Introduction
The process of tuning a game is a tricky one. If you try to do this on your own you will inevitably tune the game to your individual preferences. Also, your increasing experience of playing the game will make you think that the game is easier than it is. It's essential to get other people playing too, but not too much, or they'll get used to it as well.
Tuning can be done at the same time as play-testing and bug-hunting. During that time, anyone play-testing is going to be accumulating knowledge about the game, and formulating ways to play it.
I "borrowed" some screenshots from the interweb for this article for illustration purposes. Hope no-one minds that I used their screenshot of my game, and thanks.
I "borrowed" some screenshots from the interweb for this article for illustration purposes. Hope no-one minds that I used their screenshot of my game, and thanks.
Progression
What sort of game progression do you want, or need? Do you want a game that gets harder as you go along? Do you want the game to be fairly linear?
There are a number of ways that a game can progress. Typically the early arcade shoot-em-up games just repeated the same thing but maybe the Invaders were closer, the Galaxians a little faster, or more trigger-happy. That progression is quite easy to build into algorithms. The start positions and numbers of enemy can be used, and the chance of firing calculation can include the level number.
A puzzle game would start easy and then introduce more devious puzzles, better hidden items, more stages to solve the puzzle, that kind of thing. Again, size matters, the game area could be increased, for example.
Story-type games don't necessarily need to get harder; the complexity of figuring out the plot will keep it fresh. These games must be tough to manage, there are so many ways that the players could fail to engage the plot. I played "Dear Esther" the other day. It felt fairly linear, and I was just admiring the scenery. It's a nice 3D engine and it leads you through the story pretty well. The only time I got stuck was when I was trying to walk round the underground waterfall pit and kept slipping down. Fortunately there's youtube!
Same or Similar?
Most arcade games appear to play very similarly every time you play. Some are identical: we used to count shots in Space Invaders and we knew that on the 23rd shot you had to shoot the flying saucer for 300 points, and every 14th shot after that would always yield 300 points - it's not really random at all, but if you don't know how it's doing it then it'll look random.
Personally, I always favoured the idea that the game would be different every time, but in a controlled way. For example: in Gribbly's Day Out the Gribblets all start in their designated positions but can then choose to jump around randomly. By the time the player has found them, some could be a fair distance from their start positions. The Paradroid ships are populated with each robot randomly chosen from 4 different consecutive types, some of which are also increased by the ship number. In Uridium: some formations are generated behind the player, so it depends which way you're flying at the time, and Uridimines might appear any time. For me that creates a better long-term game because you don't know exactly what's coming. We let the SID chip on the C64 generate the random numbers, so they were not reproducible. Anyway, my point is that if the game setup is different every time then your playability rating per level is going to be a little more fuzzy and should be noted as such. It doesn't matter either if occasionally a more difficult level sneaks in early.
You could argue then that my games might be unsuitable for competitions as you couldn't guarantee the difficulty, but overall the randomness balances out - the house always wins in the end!
In this internet age if you wanted to synchronise two or more remotely connected computers you would use seeded random numbers to ensure that they all behaved the same. We therefore use both seeded and unseeded random numbers nowadays. You can have an array of unseeded numbers for things that don't matter, such as explosion fragments that don't interact with anything, and use seeded numbers for generating game elements that need to be the same on all machines. That doesn't prevent you from sending the random number seed from one machine to the other to get a different synchronised game every time.
Chasing Perfection
It goes without saying that a game must be able to be played all the way through. There must not be any bugs or mapping mistakes that would cause the game to not be completable (my new word, Susie Dent, please get it added into the OED). There was no online update capability back then, everything had to be working before the first run of tapes was made.
I also want to make this very clear: updates inconvenience your users. Sure, it's great to get those bugs fixed, but there shouldn't be any, and it takes a lot of time and effort on the part of every user to keep their software up to date. I get 1.2Mb/s here. VS2017 with all the trimmings took almost a day to download, on each computer.
