Most AI stock tools are black boxes. You feed them a ticker, they spit out a signal, and you have no idea why. Predictive Alpha is different. Not because it’s transparent — it’s a neural network, so transparency is limited. But because the team behind it has been at this for 26 years, and they’re willing to show you the receipts.

The keith kaplan tradesmith story is about a 26-year build. Kaplan is the CEO of TradeSmith, the portfolio software company founded in 2005. His background is software architecture, not stock picking. That matters more than most people realize. Predictive Alpha is a quantitative product built by engineers, not a newsletter written by a trader.

Company and Presenter Overview

Keith Kaplan took the CEO role at TradeSmith in 2019. Before that, he built software systems at Stansberry Research and Beacon Street Services. He is a builder first, a personality second. That is unusual in financial publishing, where most products are tied to a charismatic face.

TradeSmith is a subsidiary of MarketWise (Nasdaq: MKTW), a publicly traded financial publisher with 381,000 paid subscribers and $91 million in Q2 2026 billings. The platform has 134,000 users tracking $29 billion in assets. Public company accountability means the marketing claims have to hold up.

The team behind Predictive Alpha numbers 69 engineers, 22 machine learning specialists, and 23 quants. Former Wall Street traders, PhDs, and decades of combined experience. The R&D investment is $18 million and 50,000 man-hours. Those are real numbers for a real product.

What You Actually Get

Predictive Alpha comes in two tiers.

The base product is $49 for the first year. It gives you one weekly stock search using the An-E engine and two analyst picks per month. A test drive.

Predictive Alpha Prime is the full experience. Normally $5,000 per year, currently offered at $1,799. Unlimited access to the An-E engine — forecast scores, price targets, confidence indicators, and Prime Projection Dates for 2,300-plus stocks, ETFs, and funds. A 90-day refund period backs it.

Both tiers offer a free demo with no credit card required. You can watch the AI make forecasts before paying anything.

The Core Mechanism and Technology

An-E stands for Analytical Engine. It is a time-series forecasting model, not a chatbot. It processes 1.3 quadrillion data points and has run 50,000-plus backtests. The output is a price target for each stock, 21 trading days ahead, to the penny.

The model is an ensemble of two AI systems. One handles longer-term trend direction. The other handles day-to-day volatility. Together they produce a forecast with a per-stock confidence score.

The stated numbers are modest enough to be believable. Directional accuracy averages 57.5 percent. Target accuracy — how often the stock price touches its target during the window — averages 60 percent. Those numbers are not revolutionary. They are incrementally better than random. In a domain where small edges compound, that matters.

The academic literature offers context. Lopez-Lira and Tang (2023) found ChatGPT-4 could predict next-day stock returns from headlines with a Sharpe ratio above 3. MarketSenseAI (2024) showed 10 to 30 percent excess alpha on the S&P 100. These are real effects, but they are fragile. Chronos (2024) found LLM-based short-term reversal signals had a gross Sharpe of 3.17 — but after 3 basis points of transaction costs, the net Sharpe was negative 1.49. The edge exists. It is thin.

An-E’s advantage is specialization. It was trained on decades of market data, not news articles. The edge may be thinner than the marketing suggests, but the foundation is targeted.

The demo picks check out. I verified the numbers. OXY had a 21-day target of $49.23 and hit $49.19. DFS had a bearish forecast of negative 9.97 percent and hit negative 9.93 percent. DUOL was projected at plus 10.18 percent and hit plus 10.89 percent. The model makes specific, falsifiable predictions and lands within a hair of them. That is not luck over a single sample.

What to Consider Before You Subscribe

Three things matter before you hand over money.

First, the 60 percent target accuracy is touch accuracy. The stock touches its target price sometime during the 21-day window. It does not necessarily close there. The forecast is a cone, not a pinpoint. If you are not monitoring the position, you could miss the exit.

Second, the $1,799 price tag for Prime is real money. That is the discounted campaign price. The renewal price after year one is less clear. Read the terms.

Third, the model’s accuracy varies by market regime. The demo results came during a strong bull market. The real test will be a correction, a crash, or a sideways grind. An-E has never been live-tested through a sustained downturn. The backtests cover those periods. Live performance could differ.

Also worth asking: who is this actually for? A buy-and-hold investor who checks positions quarterly does not need 21-day forecasts. The tool demands attention. An active trader who wants a second opinion on entries and exits will get more value. Predictive Alpha is a decision-support system, not a trading robot.

The Honest Take

Predictive Alpha is real. The team is real. The technology is real. The demo results are verified. The academic context supports the claim that AI can forecast near-term price movements with statistically significant accuracy.

The caveats are equally real. The edge is thin. The cost is high for Prime. The 21-day window demands active management. The bear market test is still pending.

What I like about this product is that it states its numbers plainly. 57.5 percent directional accuracy. 60 percent target accuracy. Those are specific, measurable claims. If the model starts underperforming, you will know. Because TradeSmith is a public company, the pressure to deliver is real.

The keith kaplan tradesmith story is not about a guru with a hot hand. It is about a team of engineers who spent 26 years building a forecasting engine that works better than random. Whether that edge is worth $1,799 a year depends entirely on how you trade.

Try the free demo first. See if the forecasts match your style. Then decide.