In my last post I began looking at a subset of NES games as potential scope models for solo game devs, especially bedroom devs.
The idea was that modern tools might balance out a lack of budget.
In this post we'll break the monolithic data table down and see if we can draw some focused conclusions.
tl;dr 👈
To match Black Box Era games a developer should expect to:
- 💰 Move fast. Spend no more than one year on any one game.
- 🦠 Focus efforts. Promise either a great narrative or great mechanics, not both.
- 🖌 Put effort where it matters. Downplay environmental detail in exchange for detailed player characters and enemies. Emphasize sound effects rather than music.
- ⚖ Scope by Genre
- 📘 Narrative Games (including RPGs and MacVentures) are always big, with 2K to 10K words and more visual and audio content than other genres.
- 🛡 RPGs need at least 400 screens-worth of explorable space, at least 40 distinct enemy types, around 40 items/equipment, and 10 or fewer NPC types.
- 🏰 Metroidvanias need at least 500 screens-worth of explorable space and at least 14 well-defined mechanics, but don't require RPG elements.
- 🦘 Platformers (non-Metroidvania) need 200 screens-worth of playable space, generally across 12ish levels, with about 30 enemy types and around 16 collectables/power ups. Expect to write less than 500 words and to have few NPCs.
The Games In Question 🎮
As a reminder, here are the titles we're looking at:
| Game | Genre | Year | Studio |
|---|---|---|---|
| Elevator Action | Platformer | 1983 | Taito |
| Lode Runner | Platformer | 1983 | Doug Smith |
| The Tower of Druaga | Dungeon Crawl | 1984 | Namco |
| Urban Champion | Beat 'Em Up | 1984 | Nintendo |
| Hydlide | RPG | 1984 | T&E Soft |
| Castlequest | Metroidvania | 1985 | ASCII Corp |
| Ice Climber | Platformer | 1985 | Nintendo |
| Super Mario Bros. | Platformer | 1985 | Nintendo |
| Déjà Vu | MacVenture | 1985 | ICOM Simulations |
| Castlevania | Platformer | 1986 | Konami |
| Dragon Quest | RPG | 1986 | Chunsoft |
| Deadly Towers | Action RPG | 1986 | Lenar & Tamtex |
| The Legend of Zelda | Action RPG | 1986 | Nintendo |
| Metroid | Metroidvania | 1986 | Nintendo |
| Castlevania 2: Simon's Quest | Metroidvania | 1987 | Konami |
| Faxanadu | Metroidvania | 1987 | Hudson |
| Super Mario Bros. 2 USA | Platformer | 1987 | Nintendo |
| Shadowgate | MacVenture | 1987 | ICOM Simulations |
| Mega Man | Platformer | 1987 | Capcom |
| Zelda 2: The Adventure of Link | Metroidvania | 1987 | Nintendo |
Starting Small 🦐
I noted last time that the list of games was about evenly split between games that seemed to me large and ones that seemed small, with a few that felt in-between.
I first broke the games down this way, completely subjectively.
Here are the titles I'd call "Small" NES games:
| Game | Team Size | Dev Time | Time to Beat | Mechanics | Visual Assets | Songs | Sound Effects | Word Count |
|---|---|---|---|---|---|---|---|---|
| Elevator Action | 1 | 1 | 0.5 | 2 | 8 | 2 | 9 | Negligible |
| Lode Runner | 5 | 4 | 3 | 4 | 8 | 1 | 8 | Negligible |
| The Tower of Druaga | 4 | 1 | 2 | 6 | 8 | 5 | 10 | 500 |
| Urban Champion | 4 | 1 | 0.5 | 2 | 8 | 3 | 14 | Negligible |
| Hydlide | 5 | ? | 3 | 6 | 8 | 3 | 7 | Negligible |
| Castlequest | 4 | 0.5 | 2 | 6 | 32 | 5 | 5 | Negligible |
| Ice Climber | 7 | 1 | 2 | 5 | 8 | 3 | 17 | Negligible |
| Median | 4 | 1 | 2 | 5 | 8 | 3 | 9 | <500 |
| Mean | 4 | 1 | 2 | 4 | 11 | 3 | 10 | <500 |
| Mode | 4 | 1 | 2 | 6 | 8 | 3 | N/A | <500 |
Observations
- None of these games are later than 1985 and only one seems to use a mapper chip. So, small.