8-Bit Days
Once I left my IT job to write games I kept in touch with my friends there and used to visit them most weekends with a work-in-progress version of the game I was writing at the time. It is invaluable to seek other people's overall opinions.
I've been looking at a few youtube videos of some of my old games being played, or reviewed. Clearly many of these reviewers did not have the original instructions manuals that came with the games. This leads them to get rather frustrated as they don't know what to do or how to do it. The manuals are important, and worthy of their own blog page.
Arcade games need to attract people to play them, so they can have quite detailed demo or instruction sequences, but generally they also try to keep things fairly simple as they are only trying to provide a 5 or 10 minute player experience.
Computer games can be more involved, typically an 8-bit game might take one or two hours to go through all the game, once you've learned to play it, and studied the manual for the rules of the game and any hints we might have included in the text. I always used to write the manuals, including the back-story, for my games. Steve Turner did likewise.
While watching a chap play Gribbly's Day Out on youtube, I realised that he didn't have the manual and despite the instruction to "Rescue the Gribblets" as he started, he didn't know what a Gribblet was. He started to fly around and mess with the power grid, which is advanced stuff that you don't need to do much at the beginning, and ignored the jumping creatures that he was meant to rescue. If I had put a picture of a Gribblet next to the message he'd at least have known what he was looking for. On the other hand, should I have any sympathy for someone who hasn't got the manual? I'm all in favour of good official documentation. Nowadays there's a lot of unofficial documentation online and it's just spread all over the place.
When I watched another chap playing Paradroid, he hadn't realised that if you ram another robot, or roar into it guns-ablazing, you have to back off when it explodes. Explosions can damage the smaller robots. There are sound effects that allude to the damage, and if you watch the rotation speed of the player icon you can see that you are taking damage. He also didn't know what the re-energiser blocks were, so didn't know he could recover his energy.
How much information do you have to give the new player? This depends on whether they can blunder about in the game for a while and figure it out for themselves. Again, the arcade machine doesn't mind collecting a few swift 10 pences while we get used to the game, as long as we do. I don't spend a couple of days on the instruction manuals and stories so that they are ignored, or Heaven forbid, not paid for.
How much information do you have to give the new player? This depends on whether they can blunder about in the game for a while and figure it out for themselves. Again, the arcade machine doesn't mind collecting a few swift 10 pences while we get used to the game, as long as we do. I don't spend a couple of days on the instruction manuals and stories so that they are ignored, or Heaven forbid, not paid for.
My 8-bit Testing
My first 4 games were conversions of Steve Turner's Spectrum games: 3D Space Wars, Seiddab Attack and Lunattack (twice). For the most part these games were just plotting gaphics as fast as they could, I don't remember any speed regulation in the main game loop at all, and you can feel the control mode struggle when there are a lot of big ships on screen in 3D Space Wars. The other two titles had more graphics on screen for the background all of the time so the main loop was better regulated. It'll be interesting to see what they do on the super-speed Spectrum and C64.
I did put a speed control on my Dragon 32 3D Space Wars to allow the player a bit of control, and the controller was a proportional one, though not the keyboard.
A programming tip then, rather than make your control modes constants in the code, put the values into variables so that you can adjust them on-the-fly. That way you can tune up the control mode in real time.
Steve and I used to test each other's games too. I spent a fair few hours playing Avalon, DragonTorc, and AstroClone on the Spectrum. After each session we could give feedback on whether we though anything was too tough or too easy. Once you've built up some experience on your own game you know every little detail, other people don't know the underlying algorithms so they know less than you do. For example, I knew that the Paradroid 90 sentries would fire a few rounds off and then pause, as if to reload, so I could wait until they fired a salvo, then rush them. Other people might not figure that out. That's why it's really important to watch how other people play, and ask them what they're thinking.
