls always happen as they do with a normal armoury. Nothing else about the armoury or the skills changes. It is not among my fixes; a player can turn it on with Choose my own.

Some things that happen in a dungeon are events the designers wrote out step by step: open a chest, and a trap may go off, then a line of text tells you what you found. Each step can come with a sound. In the 1987 game, a few of those sounds are never heard. A shock trap (a burst of poison gas, for instance) that is followed by a line of text goes off in silence. It happens with shock traps in three dungeons: the Giant Caves, the Dwarven Burial Grounds and the Castle of Dark.

Why: a few kinds of step do not play their sound straight away. They leave it waiting, and the game plays the waiting sound once the whole event is over. But every new step starts by clearing whatever sound is waiting. So when a shock trap is the last thing that happens, its sound plays at the end as it should; when a message comes after it, the message clears the waiting sound before anyone hears it. It was recorded in the running 1987 game: a chest in the Giant Caves set off a poison gas trap, then "You find 2 nice gems." came up, and the trap's sound never played. A shock trap with nothing after it, in the Pendragon Town Archives, played its sound as normal.

The fix plays a waiting sound before the next step clears it, so every event makes its sound. What the trap does, the text and everything else about the events are as in 1987. Gold handed over without a box works the same way: with the fix, a coin sound that would have been lost is heard too, and when the gold is the last step its sound still comes once the event window has closed, as in 1987.

This is one of my fixes.

Heroes in Phantasie III grow older as time passes (see ageing). Save in a town, switch off, and Continue the next day: in the 1987 game every hero is a little older than when you saved. Continue from the same save again, and they are older again, by the same amount, every time.

Why: the party ages when it arrives in a town, by the time it spent outside. That time is kept by the step clock, which runs while the party travels; on arrival it is turned into fortnights and added to every living member's age. The clock is cleared only when the party leaves a town, not when it arrives, and saving in a town writes the clock into the save as it stands. A Continue counts as arriving in the town, so the saved clock is applied again at every Continue.

It was seen in the running 1987 game. With a save in Pendragon holding 140 days on the step clock (10 fortnights), Continue aged every member by 10 fortnights: the Thief's age, which the game keeps in fortnights, went from 338 to 348. Saving again in town left the clock at the same 140 days, and the next Continue aged them by another 10, to 358.

Does it matter? Very little. A hero's age changes nothing in play: no attribute falls, no skill fades and nobody dies of old age (see Heroes do not get worse with age). The age only shows on the character sheet.

The fix skips the ageing on the arrival straight after a load, because that time was already counted when the party arrived before the save. Any time the party spends on the road after the load still ages it as in 1987. It is not among my fixes; a player can turn it on with Choose my own.

Out on the map, stepping onto a town asks "Will you enter?". Behind the scenes the game keeps a note of which town is the current one: the town the party last visited, which decides things such as which town's name is shown and which armoury stock is used. In the 1987 game, simply stepping onto a town and answering No still makes it the current town, even though the party never went in. It stays that way after the party walks on.

Why: the part of the program that notices the party standing on a town writes the town's number into that note every time, before the question is asked, whether or not the party then enters.

You will almost certainly never notice. Nothing that a player can see has been found to follow from it. It was confirmed in the running 1987 game by reading the game's memory: with Asmith the current town, the party stepped onto Pendragon, declined to enter, and the note read Pendragon, and still did after the party stepped off.

The fix makes a town the current one only when the party goes in. It is not among my fixes; a player can turn it on with Choose my own.

Phantasie III has no boats. A party that walks into the sea or a lake swims, and every step in the water can hurt: each member tests his Swimming skill, and a member who fails is hurt, which the game shows in a Damage Report. In the 1987 game, the first step out of a town after loading a save is special. If that step goes straight into the water, the party swims twice for it: two tests each, two Damage Reports, twice the chance of drowning. Every later step swims once.