- Team Size, Dev Time, and Time to Beat are all very stable here.
- Almost all teams are 4 to 5 people.
- Elevator Action is the only smaller team; Ice Climber is the only larger team.
- As a historical note, Nintendo seems to often have had larger dev teams than other studios during the '80s.
- The one-year Dev Time trend is evident here.
- Time to Beat is lower than in the main table, typically 2 hours.
- Almost all teams are 4 to 5 people.
- Mechanics follow the overall trend of 6 or fewer.
- Soundtracks are very short, typically three songs, which is in keeping with the times beyond just the NES.
- Every game has a different number of sound effects. They're more variable than in the larger table, to the point that there is no mode!
- Very little text.
- The Tower of Druaga, with its atypically lengthy intro and outro monologues, is the only game here to have an appreciable amount of text.
This is the type of text that other games would have put only in their printed manuals. - I will also note, though, that while Urban Champion has little text the text it does have is prominently displayed, as every building you brawl in front of is labelled. It almost feels like a ZZT game!
- The Tower of Druaga, with its atypically lengthy intro and outro monologues, is the only game here to have an appreciable amount of text.

Medium Games ⚖
Here are the titles I'd call "Medium" NES games:
| Game | Team Size | Dev Time | Time to Beat | Mechanics | Visual Assets | Songs | Sound Effects | Word Count |
|---|---|---|---|---|---|---|---|---|
| Super Mario Bros. | 5 | 1 | 2 | 13 | 8 | 6 | 25 | Negligible |
| Castlevania | 10 | 1? | 3 | 9 | 64 | 16 | 37 | Negligible |
| Dragon Quest | 20 | 1 | 10 | 18 | 16 | 25 | 23 | 7,000 |
| Super Mario Bros. 2 USA | 20 | 1 | 5 | 19 | 12 | 14 | 19 | Negligible |
| Mega Man | 8 | 1? | 3 | 9 | 64 | 16 | 35 | Negligible |
| Median | 10 | 1 | 3 | 13 | 64 | 16 | 25 | <500 |
| Mean | 13 | 1 | 5 | 14 | 56 | 15 | 28 | 1,825 |
| Mode | 20 | 1 | 3 | 9 | 64 | 16 | 25 | <500 |
Observations
- Almost all of these are platformers.
- Team Size is surprisingly large here! Two games are at the max of 20 devs.
- Dev Time remains standard.
- Time to Beat is pretty typical, with the exception of Dragon Quest.
- Taken as a whole, Mechanics, Visual Assets, and Soundtracks are in the midrange: 13, 64, and 16 respectively.
- Sound effects are right in line with the full data table at 25.
- Word Count betrays the oddness of these all being platformers except for Dragon Quest. I don't think there's a lot we can draw from that.
Big Games 🗻
Here are the titles I'd call "Big" NES games:
| Game | Team Size | Dev Time | Time to Beat | Mechanics | Visual Assets | Songs | Sound Effects | Word Count |
|---|---|---|---|---|---|---|---|---|
| Déjà Vu | 11 | 1 | 3 | 8 | 256 | 7 | 25 | 11,000 |
| Deadly Towers | 4 | 5 | 4 | 7 | 8 | 25 | 11 | 500 |
| The Legend of Zelda | 8 | 1 | 8 | 17 | 8 | 12 | 25 | 500 |
| Metroid | 12 | 1? | 6 | 13 | 64 | 12 | 21 | Negligible |
| Castlevania 2 | 15 | 1 | 4 | 14 | 128 | 9 | 53 | 1,000 |
| Faxanadu | 8 | 1 | 6 | 14 | 128 | 16 | 25 | 2,500 |
| Shadowgate | 11 | 1 | 3 | 10 | 128 | 15 | 13 | 11,000 |
| The Adventure of Link | 11 | 1 | 11 | 15 | 128 | 21 | 23 | 700 |
| Median | 11 | 1 | 5 | 14 | 128 | 14 | 24 | 850 |
| Mean | 10 | 2 | 6 | 12 | 106 | 15 | 25 | 3,413 |
| Mode | 11 | 1 | 3 | 14 | 128 | N/A | 25 | 500; 11,000 |
Observations
- Almost all of these are from 1987. Three are from '85 or '86.
- All of these are "more complex" genres: MacVentures, Metroidvanias, Action RPGs.