When you let a publisher play your game then, it's best to make sure you're there to show them how to play first. Of course they'll have difficulty playing it, and they'll have more trouble accepting it if it's not exactly like the last block-buster. Innovation can be a dirty word to them. I also like to get most of the work done so that they only see the near-finished product. If you show something too early then the full impact of everything you've done is lost, and you'll be tweaking stuff unnecessarily for them for ages.
I'm not convinced of releasing early versions of games on Steam or other platforms either. Yes it's nice to get some feedback but again it's like knowing what your birthday present is going to be in advance, and being able to play with it before it's finished being built.
My preference is to complete the coding and provide a demo level, like we did on magazine disks. They were always produced last minute as a cut-down of the finished game. We would remove the end sequence and replace it with a sales message. Just try and put some of your best graphics in the demo.
Gribbly's Day Out
Gribbly's Day Out had 16 levels. I didn't want them to just play through one after the other, but I did want an easy trainer level to start with, and a monster last level to end. In between I arranged the other 14 levels roughly in difficulty sequence, but... then I messed with the next level selection algorithm so that the more Gribblets you rescue, the more levels it can choose from for the next one.
You have to rescue 6 Gribblets or more anyway to progress. If, on level 1 you rescue 6 Gribblets it'll just open up 1 more level and you'll get level 2 next. If you rescue 7 Gribblets then it'll open up 2 more levels. I can't remember if it tries to give you the highest level you haven't been to next, or picks one at random that you haven't done yet. The philosophy was that the better you do, the quicker you get to the higher levels. It always saves the 16th nasty level for last though.
When I was setting up the Gribbly's map background layouts I did start to create some more deviously small gaps to fit Gribbly through. I also set up a couple of levels of tunnels with multiple small gaps. You don't want to irritate the player by then putting all 8 Gribblets at the other end of the tunnel, that's just torture, so don't be too mean. Put some easy ones to rescue too. That gives the player the dilemma of rescuing the easy ones first, or the difficult ones.
I used to practise the new levels a fair bit. If there was a gap that was too tight a squeeze and I could only get through every 4th attempt, then it needed to be eased off. You can't spend too long testing all the gaps because you'll get used to them. That's where my trusty test team came in. Once I was fairly happy with a map I could take it to them and let them try it out. If they complained too much then I'd make a note and change it for next time.
The only other parameters that changed per level of Gribbly's were the homing instincts of Seon, the ultimate meanie, he got more feisty, and the life-cycle of the other meanies speeded up, so they didn't spend as long in each form. The rest of the difficulty was in the map designs.
Paradroid
The Paradroid deck maps can be accessed more or less as the player chooses, so the sequence became unimportant. I always start the player on an easier deck. The difficulty of the game then came from the size of the decks and which robots appear on them. I tended to make the cargo decks generally contain a lot of battle droids, and the 999 cyborg was always on the bridge with some big side-kicks. The decks all have names and some hint more than others at what sort of opposition to expect. Engineering has maintenance droids, the crew decks have servant droids.
Since the player can become any droid in the game, the game is balanced, or unbalanced by the player. I just had to make sure that the little droids couldn't damage the really big ones. So that was a new feature for me, bullets that don't hurt the bad guys at all.
There's no concept of spare lives in Paradroid. If you've taken over a bigger robot then I needed to make sure it could withstand a couple of hits from the big weapons, and if it gets shot from under you then you should run away fast. Since the game applies the same rules to the player as the other robots, then again it self-balances.
Uridium
For Uridium I wanted pure arcade. Having got 50 frames per second and a nice multi-colour metallic look, I went for a straightforward linear progression. There were some walls on the first deck to crash into to teach the player "the rules", and most dreadnaughts are as long as they can be.
The meanie fire decision routine included the dreadnaught number, and the number of times the last dreadnaught had been reached. It hashes a number together, then gets a random number and if that's bigger, and there's a sprite available, the meanie will fire. I also had different wave formations that I could generate, each with up to 5 ships. I decided to create smarter singleton waves for later on where 1 ship tries to line up with the player, and I had a fire instruction in the algorithm so it could "force fire", and being a singleton there were plenty of sprites available as only 1 wave was active at a time.
