How this review was done
For anyone curious about how all this was found, here is the short version.
Every ROM was run in a purpose-built emulator that keeps a record of every byte of code that executes. Automated tests then mashed the buttons, tried thousands of combinations, and forced the game down paths it would not normally take. Anything left unexplained became a lead: text that never appeared, code that never ran, or a check that compares your button presses with a list.
Those leads were then followed up by reading the code. Where a code looked real, it was tested in the emulator and a screenshot was taken.
Sounds and pictures
Later runs looked for things to see and hear. For sound, each game was played for half an hour or more, to find its sound code and the way the game asks it for a sound. Every sound number was then played through that code on its own, and what reached the sound chip was recorded. Where the attract sequence asks for a sound, that sound was played both ways, by the real game and by the sound code alone, and the two were compared write for write. Lists that did not match were left out. That gave sounds for 88 of the games. The rest ask for their sounds in ways the run could not drive yet.
Each game was also run as a Western and as a Japanese console, with the same buttons pressed, and the screens were compared. The emulator kept the whole picture each game drew, including the parts outside the Game Gear’s window. It also recorded where every byte of the cartridge went, which is what the strip in At a glance shows.
The games that only had leads
After the first pass, 149 of the 233 files had no confirmed finding. Each one went through the same steps:
- Read the leads. Each ROM got a one-page summary of its unseen text, forced screens and button checks.
- Compare with published codes. The SMS Power cheat page for each game was read directly, 159 pages in all.
- Look in the ROM. Most codes and passwords are stored as a table, so each one could be searched for.
- Test it. The strongest ones were run in the emulator.
| What happened | Files |
|---|---|
| A hidden feature was tested and worked | 15 |
| Published cheats, credited or linked in the game’s section | 35 |
| Only ordinary level passwords published | 11 |
| Only cheat-cartridge codes published | 37 |
| Nothing published and nothing confirmed | 51 |
| Total | 149 |
Try it yourself
Every game has a technical report in the ROM Reports folder, named after its ROM file.
- Open
REPORT.mdin the game’s folder. - Read Hidden or unused content. Items marked Confirmed are the strongest, and items marked Possible are leads worth following.
- Open
text.txtto see all the text in the ROM, and which parts never appeared in testing. - Look at
forced_1.pngtoforced_3.png, which show screens reached by forcing the code down other paths. - Open the
soundsfolder, if there is one, for every sound the review recorded from that game. - Try the buttons in an emulator with a debugger, such as Emulicious or Mesen. Set a breakpoint on the address given under Where, and see whether it runs.
- Try it on a real Game Gear, and allow for the timing being different.
Round 2
A second round on 12 November 2026 went back over everything left unfinished. The text in every ROM was read again with a better decoder, and the most likely passages were read by hand, which turned up developers’ messages, debug screens and leftover source code. Recorded sound such as speech was caught and rebuilt. Each cartridge’s build stamp, header and built-in high-score table were read, the regional versions were compared byte by byte, and two emulators were joined by a virtual link cable. Open leads were followed up too, and one of them turned out to be Crystal Warriors’ code for the ending.
The last part of the round finished on 13 November. The lists of addresses that each game jumps through were rebuilt for all 231 ROMs, and 3,105 of the entries that never ran in testing were forced one at a time from saved moments. Almost all of them gave restarts, logos, menus, ordinary play or a broken picture. Four games showed something their developers left in: the placeholder message in Ecco: The Tides of Time, two of the 36 crash screens in True Lies, and error screens in The Incredible Hulk and Mortal Kombat II.
Round 3, later on 13 November, finished the plan’s last open items. Two emulators with a finer link played Squinky Tennis to the end. Five codes were explained or seen working, among them Solitaire Poker’s lower target score, Dragon’s infinite lives, Superman’s ZAQ cheat and Star Wars’ list of scenes. Kawasaki Superbike Challenge’s TRACK LEN ERROR was put on screen, and Mean Bean Machine’s sound test was brought up by running its screen script. A new way of asking for sounds gave Alien 3 a list of 86. The 165 test-run bug reports left from the first review were traced again: 31 turned out harmless, one became a confirmed bug in Desert Strike, and the rest now say what the trace found.
A last pass on 14 November chased the leads round 3 had left. The inputs for VR Troopers’ cheats menu, Ecco’s EPROM menu and George Foreman’s cheat mode all worked once the timing was right, X-Men’s mystery number turned out to be the player’s position, and Jurassic Park’s Japanese sound test was found still sitting in the Western cartridge.
Rounds 5 and 6, on 18 and 19 November, looked at things the earlier rounds had not touched. The emulator logged every line of the picture to see how the games use the hardware: scroll effects, colour changes partway down the screen, stereo and how busy the processor is. A scan of the spare space found the PC program in Beavis and Butt-head. The password checks of several games were read and tried, explaining how Mean Bean Machine’s and Shinobi II’s passwords reach their sound tests. Every game was switched on with cleared and with random memory, its random-number routine was looked for, and the recorded button presses behind its attract demo were traced.
Where this guide’s content came from
- The words: everything was written for this guide. It was also checked word for word against all the online material read during the review.
- The screenshots: every one was taken from the ROMs in the review’s own emulator. None came from a website. The game graphics they show still belong to the games’ publishers.
- The cheats: every code printed here was found in the ROM or tested in the emulator. Where someone else published it first, the section credits them. Codes that could not be confirmed were left out, and the section links to the site instead.
- The quotes: the text quoted from the ROMs is exactly as the cartridges store it, and belongs to the games’ makers.
- The sounds: every recording was made from the games’ own sound code in the review’s emulator. The music and effects still belong to the games’ publishers. That is why this guide includes only the few sounds nothing seems to ask for, as short clips, and not whole soundtracks.
What this guide cannot tell you
- A quiet section does not prove there is nothing there. It means this review did not find anything.
- Late-game code is under-tested. Automated play seldom finishes levels, so secrets that need you to complete a game may have been missed. Six games were played from their later levels using their own cheats, but most were not.
- Some online checks are second-hand. SMS Power’s pages were read directly. The Cutting Room Floor and Sega Retro refuse automated page reads, so those comparisons come from search summaries.
- “Not found online” does not mean new. It means the search did not find it.
- A sound nothing asks for may still be used. A game can work a sound number out in ways the checks miss, and a sound test can play anything.
- The recordings are close, not perfect. The Game Gear’s stereo setting was ignored, so they are mono, and none of them was checked on a real Game Gear.