Why: when the party leaves a town, the game redraws the map and works out what the party's new square does to it, the swim included. Then the main walking loop works out the same square again. Normally a flag makes the first of those two harmless. But that flag is only set once a "Will you enter?" question has been answered, and straight after a load nobody has answered one yet, so both count.

It was seen in the running 1987 game. From a save in Pendleton, a town by the sea, the party continued, left the town and took one step north onto the water: the game ran the swim twice and showed two Damage Reports. The next step showed one.

The fix makes that first step out after a load swim once, like every other step. Everything else about swimming, how likely a member is to be hurt and by how much, is as in 1987. This is one of my fixes.

Kill Nikademus (or Lord Wood) and you see the ending: a storm, a page with a god, the Ode to Joy. Then the party stands in Pendragon with its new rank, the dead alive again with 1 hit point. And then, in the 1987 game, play simply goes on. There are no credits and no way out, and nothing has been written to the game disk. Switch off now and the victory is gone: the next Continue brings back the party as it was at the last save. It is kept only if you think to save in the town afterwards.

Why: the part of the program that plays the ending heals the party, gives every member the new rank, score and class, puts the party in Pendragon and hands back to the game. It never calls the save.

It was seen in the running 1987 game. After the Nikademus ending every member was a Superhero and the dead Fighter was alive with 1 hit point, but the disk had not been written: after a reboot and Continue, the Fighter was dead again with his old rank and score. After the other ending, against Lord Wood, saving in the town afterwards kept the new ranks through a reboot.

The fix saves the game straight after the ending, exactly as the town's Save would, as soon as the party arrives in Pendragon. The ending itself, its pictures, words and music, and what happens to the party are as in 1987. This is one of my fixes.

Every hero in a fight stands in a rank: the front rank, the middle, or the back rank. After the ending, the 1987 game puts every member in the front rank, fighters, casters and thief alike, and play goes on that way. The new ranks are kept in the heroes' records and saved with them, so they last until you change them. What a player notices is that the best fighters in the land suddenly miss blow after blow. I remembered exactly this from 1987 and for nearly forty years put it down to my heroes getting old. Age has nothing to do with it.

Why it hurts: a hero's bonus to hit does not come from standing in front as such. The game works out the party's average rank, rounded down, and gives a hero 30 points (out of 100) on his chance to hit for every step he stands ahead of that average. In a usual party, with the fighters in front and the casters behind, the average is the middle or back rank, so a fighter in front gets at least 30 on every blow. With everyone in front, the average is the front too, and the bonus is gone. The same rule works for the monsters: a hero behind the party's average is harder to hit, one ahead of it easier. With everyone on one line nobody gets either, so the casters lose the protection the back rank gave them.

It was seen in the running 1987 game: after the Nikademus ending, the party's ranks went from a mix of back, middle and front to front for all six. And a player who beat Nikademus in my recreation on 4 November 2026 found his fighters missing a lot; his saved game had all six heroes in front. The whole story is in Heroes on the same line lose their bonus.

In the 1987 game the cure is easy once you know: in the next fight, use Move to put the casters and the thief back where they belong, and it sticks. The fix makes that unnecessary: each member keeps the rank he held in the final battle. The ending's victory picture still lines everyone up as it did in 1987; only the ranks the party walks away with change. This is one of my fixes.

In Phantasie III a member who dies stays dead until he is raised, and the game's way to raise him is Resurrection, a spell with a lasting price: "The shock of dying will permanently lower the victim's Constitution one or two points," as the manual puts it, and the program does exactly that. Constitution is one of a hero's attributes, so every death and raise leaves him a little weaker for good.

In the 1987 game there is a much cheaper way. Cast any Healing spell, even Healing I, the cheapest spell in the game, on the dead member outside a battle, on the road, in a dungeon or in town, and he gets up again, with no Constitution lost at all. A Priest with Healing I is quietly a resurrection service.