Similarly the Uridimines linearly progress in their chance of production, and I could add more or less Uridimine ports on the levels easily enough, you could really put as many as you want wherever you want.
The data for the backgrounds was kept with the formation data for that level. That way I could shuffle them about in the source editor quite easily and arrange them in what I thought to be order of difficulty. It also allowed me to slot in another easy level at the beginning just to give players a warm feeling that they can progress.
Finally, since the Manta control allowed it to fly on its side then I could start to utilise some really narrow gaps, and not just on their own. It became like Gribbly's. While I was creating the background map data for a level it would be the first ship and I could have as many plays at it as I wanted, and without any formations attacking. That gave me time to see that the Manta could get from one end to the other. I had become a competent flyer and needed to back some bits off a bit, but by the time the player gets to level 15 they've probably played on the previous levels for a fair amount of time, so I felt if I could do the level, and I'm not that great a player, then plenty of others will be OK.
I did 16 more levels for Gribbly's 2nd Day Out, and Uridium Plus, and I tried to make the later levels a tad harder again than in the original. For the record, I clocked Uridium twice to reach Iron 3 times before being beaten, but only clocked Uridium Plus once, having copped it on level 10 or so for the second time. My top score on Uridium was 537,035.
Alleykat
For Alleykat, I wanted the player to be able to choose the races that they wanted to compete in. That way, if there turned out to be a type of race they couldn't get on with, they could avoid it altogether. So in some ways the races pan out like the Gribbly's levels, but you don't have to play them all, though you can. You can only enter races you can afford, so since the entry fees ramp up, along with the prize money as the races progress then you can't go straight to the difficult stuff, and you need to complete the races.
The different background types, of which there were 8 graphics sets, each had a chance threshold that a player bullet would destroy it. For the weaker types on the early levels it's a given that one shot from the right height will destroy the background piece. I then rigged some tougher types that they'd maybe have a 40% chance of going, so you might need 3 shots on average, but it could go in 1, or it might take 7. That meant that I didn't have to store how many shots each background piece had taken. You could also say it was a bit more realistic.
I tried to set the values on most backgrounds so that they would likely blow up before you got there (which would take them out for sure, while damaging the player). The Skullnia backgrounds are impervious to bullets though. That meant that I could create tracks that had to be negotiated rather than destroyed, at least for the first lap. As the race goes on the other ships start taking out the background pieces for you.
The actual backgrounds are algorithmically generated. I could set the chance of a blank piece and the chance of a high or low piece of background at each hurdle. I also thought I could set the distance between the hurdles too. The algorithm also has to ensure that at least one passable piece of background is in each hurdle, For the last race I also made sure there was a nice big impervious-to-bullets skull that you have to fly into the mouth of right in front of the player at the start.
There are also energy pellets on the ground in the races, again algorithmically generated by random chance so I could make plenty or few in each race Those of you with the manual might also know that finishing a race with full or nearly full energy gives you a bigger energy tank for the next race meaning you can take more damage. That gives the player more goals to aim for. Topping up your energy as the race season progresses gives you a stronger ship by the end. With more sprites I might have been able to show that on the screen. If only I'd had the sprite multi-plexor then.
Morpheus
The Morpheus tuning up was completely different. Previous games had been using a lot of the RAM space for game maps, and although Alleykat generated its maps, I had 8 2K character sets in there to take the place of the maps. For Morpheus I wanted lots of sprites. The attack waves are divided into 50 levels, so I had to initially invent 50 sets of attack waves and then try to get them into some semblance of difficulty order. I had also built into the game a very flexible ship feature system so the player could be using a well-armed and well shielded ship, or they might have not developed it, or they might have unbalanced it with bad choices. Success in the game requires the player to develop the ship with good choices of equipment, and keep upgrading as better kit becomes available. Probably I wouldn't go for the giant hull, that was to demonstrate that you can go too far.