- Team Size here is typically 11, which is indeed big compared to the main table.
- From my contracting experience, many current indie teams are about this size.
- Dev Time is still a year.
- Time to Beat begins to trend upwards: here the norm is 5 to 6 hours.
- Mechanics are larger here, too: 14 typically.
- Although, even here we still have two games at a sub-10 Mechanics count.
- Visual Assets are around 128, which is about as big as they come in the Black Box Era.
- Sountracks are typically around 15 songs, which is actually one song less than the Medium Games.
- Sound effects are 24 to 25 again, although we do have Castlevania 2 with more than twice that.
- Word Count is consistently 500 words or above, except for Metroid.
- In fact, the typical range here seems to be 850 to 3,500 per game.
- If we excluded Metroid, it would be 1000 to nearly 4000.
Size-Based Conclusions ◾◼⬛
I don't know if it's the size per se or the premises, but as a player I don't find any of the Small Games particularly compelling. Just speaking for myself, I might have a hard time staying focused working on one of them for more than a few months.
Beyond that:
- The biggest differentiator between Small Games and Medium-to-Big games appears to be Mechanics count, not content.
- Small games have strictly less than 7 Mechanics.
- Larger games typically have 13 to 20 Mechanics.
- The biggest differentiator between Big Games and Small-to-Medium games is Word Count.
- Smaller games have simple stories communicated in less than 500 words.
- Big games have from 500 to 10K words.
- Accordingly, if you really want a complex story you should know that you're making a bigger game.
- In general, the bigger the game, the bigger the team and the more assets.
- That's surely not surprising!
- What is surprising is that development time does not increase.
It seems clear that it was important for the business models of the day to not go beyond a year.
Now let's break down the data according to genre.
Narrative Games 📘
Let's start by examining games that double-down on their text, games with over 1000 words.
| Game | Team Size | Dev Time | Time to Beat | Mechanics | Visual Assets | Songs | Sound Effects | Word Count |
|---|---|---|---|---|---|---|---|---|
| Déjà Vu | 11 | 1 | 3 | 8 | 256 | 7 | 25 | 11,000 |
| Dragon Quest | 20 | 1 | 10 | 18 | 16 | 25 | 23 | 7,000 |
| Faxanadu | 8 | 1 | 6 | 14 | 128 | 16 | 25 | 2,500 |
| Shadowgate | 11 | 1 | 3 | 10 | 128 | 15 | 13 | 11,000 |
| Median | 11 | 1 | 5 | 12 | 128 | 16 | 24 | 9,000 |
| Mean | 13 | 1 | 6 | 13 | 132 | 16 | 22 | 7,875 |
| Mode | 11 | 1 | 3 | N/A | 128 | N/A | 25 | 11,000 |
Observations
- These games are RPGs, Metroidvanias, and MacVentures.
- These came out in '85 through '87.
- Team Size, Dev Time, Time to Beat, Mechanics, and Sound Effects are all typical.
- Visual Assets and Soundtracks both tend large.
- Word Count, by selection, tends very large, from Faxanadu's 2,500 to Shadowgate's 11,000.
- My main takeaway here is that story-focused games on the NES are also in general content-heavy. These are big games.
- Yet, they're made by typically-sized teams in a typical amount of time.
- Shadowgate and Déjà Vu may look like they have less content than the others due to their typically shorter Time to Beat, but I expect that playtime is actually quite variable. MacVenture gameplay is largely trial-and-error and not everyone will figure out what to do at the same rate. Also, the vast majority of content in these games is optional: you do not need to collect, examine, or interact with every available object.

RPGs 🛡
And now for Narrative Games' next of kin.
| Game | Team Size | Dev Time | Time to Beat | Mechanics | Visual Assets | Songs | Sound Effects | Word Count |
|---|---|---|---|---|---|---|---|---|
| Hydlide | 5 | ? | 3 | 6 | 8 | 3 | 7 | Negligible |
| Dragon Quest | 20 | 1 | 10 | 18 | 16 | 25 | 23 | 7,000 |
| Deadly Towers | 4 | 5 | 4 | 7 | 8 | 25 | 11 | 500 |
| The Legend of Zelda | 8 | 1 | 8 | 17 | 8 | 12 | 25 | 500 |
| Castlevania 2 | 15 | 1 | 4 | 14 | 128 | 9 | 53 | 1,000 |
| Faxanadu | 8 | 1 | 6 | 14 | 128 | 16 | 25 | 2,500 |
| The Adventure of Link | 11 | 1 | 11 | 15 | 128 | 21 | 23 | 700 |
| Median | 8 | 1 | 6 | 14 | 16 | 16 | 23 | 700 |
| Mean | 10 | 2 | 7 | 13 | 64 | 16 | 24 | 1,757 |
| Mode | 8 | 1 | 4 | 14 | 8 | 25 | 23 | 500 |
Observations
- You'll note that there are some Metroidvanias in here, but I think all of these are arguably RPGs as well.