Why: Healing cast outside battle never asks whether its target is alive. It adds the hit points it rolled, mends wounded body parts, and then works out the member's condition again from his body and his hit points. A body with hit points above 0 and nothing missing counts as alive. There is some small print: the hit points have to end above 0, so a low roll on a badly hurt body may need a second cast, and a member who died of a lost part, such as a head that is gone, needs Healing III or IV, which can grow it back. In the middle of a fight Healing does not raise anyone. It was seen in the running 1987 game: Healing III cast on a dead member with no head mended the head and brought him back, his Constitution unchanged.

The fix makes Healing refuse a dead member, with the game's own line "This character is unable to do this.", the one a potion gives when it is used on the dead, and no magic points are spent. Raising the dead is then Resurrection's job, as the manual describes. I chose to refuse rather than raise at a Constitution cost, because that would only make Healing a cheaper Resurrection. More in Healing I raises the dead, for free.

This one is not among my fixes: played that way, Healing still raises the dead as in 1987. A player who wants death to cost something can turn it on with Choose my own.

Every hero has a maximum number of hit points, and healing is meant to fill him up to it and stop. In the 1987 game, a member who somehow has more hit points than his maximum loses the extra when he is healed outside a battle: cast Healing on him or give him a Healing Potion, and he ends up with exactly his maximum, which can be far fewer than he had. A Magic Potion does the same to magic points.

Why: outside battle, the Healing spell works out the room left, the maximum minus the hit points now, and adds at most that much. Above the maximum that room is a negative number, so the "heal" takes hit points away. The potions set the value to the smaller of two numbers, the value plus the potion's gain or the maximum, and above the maximum that is always the maximum. It was seen in the running 1987 game: a Thief with 1000 hit points and a maximum of 8 fell to 8 from Healing I, and again from a Healing Potion; with 20 magic points and a maximum of 4, a Magic Potion left him at 4.

Most players will never see it, because a hero has to be above his maximum first, and how that comes about in a game nobody has edited is not known: it was tested with an edited save.

The fix makes a heal or a potion outside battle leave a value that is already above its maximum as it is: it never takes anything away. The fix covers only healing outside battle: Healing cast during a fight still cuts a member above his maximum down to it, as in 1987. It is not among my fixes; a player can turn it on with Choose my own.

An inn is where the party rests: hit points and magic points are filled up, wounds mend, and everyone grows a little older. A dead member does not rest. The inn says he "is resting in peace", and he is not meant to get up. But in the 1987 game, a dead member whose hit points are still above 0 walks out of the inn alive in the morning. It happens at the town inns and at the roadside inns alike.

Why: the inn's rest does skip a dead member, so he gets no healing and does not age. But at the end of the night the inn works out every member's condition again from his hit points and his body, and hit points above 0 with nothing missing count as alive. It is the same slip that lets a Healing spell raise the dead for free (see Healing I raises the dead, for free). It was seen in the running 1987 game at the Pendragon inn: a dead Wizard with 5 of his 8 hit points "rested in peace" and came out alive, still with 5 hit points and no older.

Most players will not meet it. Which ways of dying leave a body with hit points above 0 and nothing missing in a game nobody has edited is not known: the dead Wizard was set up in an edited save.

The fix keeps a member who was dead when the rest began dead after it, until he is raised. Everyone else rests exactly as in 1987. It is not among my fixes; a player can turn it on with Choose my own.

A hero carries up to nine items, and separately has a weapon, a shield, armour and a bow readied: the ones he actually uses in a fight. In a town, Trade Item lets one member hand an item to another. In the 1987 game, if he gives away something he has readied, he goes on using it. The item leaves his pack, but as far as the fight is concerned it is still in his hand, until he readies something at least as good. Nothing on his character sheet shows it, because the sheet lists only what he carries.