Morpheus does end, but I can't remember how. There are 50 levels, and although I tested each of them individually and could beat them all, I am ashamed to say that I was only good enough to get to level 44 of the finished article in my best sitting. I was OK with that because by then we had some better games players than I in Graftgold, so I figured someone would beat it. Steve used to get letters from players that had beaten Avalon or Dragontorc, usually within a week of release. We didn't think that would be possible.
Intensity
Each level in Intensity is a discrete battle. There are almost 80 levels in the game, arranged in a 5 high x 16 wide rectangle. You start in the middle on the left and have to get to the right-hand end. The 5 horizontal strands represent different levels of difficulty, the top alpha route being the hardest. Some levels allow you to progress to the right, some only allow up or down, so the route is not necessarily simple to get to the end. You will have to play more than 15 levels for sure. The rewards are higher on the alpha strand too, but you need the better ships that can fly over higher obstacles.
When I was designing levels I was trying to make more difficult backgrounds all the time. I was able to design new backgrounds always as the first level before moving the data into it's designated place. I had to decide where the levels would go. You can firstly design easy or difficult features, but then the level takes into account where it is in the map for the meanie generation and transformation times. Also, the background layout plays a part as the meanies can get on the decks more frequently if there are bigger gaps to get through. Bigger decks mean they splat on the underside, like flies on the windscreen.
By placing the crew exit hatches carefully you can force the player to keep moving the drone or face the consequences, or keep the exits close for an easier level. You need to know what elements to alter to make puzzles or add to the difficulty.
Our 16-bit Testing
Our forays into Atari ST and Commodore Amiga territory required us to form teams. This usually meant the main programmer, plus 2 or 3 graphics artists and nearer then end, the sound and music expert. There was also Dominic Robinson with system support for the game OS he had written, allowing us to write in a pre-emptive multi-tasking environment that was lean for games.
I'll miss out Rainbow Islands from here because we were not tuning the game up other than to ensure that we had replicated the game as closely as possible. We had a lot of documentation from Taito, which gave very detailed speed and algorithm information for the player controls and the meanies. The rainbows on their own deserve a blog page.
I'll miss out Rainbow Islands from here because we were not tuning the game up other than to ensure that we had replicated the game as closely as possible. We had a lot of documentation from Taito, which gave very detailed speed and algorithm information for the player controls and the meanies. The rainbows on their own deserve a blog page.
Paradroid 90
For Paradroid 90 then I would generally design the ship layout, since it all had to fit together with no overlapping lifts on the side view, and the connecting data had to be fed into the assembler and debugged. We designed the levels on an editor written in STOS for the Atari ST. I had the graphics artists design some of the background characters to create features, and decorate the backgrounds. I did some of the background pieces myself just because sometimes it's easier to get what I wanted.
Having a team then means that you can let everyone have a go at playing the game, whilst reminding them that they're paid to do a job, and not just play. They have to know when to stop. Actually, since for most of the development life of a game it would only run on the programmer's kit anyway, we wouldn't make up playable disks until near the end of the project. Game tests generally consist of one group of levels assembled in rather than loaded. Therefore, to play the game they'd need to play on the programmer's kit.
Paradroid 90 deck designs followed the same idea as on the C64: each deck had a purpose and was accordingly populated. The cargo decks could be bigger with even more robots on them. I cut down on the types of robots because we had to do over animated graphics in 8 directions, and that was eating up a lot of memory. I then diversified what I could do with them. I could have patrolling sentries on specific short routes, and I could have static sentries that would guard locations, making them harder to ambush.
As each ship was taking shape we could test them as the first ship and prove that we could play them all the way through. The overall game design still balanced itself nicely. I made the transfer game AI more intelligent to make it more challenging for the player. The bigger decks made for better encounter areas so I was quite free and easy with big mean robot distribution.