- These span the gamut of my Small-Medium-Big sets.
- Most of the few games that shipped with more than just a printed manual are in this set. This is a trend that I believe continues into later NES Eras as well.
- Contrary to RPGs in later time periods, these are not all Narrative Games.
- Hydlide has about as much narrative as Super Mario Bros.
- Team Size and Dev Time are a bit large at 8 to 10.
- Time to Beat is definitely high at 6 to 7 hours.
- I believe this shows that even back in the day, RPGs tended to have more content... and grinding.
- Mechanics counts are high at around 14, even including the two early, low-mechanic games Hydlide and Deadly Towers.
- Later games here have lots of Visual Assets while earlier games do not, leading to a mean right in line with the full table: 64.
- Soundtracks tend long with 16 songs, some even breaking 20.
- This is certainly something we would see more of in the 16-bit days when RPGs placed a lot of emphasis on their soundtrack.
- Sound Effects are standard with 23.
- Word Count of 700 to 1,700 is unsurprisingly well above average. Yet, it remains far below the Narrative Games as a result of several story-light RPGs.
- Taken together, it's clear that it is possible to make a small NES RPG, but even in the Black Box Era most RPGs are big games.
- It's also worth noting that the smallest Black Box RPGs we've looked at, Deadly Towers and Hydlide, were quite unpopular in North America.
I was able to compute more detailed data for some of these titles. Let's take a closer look, beginning with graphics.
👾 Note that the tiles we are counting here are 8x8 pixel hardware tiles, not the 16x16 metatiles that most games used as their fundamental unit of mechanical interaction.
| Game | Sprite Tiles | Background Tiles |
|---|---|---|
| Dragon Quest | 108 | ~1348⸸ |
| The Legend of Zelda | ~227 | 156 |
| Castlevania 2 | ~240 | ~180 |
| Faxanadu | ~3,560 | ~1024 |
| The Adventure of Link | >128 | >128 |
| Median | 227 | 180 |
| Mean | 853 | 567 |
Warning ⚠
- I've done my best to make an accurate tally but all of these numbers are rough and, except for the count of Sprite Tiles in Dragon Quest and the Background Tiles in Zelda 1, they're likely incorrect. However, most of them should be in the right ballpark.
- ⸸One that might be wildly wrong is the very high tile count for Dragon Quest's Background Tiles. The vast majority of these come from the large, colourful enemy graphics. The game uses several tricks to reduce unique tiles and it's possible I've missed some of those tricks.
- Even though Faxanadu's Sprite Tile count is also very large, I'm confident that it's close to correct.
- I wasn't able to get a definitive count of tiles for The Adventure of Link. All I could confidently say is that both are over 128 and under 4,096. 😅
Graphical Observations
- One thing not captured by the numbers alone: with the exception of The Adventure of Link, all of these games show a similar graphical priority. Environmental detail is downplayed in exchange for character detail, especially player characters and the enemies the player faces. Dragon Quest in particular is a great example of this:
- First, unlike later NES Dragon Quest games, enemies of the same type in Dragon Quest vary by more than palette, with visibly different equipment.
- Second, 48 of the 108 Sprite Tiles depict variations in the player character.
By contrast, the next most complex character graphics have only 10 (Royal Soldiers) or 8 (Princess, Dragonlord) Sprite Tiles.
- In keeping with this, more memory is spent on Sprites than on Backgrounds, often at a proportion of nearly 1.5:1.
- Although Dragon Quest bucks this trend, that's because it uses Background Tiles to represent enemies in combat.
- Faxandu is way more graphically detailed than any of the other games.
- Honestly, I never noticed this as a player, although its distinctive sepia aesthetic always stood out. Aparently others did, however.
