ASCEND
Saturday · 2026-09-26 · ASCEND-EQ 1.0

AI Stock Market Predictions

The same six-stage pipeline that prices sports markets, retrained on equities. Every trading session the engine scores our US large-cap, megacap-tech and ETF universe across eight weighted factors, calibrates a probability that each call resolves correct, and compares it against the probability already priced into the tape. It publishes six conviction calls, four swing setups and three volatility reads — each with an expected move, an invalidation level and a fractional-Kelly position weight.

Markets are closed today — this is the next-session card, published early.

Conviction hit rate
58.9%
Swing hit rate
55.7%
Volatility hit rate
55.3%
Avg. return / idea
+1.4%

Rolling window anchored to a 1,000,000-idea Monte-Carlo validation. Blended accuracy 57.2% against a 38.5% break-even at the published invalidation distance. Directional equity models live in the mid-50s — anyone quoting you 80% is quoting a curve fit.

Measured accuracy

No settled ideas have been graded in the published ledger yet. This panel fills in from the settlement ledger as results land — nothing is shown before it is measured.

Market regime field

Assets move in clusters

Candle structures group equities, commodities, miners, agriculture, energy and bonds; the surrounding wave reflects changing volatility conditions.

Unlock the desk — $50 / month upfrontSee the sports engine →

Live market board

The real tape behind the desk: last price, session move, five and twenty session momentum and realised volatility, ranked by how much each name is actually moving today.

Reading the tape…

Today's free conviction call

One idea from the top of the card, released publicly so you can judge the work before paying for it. The remaining twelve are members only.

#1 · Conviction Call · Consumer Tech

AAPL — Apple

Cash equity · 10 sessions

Long @ 341.07
6% of risk capital
Live quote · closed · 20:00 UTC
Target
383.98
Invalidation
308.05
Expected move
12.58%
Win prob.
60.6%
Tape implies 57.2% · edge +3.4%Confidence 67/100

Sector Breadth and Price Momentum (20/60d) carry the call (Sector z=+1.61, Price z=+0.89, Positioning z=+1.12). Composite signal +0.98σ against a tape implying 57.2% for the same outcome over 10 sessions. Anchored to a live print of 341.07 USD (+1.53% on the session, 22% realised vol). RSI 66, MACD histogram +0.71, 50-day slope +1.1%, relative strength vs SPY +6.42%. Stop set at 1.5x the 20-day ATR (6.95). Measured factors this session: 5 of 8.

  • Sector Breadthz +1.61
  • Price Momentum (20/60d)z +0.89
  • Positioning & Sentimentz +1.12
Verifying your access…

The factor stack

Eight weighted inputs, published in full. Weights sum to 1.0 and are refit quarterly against out-of-sample sessions — never against the window being reported.

17%Price Momentum (20/60d)Risk-adjusted trend strength across two lookbacks, volatility-normalised.
16%Earnings Revision BreadthDirection and dispersion of forward EPS revisions across covering analysts.
14%Valuation vs SectorForward multiple against the sector median, growth-adjusted.
13%Institutional FlowDark-pool prints, block accumulation and 13F drift.
12%Options Skew & GammaPut/call skew, dealer gamma positioning and term-structure slope.
11%Macro Regime SensitivityBeta to real rates, dollar and credit spreads under the current regime.
10%Sector BreadthShare of sector constituents confirming the same direction.
7%Positioning & SentimentShort interest, retail crowding and news-sentiment residual.

How a session is built

Six stages, run in the same order every session. Nothing is published until the calibration stage agrees with the sizing stage.

1 · Universe screen

Liquidity, spread and borrow floors remove anything where the published entry and invalidation could not realistically be filled.

2 · Factor scoring

Eight weighted inputs — trend quality, earnings revision drift, flow, breadth, volatility term structure and more — collapse into one composite score.

3 · Calibration

The composite is mapped to a probability using out-of-sample sessions only, so the number quoted is the number that historically resolved.

4 · Edge test

That probability is compared against what the tape already prices. Ideas without a material gap are discarded, however good the score looked.

5 · Risk sizing

Survivors get a fractional-Kelly weight capped well below full Kelly, plus an explicit invalidation level so a loss is bounded.

6 · Publication

Thirteen ideas post before the open with the full reasoning attached. Nothing is edited after the fact — the record stands as published.

Frequently asked questions

What subscribers ask before they pay for the equity desk.

What exactly do I get each session?

Thirteen published ideas: six conviction calls, four swing setups and three volatility reads. Each carries a calibrated probability, an expected move, an invalidation level and a fractional-Kelly position weight, posted before the US open.

How is the AI stock model different from a stock screener?

A screener filters on static conditions. ASCEND-EQ scores the universe across eight weighted factors, converts the composite into a calibrated probability, then only publishes an idea when that probability is materially better than the one already priced into the tape.

What accuracy should I expect?

Blended validated accuracy sits in the mid-50s across a large Monte-Carlo sample, against a 38.5% break-even at the published invalidation distance. Directional equity models that claim 80% are curve fits — we publish the real number and the sample size behind it.

Which markets does it cover?

US large caps, megacap technology and liquid ETFs. Names must clear liquidity and spread floors so that published entries and invalidations are realistically fillable.

How does billing work?

Access is paid upfront: $50 for a month, $135 for a quarter or $480 for a year. There is no auto-renew trap — when a term ends, access ends unless you renew.

Is this financial advice?

No. Ascend publishes quantitative research, not investment advice, and is not a registered investment adviser. Every idea can lose. Size positions accordingly.

Keep reading

AI Stocks FAQ — model, edge & accuracyThe eight-factor stack, in fullHow a session is builtRisk, sizing & drawdownThe model behind both enginesValidation & backtest auditGrade a pick for $1.99AI Picks & Insights research

Risk disclosure

Ascend publishes quantitative research, not investment advice. Nothing here is a recommendation to buy or sell any security, we are not a registered investment adviser, and past model performance does not predict future returns. Every idea can lose. Size positions accordingly and never risk capital you cannot afford to lose.