Why: the four readied items are kept apart from the nine carried ones. After a trade the game runs its re-equip routine on both members, and that routine only ever puts things on: it readies a carried item when it is at least as good as the readied one and the hero meets the item's requirement (for a bow, enough Strength plus Dexterity). It never takes anything off, and never checks that the readied item is still carried. A fight works out each hero's blows from the readied items alone, so the missing item hits as if it were there. It was seen in the running 1987 game: a Fighter traded his readied Mace to the Thief, and the Mace left his pack but stayed his readied weapon.

Taken further, one item can serve a whole party: pass the best bow in the game from hero to hero, and every one strong and nimble enough for it ends up shooting with it, though only the last carries it. That story is in One God Bow for the whole party. The other way to share items, Distribute, never leaves such a ghost: it takes every readied item off before it gathers the packs, so it also clears any ghost a trade left.

The fix makes the giver stop using what he no longer carries: before the game re-equips him, his readied slot is cleared if he has no other copy of the item, and the re-equip then readies the best of what he still carries and can use. The taker is re-equipped as in 1987. It is not among my fixes; a player can turn it on with Choose my own.

Buy a new sword, shield, suit of armour or bow in a town's armoury and you would expect the hero to start using it. In the 1987 game he does not. The purchase goes into his pack, but his readied weapon, shield, armour and bow stay what they were. He walks out still wearing his old Robes with the new Leather in his bag, and fights that way, until something else makes the game re-equip him: a trade, or the Distribute screen. A player who does not check the character sheet may never know the money bought nothing yet.

Why: the armoury does call the game's re-equip routine, the same one a trade uses, which readies the best item a hero carries and meets the requirement for. So the intent to ready a purchase is there in the code. But the armoury works on a copy of the hero's record while you shop, and as you leave it copies that working record back over his real one, undoing what the re-equip had just done. It was seen in the running 1987 game: a Fighter wearing Robes bought Leather in Pendragon, it went into his pack, and his readied armour stayed the Robes until a later trade re-equipped him.

The fix keeps the re-equip the game already does, and runs it on the record that holds the new items, when the hero finishes shopping. A purchase is readied if it is the best he carries and can use; something whose requirement he does not meet, or that is no better than what he has, stays in the pack as before. Because the re-equip runs as he leaves, leaving the armoury readies his best gear even on a visit when he bought nothing. Prices, stock and everything else about the armoury are as in 1987.

This is one of my fixes.

A hero can carry nine items. Try to buy a tenth in the armoury and the 1987 game refuses with "He cannot carry anymore!", shown for a moment before the party's gold line is drawn over it.

It is a small thing, but it is the only line in the game about a party member that names a sex. A hero's record holds no sex at all, so the game cannot know whether "he" is right. The same refusal in Trade Item, when the hero receiving an item already has nine, reads "This character can't carry anymore."

The fix uses that trade line, the game's own words, in the armoury too. The line is longer than the armoury's box, so my recreation shows it across two rows where the gold line stands and then puts the gold line back. Only the words change: the nine item limit and everything else about buying are as in 1987.

This is one of my fixes.

When a hero has gained enough experience, he can go up a level at the town's training hall (see training). Along with the new level he gets three skill picks. Each pick shows every skill with a new value on offer, and the skill you pick takes that value. In the 1987 game the offer can be lower than the skill the hero already has, and taking it lowers the skill. A hero with an Attack of 250 can see it fall to 199 on one pick and 173 on the next. Worse, a pick cannot be declined: Cancel simply takes whichever skill is highlighted.

Why: the offer is worked out from three things: the skill as it stands now, a number in the training table for the hero's class and level, and a part that comes from his attributes. Roughly, it lands halfway between the skill now and the table's number, raised or lowered by the attribute part. For most skills and most levels that is a rise. But a skill already well above what the table and attributes support is pulled down, and a few entries in the tables are oddly low. The Ranger's Attack entry for level 15 and up, for instance, drags a Ranger's Attack down with every pick (see Heroes do not get worse with age). It was seen in the running 1987 game: a Fighter with Attack 250 trained, the Cancel key took the highlighted Attack row twice, and his Attack went from 250 to 199 to 173.