- The numbers of tiles these games use is highly variable, anywhere from 256 to 4,500.
![]()
An Aside: The Problem with Using CHR-ROM to Estimate Asset Creation
This table underlines the limitations of my estimating graphical detail by the size of cartridge chips.
Faxanadu and Castlevania 2, two games with reasonably accurate tile counts, have exactly the same CHR-ROM available; yet, Faxanadu has an order of magnitude more tiles than Castlevania 2.
Related, a single 8KiB CHR-ROM chip, such as that used by Hydlide, can hold up 512 tiles assuming the space is devoted entirely to tile data. But Hydlide does not appear to have anywhere near this number of tiles.
Now let's look at some mechanical details.
| Game | Areas | Screens per Area | Total Screens | NPC Types | Enemy Types | Items, Etc. |
|---|---|---|---|---|---|---|
| Dragon Quest | 13 | 4 | 130 | 9 | 40 | 52 |
| The Legend of Zelda | 10 | 23 | 522 | 6 | 34 | 34 |
| Castlevania 2 | 6 | 28 | 334 | 7 | 32 | 28 |
| Faxanadu | 33 | 16 | ~540 | 25 | 38 | 37 |
| The Adventure of Link | >17 | ≠ | 489 | 12 | 50 | 25 |
| Median | 12 | 20 | 498 | 9 | 38 | 34 |
| Mean | 16 | 18 | 403 | 12 | 39 | 35 |
Notes
- Scenes per Area in Dragon Quest vary, but the majority have exactly 4.
- Most of Castlevania 2 takes place in a very large open world. There are also 5 distinct Mansions.
- The Adventure of Link has many areas; I counted only the most significant:
- 9 towns
- 7 palaces
- 1 large overworld
- Screens per Area in The Adventure of Link vary too greatly to make any definitive statement.
Mechanical Observations
- These games divide their world into areas very differently, but a common theme is having an open overworld and other locations branching from it.
- Having between 10 to 20 important non-overworld locations is common.
- Faxanadu has no overworld per se and is notably more linear.
- Typically there are 20 screens-worth of explorable space in each non-overworld location.
- Dragon Quest has much less than this, which corresponds to Dragon Quest also having fewer Total Screens and more grinding.
- A developer should expect to create 400 to 500 screens-worth of explorable content to match the topographical complexity of Black Box Era RPGs.
- NPC types are modest, typically less than 10.
- Faxanadu is an exception here, as it places emphasis on NPC variety, including large portraits with simple facial animations.
However, two of the Faxanadu NPC types counted here were not actually used in-game.
- Faxanadu is an exception here, as it places emphasis on NPC variety, including large portraits with simple facial animations.
- A developer should expect to create 40 enemy types to match Black Box Era RPGs, four times the number of NPC types.
- Taking consumables, equipment, collectibles, quest objects, and other similar objects together, a developer should expect to create 38 items to match Black Box Era RPGs.
Yay, these are some nice, actionable numbers!
Metroidvanias 🏰
And now for a genre that has recently been in the spotlight.
| Game | Team Size | Dev Time | Time to Beat | Mechanics | Visual Assets | Songs | Sound Effects | Word Count |
|---|---|---|---|---|---|---|---|---|
| Castlequest | 4 | <1 | 2 | 6 | 32 | 5 | 5 | Negligible |
| Metroid | 12 | 1? | 6 | 13 | 64 | 12 | 21 | Negligible |
| Castlevania 2 | 15 | 1 | 4 | 14 | 128 | 9 | 53 | 1,000 |
| Faxanadu | 8 | 1 | 6 | 14 | 128 | 16 | 25 | 2,500 |
| The Adventure of Link | 11 | 1 | 11 | 15 | 128 | 21 | 23 | 700 |
| Median | 11 | 1 | 6 | 14 | 128 | 12 | 23 | 700 |
| Mean | 10 | 1 | 6 | 12 | 104 | 13 | 27 | 880 |
| Mode | N/A | 1 | N/A | 14 | 128 | N/A | N/A | <500 |
Notes
- Castlequest, AKA Castle Excellent, is an early entry from 1985 and arguably not a full Metroidvania.
- It's historically interesting that three important Metroidvanias all came out in 1987: Zelda 2, Castlevania 2, and Faxanadu (AKA Dragonslayer 2.5).