Whilst I could get to the standard end of the game in one (long) sitting, I failed to beat the secret extra pirate ship, though I had beaten it standalone. Again, I was happy that someone somewhere would beat it. Anyone?
Fire and Ice
Fire and Ice was my first original 16-bit title, I suppose you might say. I had less steer on this new design as platform games aren't my forte. It's nice to change genres though to keep things fresh. Phillip Williams was the lead artist and took a great interest in the design of the sprites, the backgrounds and the puzzles, as did Jason Page. They were happily throwing ideas into the mix.
Since the arrangement of the lands had to go from icy cold to desert hot, we had little flexibility to swap things around to better serve the difficulty. We chose the lands based on themes and put them in latitudinal sequence. I had made the mistake of putting the slippery levels first. Seems we put penguins in the North Pole too, sorry penguins!
Each land had its own Coyote movement constants to decide accelerations, top speeds, friction and drag. Apart from the ice and undersea sections the values were broadly the same. If I was doing a sequel I might try Japan, Peru, Australia and Antarctica, with polar bears. I'd like to represent Wales somehow too.
Each land had its own Coyote movement constants to decide accelerations, top speeds, friction and drag. Apart from the ice and undersea sections the values were broadly the same. If I was doing a sequel I might try Japan, Peru, Australia and Antarctica, with polar bears. I'd like to represent Wales somehow too.
When we went to 16x16 pixel graphics we needed a new map editor. Steve wrote one on the PC. The graphics were done in DPaint, also on the PC. Phillip did most of the original graphics and I programmed all of the meanies. I do remember that we weren't happy with the Coyote's speed so I went to London to meet up with Jaz Rignall to see what he thought of the demo. The result of that discussion was that I speeded up the Coyote. This changed the jumping speed, gravity and the sideways accelerations and speeds, basically they're all related. After we were happy with that I tried to play one of the first developed levels and realised that the jumping meanies were affected by the gravity change and I had to recalibrate them. The next related area was all of the ice weaponry, all of that needed adjusting too. Best to get your physics constants all sorted out early!
Fortunately making the Coyote quicker didn't mess up the jumping gaps as his overall jumping range was kept similar, otherwise we might have had lots of map adjustments to make as well. Fire and Ice doesn't require the pixel-accuracy of, say Manic Miner, otherwise we'd have had to be even more careful. The Coyote top speeds were all in whole pixels, the accelerations had accuracy down to 65536 of a pixel, as did the co-ordinates. The plot routines clipped off the sub-pixel positions.
The majority of tuning adjustments came from adding more meanies until there felt like too many, then taking one or two out again. Meanie start positions were entered into my assembler code, not the map editor, so I kept a tight rein on those. Similarly the strength of the big bosses needs to be regulated carefully. I hate meeting seemingly invincible big bosses, especially if you can't see you're making progress against them. I do like the Flying Shark boss tanks that start flaming so you can at least see you've damaged them.
We did end up making the first icy lands as easy as possible, most meanies don't fire back, just because the control mode is a bit of a handful on ice, and I was not trying to create a shoot-em-up. The team's input on this game was vital for more variety. Phillip and Jason especially were coming up with plenty of ideas.
Uridium 2
This one was the toughest of them all.
We started off with a dual-playfield version. This gave us 7 colours for the front layer, and 3 for the back lower layer, which scrolled in parallax. Phillip Williams created a prototype ship which looked nice, though with the limited colour choices it looked a big cartoon-like. He did get the scale nicely. We then realised that most of the moving objects would also have to come out of the same 7 colours, and that made them blend in all too well. Reserving 1 or more colours for obstacles was also impractical, as was reserving some colours for the sprites.
The smooth parallax scrolling did look lovely, but we reluctantly decided to use the 5 bit-planes for 1 32-colour layer. Unfortunately Phillip left Graftgold shortly after this so he didn't get to do any 32-colour backgrounds.
