Decoding notes — One Piece Going Baseball (J)

End-goal: produce an English-translated patched ROM. This file documents progress so far.

Confirmed character mappings (9)

Source: matching menu screenshots to script strings by repeat-pattern signature (positions of identical characters) and string length.

Both anchor strings are unique matches in the 596-string script under length + repeat-pattern constraints, so the mappings are high-confidence.

What the script does NOT contain

The 596-string pointer table at 0x4ED7F8 holds story dialogue only. The following are sprite/tilemap graphics, not script strings, and attempts to find them in the dump fail:

• Mode-select menu items (ストーリー, 対戦, ミニゲーム, エディット, オプション, あそびかた) — the unique ストーリー match at #074 is almost certainly dialogue containing the word “story”, not the button label itself.
• Team names (ルフィ海賊団, イーストブルー, アラバスタ, etc.)
• Screen headers (モードを選んでね, 味方チームを選んでね, 誰と対戦する？)
• Team-description blurbs (パワフルなパンチをうつチームだ！ etc.)

To localise the menu/team strings you’d edit the sprite tiles directly, not the script.

Encoding correction

The original dumper docstring claimed:

0x10XX = hiragana, 0x20XX = katakana, 0x30XX = small kanji/symbol

That’s empirically wrong. The 0x10 page contains hiragana and common katakana (ト, リ, ン, グ, ラ, ド, イ are all 0x10XX). The 0x20 page also contains a mix — ス shows up as 0x2000 in #074. The pages appear to be ordered by glyph frequency, with the same character sometimes present in multiple pages as alternate visual styles (body-text vs title-text).

ROM cross-comparison

All three ROMs (J + Chi fan-patch + K fan-patch) are present in rh_uploads/JKC/Menus/.

• J and Chi share the same pointer-table layout (596 strings at 0x4ED7F8). Their script.json dumps line up slot-for-slot, so slot #N in J ↔ slot #N in Chi is the same dialogue line translated. See script_aligned.txt for a side-by-side dump.
• K restructured the ROM — its pointer table is at a different offset, so the J dumper produces garbage when run on the K ROM. Cross-referencing K would require a separate K-specific dumper.
• Diff between J and Chi: ~254 KB of bytes differ, scattered across many regions. Notable:
  • 0x4EB000-0x4ED7F8: script text bank (translated, expected)
  • 0x0DB000-0x10A000: ~200 KB heavily modified — likely the primary font tile region, with Chi glyphs replacing J glyphs
  • 0x5C5000-0x5CC000: ~28 KB that was empty (0x00 padding) in J but contains data in Chi — likely an added font block for extra Chinese chars beyond what fits in the original tile region

What blocked the “Chinese-as-Rosetta-stone” plan

The plan was: dump Chi, render its glyphs, read them as Chinese, translate to English. It stalled at the “render its glyphs” step:

• The 28 KB block at 0x5C5000 didn’t render as recognizable Chinese characters in any standard format (1bpp/2bpp at 8×8, 12×12, 14×14, 16×16, both bit orderings, multiple alignment offsets all tried).
• LZ77 decompression at every 4-byte-aligned 0x10 byte produced zero valid blocks in the 2 KB–64 KB range.
• The font data is probably in a custom format used by the original game engine (or the patcher’s custom encoder), not one of the GBA-standard tile formats.

What blocked the J font extraction

Same issue from a different angle:

• The J font isn’t visible in any standard tile format when rendering candidate regions (0x040000-0x180000 scanned).
• LZ77 / BIOS-decompressible blocks: zero found.
• The font is likely either custom-encoded, or referenced through a pointer that we haven’t traced yet (would need to disassemble the ARM/Thumb code that calls the font-render routine).

Practical paths forward

Ranked by user effort:

1. (Lowest effort) Take 5-10 story-mode dialogue screenshots from the game in any GBA emulator (mGBA, VBA-M) and upload them. Each dialogue line maps 1:1 to a script entry. With those, the top ~30 hiragana codes can be locked down, which covers 60-70% of every dialogue line. That’d be enough to translate the entire script via context.

2. (Medium effort) Find the CGP/PGCG fan-translation team’s documentation or .tbl file (searches indicate the Chinese patch is widely distributed but I couldn’t surface the dev resources via public web search).

3. (High effort, technical) Disassemble the J ROM’s text-rendering routine to locate the font pointer in code. This is a multi-day ROM-hacking project and unlikely to be the cheapest path.

For an English translation, option 1 + a translator (human or machine) for the resulting hiragana dialogue is the recommended path.