Randsight

Joint-model set prediction and damage calculator for Pokemon Showdown Random Battles.

# Chrome Web Store listing — paste-ready copy Everything below is written to be pasted into the developer dashboard as-is. Where a field has a hard character limit, the count is given and the text is inside it. --- ## Name (45 char limit) ``` Randsight ``` `13/45`. Leaving it short on purpose — the store appends nothing, and a name padded out with keywords ("Randsight — Pokémon Showdown Random Battle Set Predictor") reads as SEO rather than as a product. --- ## Short description (132 char limit) ``` Live set prediction for Pokémon Showdown Random Battles: what the opponent is probably running, and what it does to you. ``` `119/132`. Alternatives if you want a different emphasis: ``` Predicts the opposing Pokémon's moves, item and Tera in Random Battles — and updates every time they reveal something. ``` `117/132` ``` A Random Battle overlay that keeps a running estimate of the opposing team, with damage numbers and switch advice. ``` `113/132` --- ## Detailed description ``` You are nine turns into a Random Battle, looking at a Gholdengo, and the game comes down to one question: does it have Focus Blast? You can't know. But you can know the odds — and they are not the odds they were on turn one. Randsight watches your battle and keeps a running estimate of what the opposing team is holding. Every move they use, every item that gives itself away, every Tera type spent narrows it. By the time the decision matters, you are usually not guessing. WHY IT ISN'T A LOOKUP TABLE Random Battles don't roll moves independently. The generator picks a role first, draws moves to fit that role, and then chooses an item based on the moves it landed on. So what you have already seen genuinely changes what is still possible. A Great Tusk that has shown Rapid Spin is a different Pokémon from one that has shown Bulk Up — different remaining moves, and different item odds too. Tools that show the same static percentages every game can't tell those two apart. Randsight models the relationship, so the numbers answer your battle instead of describing the average one. THREE TABS, THREE QUESTIONS SETS — what can they still be carrying? One card per opposing Pokémon: the odds on each move slot you haven't seen, the role, the item, the ability, the Tera type, and their speed at their actual level. Anything that has stopped being a question collapses to a single line, so what's on screen is only what's still undecided. Hover anything to find out what it does. DAMAGE — do I survive, and do I kill? The current matchup in both directions, opening with the answer in plain words: "You survive — worst is Headlong Rush at 64%, about 2 hits." Their moves are ranked by threat, not raw damage: how hard it hits, discounted by how likely they are to actually have it. A normal calculator can't rank a move it doesn't know about yet. And the numbers are computed across their whole item and ability distribution rather than one hopeful guess. Immunities are shown, not quietly skipped. SWITCH — who can come in? Your six, ranked by how much HP is left standing after the worst hit they can realistically land, with a mark for who outspeeds. It leads with the answer. If nothing on your bench survives, it says so instead of recommending a Pokémon that dies. IT READS THE BATTLE, NOT JUST THE DATA Three things no set file can tell you, taken straight from what happened: Terastallization is once per side. After they use it, Randsight stops offering Tera types for the rest of their team. A Choice item locks them into the move they just used — so the panel stops pretending the others are live this turn. A move used to its last PP cannot happen again. WHAT IT WON'T DO It won't tell you it knows. Uncertainty stays visible: a 34% is drawn as a 34%, not rounded up into a recommendation. Where a number depends on something unrevealed, it says which assumption it made. It doesn't play for you. There is no auto-anything — it reads the page, it never sends a move, and it never touches the socket. Doubles damage doesn't yet model spread-move targeting or redirection, and the panel tells you so rather than quietly being wrong. FORMATS All fourteen Random Battle formats with published prediction data: gen 1 through gen 9 singles, gen 9 doubles, Champions doubles, Baby, BDSP, and Let's Go. Blitz and unrated lobbies work too. Multi and Free-For-All fall back to the closest format and are labelled as approximate rather than presented as exact. PRIVACY It collects nothing. No account, no analytics, no telemetry, no server — there is nowhere for your data to go, because there is no "there". Two kinds of request leave your browser: the public set-data files from pkmn/randbats, and Pokémon sprites from Showdown's own sprite server. The only permission it asks for is storage, used to remember where you left the panel. The full policy is linked below and also ships inside the extension. Open source under the MIT licence, and the damage calculator is Smogon's own, included unmodified so you can read it. Not affiliated with Smogon, Pokémon Showdown, or Nintendo/Game Freak. ``` --- ## Category **Primary:** Entertainment Not "Productivity" — a reviewer who opens it sees a game overlay, and a mismatched category invites a closer look for the wrong reason. ## Language English (United States) --- ## Single purpose (required field) Reviewers reject vague answers here. This one is narrow and matches what the code actually does: ``` Randsight has a single purpose: to display predicted set information for the opposing team during a Pokémon Showdown Random Battle. It reads the battle state already present on play.pokemonshowdown.com and renders an overlay panel on that page. It has no other function and operates on no other site. ``` --- ## Permission justifications ### `storage` ``` Used only to remember the overlay's position, size, theme, which cards the user expanded, and a cached copy of the public set-data files so they are not re-downloaded on every battle. All of it stays in chrome.storage.local on the user's own machine. Nothing is transmitted. ``` ### Host access Two separate things a reviewer will ask about, and they must not be conflated — the manifest declares `host_permissions` for the two data hosts, and separately declares content-script matches for the two Showdown hosts. ``` The extension does two distinct things with hosts, and asks for the narrowest access that allows each. Content scripts are declared statically in the manifest for exactly two hosts, play.pokemonshowdown.com and replay.pokemonshowdown.com, because the extension's entire function is to draw an overlay on the page where the battle happens. They cannot be extended to any other site at runtime. No tabs permission is requested and the extension cannot see any other tab. On those two pages it reads the battle state the page already holds and writes only its own overlay element. host_permissions are declared for two hosts and nothing else: data.pkmn.cc and raw.githubusercontent.com/pkmn/randbats. These are the two locations of the public random-battle set-data files the prediction is built from. They are fetched as JSON, parsed with JSON.parse, and never executed. No user data is sent with those requests; they are plain GETs for static public files. ``` ### Remote code ``` None. All code is contained in the package. The extension fetches two kinds of remote resource, both of which are data and neither of which is executed: JSON set-data files from data.pkmn.cc (parsed with JSON.parse) and PNG sprite images from play.pokemonshowdown.com (used only as a CSS background). The damage calculator is a vendored, unminified copy of @smogon/calc included in the package. ``` --- ## Data usage disclosures Every checkbox: **not collected**. There is no analytics, no telemetry, no account, and no server belonging to this extension. Certifications to tick: - [x] I do not sell or transfer user data to third parties, outside of approved use cases - [x] I do not use or transfer user data for purposes unrelated to my item's single purpose - [x] I do not use or transfer user data to determine creditworthiness or for lending purposes --- ## Screenshots `docs/store/` — `1-sets.png`, `2-damage.png`, `3-switch.png`, three at 1280×800. Generated by `npm run shots`, so they track the shipping build rather than drifting away from it the way the old hand-taken ones did. Each already carries its own headline in the image, so the store's caption field can be left empty. The uncaptioned crops in `docs/site/` are for the landing page, which writes its own captions — don't upload those here. --- ## URLs | Field | Value | |---|---| | Homepage | `https://nabilcodhu7360.github.io/randsight/` | | Privacy policy | `https://nabilcodhu7360.github.io/randsight/privacy.html` | | Support | `https://github.com/NabilcodHu7360/randsight/issues` | Both github.io URLs need GitHub Pages switched on (Settings → Pages → `main`, folder `/docs`). The privacy field is required and the form will not submit without a URL that actually resolves. --- ## Before you hit submit - [ ] Privacy policy actually reachable at the URL you entered — open it - [ ] The zip is the one from `npm run package` (~0.74 MB, 46 files), not a hand-made one and not `randsight-repo.zip` - [ ] `LICENSE` and `THIRD-PARTY-LICENSES.txt` are inside the zip (they are, and `load.test.js` asserts it)