I was having trouble getting nice graphics from the guys that I thought we should be capable of, given that this was our first 32-colour game. Since the background graphics define the whole game arena they've got to be good. Other companies were getting the hang of the extra colours.
They graphics also have to be functional. It has to be obvious what bits of the dreadnaughts not to crash into. This might be signalled by big shadows being cast, and also their own colours. It turns out that the colour choices I made on the C64 and the starkness of the graphics were what the game needs, and all the extra colour choices were muddying the waters.
I also wanted 6 different styles for the backgrounds, but we struggled to get even one. If the background obstacles are not obvious then the player will get frustrated crashing into them. We managed to enlist the services of Stephen Rushbrook, who was quite local.
The background graphics were made from 16x16 pixel characters, drawn with DPaint, and then fed into the PC map editor. The maps tend to have to be at least partially put together by the graphics artist because they have to design the blocks and make larger constructions. No-one else is going to be able to piece them all together, it's like a custom jigsaw puzzle where all the pieces are square. You only have the pixels to see how they fit together.
If you're not the artist then you might use two blocks then you might map pieces together that happen to work now, but if the artist then changes a couple of blocks then your incorrect use of them might break. This is why I was happiest on the C64 doing my own graphics, maps and programming!
Anyway, Stephen Rushbrook tuned in to what I needed very quickly. He started with a graphics set that ending up becoming the second set of 4 ships. It had very clear lines and bright almost glowing wall pieces. This set actually was intended to be the first set, but due to publisher demand for easier ships to play on, it got preceded.
This is where the playability gets tuned. You can't necessarily change the levels that you've done to make them easier, you might end up throwing away some of the drawn features that are a tad too difficult, but no-one likes throwing away good work, so you move it on up and try to then draw something that's easier to play from scratch. Unfortunately Renegade were having trouble playing the early levels so kept asking for it to be made easier to start with.
We ended up with a new first set of ships, which looked a bit smarter than the old first set, but with no walls at all on the first couple of levels, then we added some on the next. Since we kept the next set of 4, the first level of those, which was to be the easy opening level of the game, gets easier again. Somehow Uridium without walls is like a tank without any ammunition.
By the time we got to the fourth set of 4 ships, Mr. Rushbrook was getting into his stride and came up with a solid-looking set of graphics, and some evil laser turrets. At this point it was my turn to try and make the game easier. I was play-testing all the levels and with the horizontal and vertical scrolling, the real estate was getting large. A new problem was starting to occur, the levels were easy to get lost in. A smarter bunny than I might have put more than one landing strip on each level to balance that out, or helpfully marking the final landing strip(s) on the radar, but we started to think that navigation was going to become a bigger part of the game. So be flexible with the game design if opportunities for expansion present themselves.
If you have an end sequence, you tend to want to make the player really work for it, so the final level(s) tend to end up being a little harder than they ought to be. When the development time is up, you have to go with what you've got. All you an do is rearrange things a bit. You don't want a really tough bit in the middle and then it gets easier again, you'll always lose some people who give up.
I can't remember now how many ships we gave away on the magazine demo disks, but some people have reported that they couldn't manage those levels. They were only easy levels too. Unless your game adapts to how well the player is doing then there's no way to match up with everyone's abilities, and if you do try to adapt then you end up with players that know they shouldn't try too hard and the game will go on easy on them. Plus, no-one gets bragging rights. Wouldn't work at the Olympics, would it? "I decided not to run so quickly, so everyone slowed down for me and I got the gold."
The Final Words
Sorting out the playability and progression of a game is very important. Trust your instincts. As the game designer you've got an edge over most other players in learning to play the game first, and grading the difficulty is still the same comparing one level to another. Just make sure you can do all of the levels, preferably in a single session, but at least playing them one by one. If you lose 10 lives getting through one level then it's too hard! With a little practise you should be able to clear each level without too much hassle.
Let some other people have a play, watch them, help them, and listen to their feedback.
Let some other people have a play, watch them, help them, and listen to their feedback.