- Metroid came out in 1986 and might seem like inspiration, but Castlevania 2 and Faxanadu are building off of elements earlier in their own franchises.
Observations
- Team Size is highly variable, from 4 to 15, but the average and median values end up at 10 to 11.
- Dev Time remains 1 year.
- Each game has a significantly different length.
- Though, I'll say that I don't completely trust the data from HowLongToBeat.com in this case:
I certainly didn't beat Castlevania 2 in only 4 hours! And I very much doubt I could have beaten Castlequest in 2, either. I think these are "times to beat as I read the online walkthrough"....

- Though, I'll say that I don't completely trust the data from HowLongToBeat.com in this case:
- Mechanics trend high at 14.
- This is probably core to being a Metroidvania as progress is typically gated mechanically.
- Soundtracks are a bit longer than usual at 13.
- Visual Assets tend high here as well; even the relatively simple Castlequest has more graphics than a lot of the games we've looked at.
- This seems to me like it's not part of the nature of Metroidvanias but instead was either a historical genre trend or an accident of my small sample size.
- Sound Effects hover around standard at 23-27.
- Word Count is high due to three of these five games also being Narrative Games.
- An important note not visible in the table is that while nowadays RPG elements are standard in Metroidvanias not all Black Box Metroidvanias had RPG mechanics.
I only have additional data on the three Metroidvanias already explored as RPGs, but here they are:
| Game | Sprite Tiles | Background Tiles |
|---|---|---|
| Castlevania 2 | 240 | 180 |
| Faxanadu | 3560 | 1024 |
| The Adventure of Link | >128 | >128 |
| Median | 240 | 180 |
| Mean | 1309 | 444 |
| Game | Areas | Screens per Area | Total Screens | NPC Types | Enemy Types | Items, Etc. |
|---|---|---|---|---|---|---|
| Castlevania 2 | 6 | 28 | 334 | 7 | 32 | 28 |
| Faxanadu | 33 | 16 | ~540 | 25 | 38 | 37 |
| The Adventure of Link | >17 | ≠ | 489 | 12 | 50 | 25 |
| Median | 17 | 22 | 489 | 12 | 38 | 28 |
| Mean | 19 | 22 | 454 | 15 | 40 | 30 |
More Observations
- Not too much has changed in either of these tables compared to the RPG section.
- In general, the medians are all a bit higher, reflecting these games being developed a little later in the Era.
- I do note that the average Sprite Tile count is much higher, likely for the same reason.
- Worth noting is that the amount of Items, Etc. is actually lower than in the RPG table.
- I believe this is because Metroidvanias place a greater emphasis on mechanical differences between equipment than Turn-Based RPGs. We're not just adjusting stats here!
Platformers 🦘
Last but not least, one of the most popular genres during the NES's lifetime
I've artificially excluded Metroidvanias here as including them would make the table unwieldy and we've already looked at most of them twice.
| Game | Team Size | Dev Time | Time to Beat | Mechanics | Visual Assets | Songs | Sound Effects | Word Count |
|---|---|---|---|---|---|---|---|---|
| Elevator Action | 1 | 1? | 0.5 | 2 | 8 | 2 | 9 | Negligible |
| Lode Runner | 5 | 4 | 3 | 4 | 8 | 1 | 8 | Negligible |
| Ice Climber | 7 | 1 | 2 | 5 | 8 | 3 | 17 | Negligible |
| Super Mario Bros. | 5 | 1 | 2 | 13 | 8 | 6 | 25 | Negligible |
| Castlevania | 10 | 1? | 3 | 9 | 64 | 16 | 37 | Negligible |
| Super Mario Bros. 2 USA | 20 | 1 | 5 | 19 | 128 | 14 | 19 | Negligible |
| Mega Man | 8 | 1? | 3 | 9 | 64 | 16 | 35 | Negligible |
| Median | 7 | 1 | 3 | 9 | 8 | 6 | 19 | <500 |
| Mean | 8 | 1 | 3 | 9 | 41 | 8 | 21 | <500 |
| Mode | 5 | 1 | 3 | 9 | 8 | 16 | N/A | <500 |
Observations
- Both two-player games that we've looked at are Platformers.
- The only game with a minigame, SMB 2 USA, is here.
- Almost every year we're interested in is represented!