The fix makes a pick keep the higher of the offer and the skill the hero already has, so training never lowers a skill. The offers shown, the tables, the three picks and Cancel taking the highlighted row are all as in 1987: a pick that would have lowered a skill now simply leaves it where it was.

This is one of my fixes.

To go up a level, a hero needs enough experience and the gold to pay for the training. The experience needed climbs steeply: a hero of average aptitude needs about 12 million points to get from level 16 to 17. In the 1987 game, once a hero reaches level 17 that stops. At 17, 18, 19 and 20 he can train with no experience at all, paying only the price, at most 15,000 gold a level. A level 16 hero with no experience, by contrast, is told quite correctly how many millions of points he still needs.

Why: the experience needed is worked out by multiplying a number from the level table by 1000 and dividing by the hero's aptitude. From level 17 that multiplication passes about 2.1 billion, the largest number the program's 32 bits can hold, so the result wraps round and comes out negative. Any experience is more than a negative amount, so the game decides the hero has enough. It was seen in the running 1987 game: a Wizard of level 17 with no experience was offered level 18 for 14,512 gold, trained, and was offered training again at 18, while a level 16 Monk in the same party was told he needed 13,492,388 more points.

The switch makes every level need its experience: the number is worked out whole, with no wrap. A level 17 hero of aptitude 100 then needs about 21.9 million points for the next level, and every level after that needs nearly twice as many as the one before. The price in gold, the new hit points and the skill picks are as in 1987.

This sits with the enhancements, not the fixes, and it is off unless you choose it, so With Justin's fixes plays levels 17 to 20 as in 1987. The 1987 rule does exactly what its arithmetic says, and turning the switch on changes how the end of the game plays: the last levels go from a matter of gold to about 21.9 million points for level 18 and over 127 million for level 21, at aptitude 100.

This is not a fix: nothing is wrong with the 1987 song. It is an enhancement, something the 1987 game never had, and it is off unless you choose it.

When you win, the 1987 game plays Beethoven's Ode to Joy, the melody note for note. It plays in three thin voices, flat and MIDI-like, and it uses none of the tricks of the Amiga's Paula sound chip that the opening theme does. The game's music is not a recording: it is a short file of notes and timings that the game's own sound driver plays, the same driver that plays every sound in the game.

With the enhancement, the ending plays my own arrangement of the Ode in its place, written in exactly the same file format and played by the same 1987 driver, unchanged. It is made to sound like an Amiga: a short fanfare to open, the tune on one side of the stereo with a soft pad and a bass on the other, then the tune passed across an octave up with a second line beneath it, vibrato on the lead, and a broadening finish on a held chord.

Only the song changes. The ending's pictures, words and timing are the same either way, and it is not part of With Justin's fixes: a player turns it on with Choose my own.

This is not a fix: the 1987 inn works as it was written. It is an enhancement, something the 1987 game never had, and it is off unless you choose it.

A night at a town inn heals the party: hit points and magic points are filled up, wounds mend, and a caster may gain a little more magic. It also costs time. Every living member grows one fortnight older (see ageing). In the 1987 game the inn always keeps the party for that stay, even when nobody is hurt, tired or short of magic, so everyone ages for nothing.

With the enhancement, the town inn first checks whether the night could change anything. If nobody would gain hit points or magic points, no wound would mend, and no caster has a chance of more magic, the party leaves without staying: nobody ages, and the report just says "All rested." If anything at all could change, the stay happens exactly as in 1987, ageing included. A caster's chance to gain magic at the inn comes back with every battle the party fights, so after a fight a caster who can still grow makes the stay happen.

Only the town inns change. The roadside inns are as in 1987. Age changes nothing in play (see Heroes do not get worse with age), so the difference a player sees is the age on the character sheet. It is not part of With Justin's fixes: a player turns it on with Choose my own.