The Engine

Zenith ranks the day's biggest US stock-market gainers and publishes a short thesis on the top five, about 30 minutes before the 4:00 PM ET close. The ranking and the scores are computed from market data — filings, historical outcomes, statistical models, price and volume. A language model is used only to phrase the write-up; it never produces a number. Zenith places no trades and holds no positions.

Zenith can also be wrong. Its limitations are set out below and should be read before relying on anything it publishes.

The engine answers exactly one question about each stock: will it close lower at the next close than it did at this one? Close to close, roughly 4:00 PM ET to 4:00 PM ET, one session. Nothing on Zenith is a view on where a stock goes over a week, a month, or the rest of the competition.

That basis comes from how the Stock Market Game actually fills orders. It is an end-of-day game: an order placed during market hours fills at that day's closing price, and an order placed after the close fills at the next day's close. Orders stay pending and cancelable until then.

Two things follow. A thesis is only useful if it reaches you before the close, which is why it lands at 3:30 rather than after the bell. And only the regular session counts — pre-market and after-hours moves never become a fill price, so the engine ignores them.

The theses post about half an hour before the close. That is a scheduling decision, not a trading one — we are not saying 3:30 is or is not a good moment to short a stock, and there is nothing special about the price at 3:30.

It is a compromise between two things that pull against each other. Late in the session the day's biggest gainers have mostly settled, so the list at 3:30 is usually close to the list at 4:00; earlier in the day it isn't, and names trade places constantly. But your order still has to be in before the close to fill at today's price. Half an hour leaves room to read five theses and place an order without cutting it fine.

Nothing is settled at 3:30. Stocks move in the last half hour, and the ranking can still change before the bell.

Each thesis carries a short score from 1 to 10. It ranks the day's candidates against one another. It is not a probability and not a confidence level — an 8/10 does not mean an 80% chance of anything, and a 9 sitting next to a 3 says the engine prefers one to the other, not that either is a good idea.

Where a thesis quotes figures from past cases, those describe how a group of similar past spikes behaved. They are not predictions about the stock in front of you.

A language model is currently used for one thing: rewriting the finished findings into readable sentences. The sentences carrying the odds, the expected move and the risk warnings are written by the engine and passed through untouched. Every figure the model does write is checked against the engine's own numbers before publication — if one doesn't match, the model's version is thrown away and the engine's wording is used instead.

What a model never does, in any mode:

• choose which stocks appear on the screener or in the top five
• produce or adjust the short score, the base rate, or any figure
• look anything up, browse the web, or add facts of its own

Nothing about you is ever sent to an AI provider — it receives only public market facts. See the Privacy Policy.

Nobody can predict what a stock will do tomorrow, and this engine doesn't either. It estimates, from history, how situations like this one have tended to resolve. Situations that historically resolved one way still resolve the other way often — that is what a probability means. Expect losing calls, including on high-scoring ones.

A stock that has already risen sharply can keep rising, and often does. Historical patterns can also stop describing the market in front of them. Treat a thesis as one input to your own judgement rather than a conclusion to act on.

Prices, volume and company details come from a third-party market data source. Filing information comes from the SEC's EDGAR system, and headlines from a news provider. All of it can be delayed, incomplete, or wrong, and the engine has to work around several known problems:

• Stock splits. On the day a split takes effect, percentage changes can be computed against the pre-split price, manufacturing enormous fake gains. Moves with that signature are dropped.
• Halted stocks. A stock halted from trading keeps reporting the same values day after day. Rows that repeat exactly are dropped rather than presented as a live mover.
• Ticker symbols aren't unique. The same few letters can refer to different instruments on different venues, and symbols get reused when companies leave the market. Every symbol is qualified with its exchange before it's looked up or charted, and if a symbol can't be matched confidently the row is dropped and the filing check returns nothing rather than the wrong company's documents.

These guards are imperfect. If a listing looks wrong, it may be — tell us at support@zenithscreener.com.

The screener looks at common stock on the NASDAQ and NYSE, ranked by percentage gain over the regular session. No over-the-counter stocks. To appear, a stock has to have finished the previous session at $3 or more with a market value of at least $25 million.

The price and size floors are set where they are because the Stock Market Game applies similar limits to what you may trade. They are not a promise: the game also reserves the right to block or unwind trades in unusually volatile stocks, and its rules govern, not ours.

They are measured the way the game measures them — against the previous close, not against what a stock costs right now. That difference matters most at the top of the board, where a name that has doubled today can be above $3 on screen and have closed well under it yesterday. Those rows are left off the board rather than listed with a warning, so the ranking you see is one you could have acted on.

Zenith is a research aid — a screener plus analysis, like reading a finance site before you trade. You place your own trades in the official game. A few things are your responsibility, and they matter:

• DECA expects each team to do its own research, with a portfolio that reflects its own strategy. Treat a thesis as one input to your thinking, not as your strategy.
• If Zenith's analysis informs a written entry or a presentation, cite it as a source. Written events require a signed academic-integrity statement and a bibliography. Do not present Zenith's text as your own writing.
• Rules vary by chapter and by event. Yours are between you, your advisor, DECA, and the SIFMA Foundation — check them if you're unsure.

The full version is in the Terms & Conditions.

Everything on Zenith — the rankings, the theses, the scores, the risk labels — is educational analysis about a simulated trading competition. It is not investment advice, not a recommendation to buy, sell, or short any real security, and not an offer of brokerage or advisory services. Do not use Zenith to make real-money decisions. Short selling with real money can lose more than you put in.

Zenith is not affiliated with, endorsed by, or connected to DECA Inc. or the SIFMA Foundation.

Questions about anything here: support@zenithscreener.com.