- SMB draws a distinct line across this table: after it, platformers become dramatically more complex. Three reasons for this, I think:
- First, I'm guessing the huge success of this game made platformers in general more profitable.
- Second, Super Mario Bros.' sophisticated mechanics were an inspiration to other developers.
- Third, ROM chips became much less expensive around this time, allowing for more complex cartridge games.
- Beyond this, the median, mean, and mode in this table are about the most consistent we've seen.
- This makes intuitive sense to me as I feel like this genre may be the most representative of the NES as a whole. I bet teams got platformer production down to a science by the end of the system's lifespan.
- SMB 2 USA has a huge team by the standards of this Era, but most dev teams here are actually on the small end.
- Between all the uncertainty in Dev Time and Lode Runner's lengthy production period while the lead dev was in university, I'm hesitant to make any strong statement about Dev Time.
- The three Nintendo platformers do all seem to have been 1 year, though.
- Time to Beat matches the full table at 3 hours.
- Mechanics are typically 9 but can vary significantly.
- Vissual Asset size jumps up after SMB.
- Soundtracks run the gamut from nearly no songs at all to 16.
- Sound effects are just barely under standard at 20ish.
- Word Count is universally under 500.
These are by far the smallest scoped games we've seen since the Small Games table. The Small Games are still significantly smaller, though.
Now, let's re-run the numbers for just Super Mario Bros. and later.
| Game | Team Size | Dev Time | Time to Beat | Mechanics | Visual Assets | Songs | Sound Effects | Word Count |
|---|---|---|---|---|---|---|---|---|
| Super Mario Bros. | 5 | 1 | 2 | 13 | 8 | 6 | 25 | Negligible |
| Castlevania | 10 | 1? | 3 | 9 | 64 | 16 | 37 | Negligible |
| Super Mario Bros. 2 USA | 20 | 1 | 5 | 19 | 128 | 14 | 19 | Negligible |
| Mega Man | 8 | 1? | 3 | 9 | 64 | 16 | 35 | Negligible |
| Median | 9 | 1 | 3 | 11 | 64 | 15 | 30 | <500 |
| Mean | 11 | 1 | 3 | 13 | 64 | 13 | 29 | <500 |
| Mode | N/A | 1 | 3 | 9 | 64 | 16 | N/A | <500 |
Post-Mario Observations
- Team Size jumps up to 10ish thanks to those large Nintendo teams.
- Still can't speak confidently about Dev Time.
- Time to Beat remains 3 hours.
- Mechanics are a bit higher, but not as much as I expected.
- Visual Assets appear moderate at 64.
- Soundtracks are much more consistently near 15.
- Sound effects are greater than typical at 30.
- Still no words to speak of. No need for text in an oldschool platformer!
- I really wonder how well that would hold up in today's market, though, where stories are so ubiquitous?
As in the RPG and Metroidvania sections, I have some additional details on a subset of these games:
| Game | Sprite Tiles | Background Tiles |
|---|---|---|
| Castlevania | 1024 | 1024 |
| Super Mario Bros. 2 | 1079⸸ | 1024 |
| Mega Man | 1024 | 1024 |
| Median | 1024 | 1024 |
| Mean | 1042 | 1024 |
Notes
- I'm much more confident in these numbers than I was in the RPG ones, with one exception:
- ⸸I bet my count of Sprite Tiles for Super Mario Bros. 2 is off and there are really 1024. It could be higher, though, as the CHR-ROM has room....
- To break Super Mario Bros. 2 down further,
- The 1079 Sprite Tiles are divided into:
- 8 tiles per most normal enemies, but some are more
- 24 to 72 tiles per boss enemy
- 82 to 90 tiles per each player character
- 4 tiles for most items, projectiles, and effects, but a few animated items are much more
- The 1024 Background Tiles are divided into:
- 512 for level backgrounds
- 256 for the ending cut scene
- 256 for UI and various other elements
- The 1079 Sprite Tiles are divided into:
Graphical Observations
- Wow, talk about consistency!
I guess we can safely say that 1024 tiles is a great number to shoot for for both Sprites and Backgrounds when making a Black Box Era platformer, although do note that these are all more complex games later in the Era.
And for mechanical details...
| Game | Areas | Screens per Area | Total Screens | NPC Types | Enemy Types | Items, Etc. |
|---|---|---|---|---|---|---|
| Castlevania | 6 | 18 | 108 | 0 | 33 | 16 |
| Super Mario Bros. 2 | 20 | ≠ | 539 | 5 | 30 | 32 |
| Mega Man | 11 | 24 | 227 | 2 | 20 | 13 |
| Median | 11 | 21 | 227 | 2 | 30 | 16 |
| Mean | 12 | 21 | 291 | 2 | 28 | 20 |
Notes
- Castlevania has:
- 25 normal enemies and 8 bosses.
- 1 main weapon, 7 sub-weapons, and 8 other items.
- One of the other items, treasure, has several visual/mechanical variants.
- In Super Mario Bros. 2:
- Screens per Area vary too greatly to make any definitive statement.
- Only one NPC type exists and they show up only at the end, but they have a surprising amount of detail in the closing sequence.
- For Mega Man:
- In addition to the 10 playable levels there is a large level-like area featured in the ending cutscene.
- Most often there are 24 screens per area, but it varies from 16 to 25 with the average being 20.
- There are 19 normal enemies, 6 robot masters, and 5 other bosses (counting both phases of Dr. Wily separately).
- There are 7 complex weapons, 1 tool, and 5 other items.
Mechanical Observations
If I was hoping for consistency as in the graphical department, well, I'd be disappointed. Despite all being platformers, these three games play very differently. So a lack of mechanical cohesion makes sense. That said, there are still some lesons for us here:
- A Black Box platformer does not need a huge world; instead, the emphasis is placed on mechanical finesse and challenge.
- Around 200 total screens-worth of playspace may be sufficient.
- The game is likely divided into stages, one way or another, and no more than 12 levels are needed in most cases.
- In Black Box platformers that are not Metroidvanias, NPCs are few, simple, and rarely encountered.
- A developer should expect to make about 30 enemy types to match the complexity of a Black Box Era platformer.
- Power Ups, Items, Pick Ups, etc. are fundamental to post-Super Mario Bros. platformer gameplay, but 16 collectables/power ups may be sufficient.
Let's Sum Up Our Observations 🧭
In my opinion, the most important take away for the solo dev for all of Part 2 is this:
- 1980s game studio business models seem to have relied on spending no more than 1 year on any one game.
In addition, here are some guidelines we can give for a few genres.
-
Narrative Games:
- If you're telling a story, know that you're making a big game.
- Expect to write from 2,000 to 10,000 words.
- Expect to make more visual and audio content than in other types of games.
-
RPGs:
- Most Black Box Era RPGs are big games.
- Most are also Narrative Games so those notes apply.
- Downplay environmental detail in exchange for detailed player characters and enemies.
- In fact I'd say this is probably a good idea for most genres.
- Expect to create 400 to 500 screens-worth of explorable content to match the topographical complexity of Black Box Era RPGs.
- This typically takes the form of 10 to 20 important non-overworld locations each with 20ish screens-worth of explorable space.
- Expect to create 10 or fewer NPC types.
- Expect to create at least 40 enemy types.
- Expect to create about 40 items or pieces of equipment to match Black Box Era RPGs.
- Most Black Box Era RPGs are big games.
-
Metroidvanias
- Expect to make a large explorable world, probably with 500 or so screens-worth of explorable space.
- You don't need RPG mechanics and you don't need narrative.
- You do need to design and implement more mechanics than in other genres, around 14.
-
Platformers (non-Metroidvania)
- Expect to write less than 500 words to match the narrative simplicity of Black Box Era platformers.
- Expect to create 1024 Sprite and Background tiles each to match a Black Box Era platformer.
- Mechanical finesse is emphasized over playspace, so around 200 total screens-worth may be sufficient.
- Expect to make no more than 12 levels in most cases.
- Expect to make few NPCs and to rarely use them.
- Expect to make about 30 enemy types to match the complexity of a Black Box Era platformer.
- Power Ups/Upgrades are indispensable but 16 may be sufficient.
A final, more general point strikes me about these games.
- For the mostpart, they seem to have focused on either an engaging narratives or excellent mechanics, but never really both.
If memory serves, this is something that would change in later NES Eras, but for now the games are relatively simple.

For NES games, differences in scope can sometimes be seen in the hardware itself. Photo credit to Wourme on the Select Button forums.
That's It for Part Two ⏭⏭
In the last post in the series, we'll talk about what these games didn't do. 🔎
