AI Stock Picker Trades Both Ways: GPT-6 Sol Buys 3 Stocks and Sells 2 After Earnings

In the fourth round of our public experiment, the AI stock picker trades in both directions. The AI model GPT-6 Sol builds a portfolio for a 10,000 USD demo account in StocksTrader with a horizon of 10 trading days. It buys three stocks after strong quarterly reports and sells two after weaker forecasts. All five positions open at market at the US open on 9 November and close no later than 20 November. The result goes up on this page next to a US 500 position of about the same size.

AI Portfolio Status

How the AI Portfolio Experiment Works

The portfolio was built with GPT-6 Sol, and the AI does the whole selection. It assesses the market, picks the stocks and the direction of each trade, splits the budget, and calculates the position size, Take Profit and Stop Loss. Our part is technical: we check the quotes in StocksTrader, open the positions on the demo account and record the result.

The request sent to the AI, translated from the original:

The portfolio follows six rules:

• Fixed horizon. Positions open at market on Monday 9 November at the US open. Everything still open is closed on Friday 20 November at the US open, whatever the result.
• Early exit on profit. If the total net profit of the AI portfolio reaches 850 USD before 20 November, all remaining positions are closed at once. The limit counts closed trades, open positions and costs. The US 500 position stays outside it.
• Buys and sells. New in this cycle: two of the five trades are sells. A sell earns money when the stock falls and loses money when it rises.
• Levels follow the entry. Take Profit and Stop Loss are recalculated from the actual entry prices with the same percentages.
• No second attempt. A position closed at its Stop Loss is not reopened, and losing trades are not averaged down.
• Leverage. Up to 1:2 on stocks and ETFs, up to 1:50 on currencies and crypto.

Market Backdrop for the Next Two Weeks

• US stocks. After a drop on 5 November the market partly recovered on 6 November. Single stocks reacted to their reports in very different ways: Airbnb rose about 11% and Cloudflare about 10%, while Axon lost about 18% and DraftKings about 21%. That spread gives room for single-stock trades in both directions. Whether Friday's moves continue is the trading idea this portfolio tests.
• Company reports. Airbnb revenue grew 19% and Cloudflare revenue 39%. Axon lowered its revenue growth forecast and left out the Dedrone deliveries investors had expected. DraftKings is putting large sums into its Predictions business, which holds back its forecast for adjusted EBITDA. These differences shaped the selection.
• Inflation and rates. US CPI on 10 November and PPI on 13 November can move many stocks at once. The two sells reduce how much the portfolio leans on a rising market, and they work as a partial offset only: the bought stocks can fall while the sold ones rise.
• Technology. The NVIDIA report on 17 November after the close matters for the mood in chip stocks and for AMD. In sharp moves a stop can fill at a worse price than set. The portfolio keeps each position small to limit that risk.

How the AI Split the Budget

• Core block: buy ABNB and NET. Both companies reported strong quarters, and the trades count on the market continuing to value their businesses higher. The block comes to about 3,892 USD.
• Tactical block: buy AMD. The trade counts on a partial recovery after a sharp fall: selling pressure eases and interest in chip stocks returns. The main risk is a further fall in technology stocks. The block takes about 3,143 USD.
• Short block: sell AXON and DKNG. The trades count on pressure holding after investors revised their expectations. Axon lowered its forecasts for revenue growth and free cash flow, and DraftKings is spending heavily on Predictions, which holds back its current profit. The main risk for both is a strong rebound after the fall that has already happened. The block comes to about 3,046 USD.

The portfolio combines buys of AMD, Cloudflare and Airbnb with sells of Axon and DraftKings. It counts on a partial recovery in AMD, a continued positive reaction to the Cloudflare and Airbnb reports, and lasting pressure on Axon and DraftKings. The main risk is a change in market mood: the bought stocks can fall and the sold ones can rise. The sells reduce how much the portfolio leans on a rising market, and they give no guarantee against a loss.

AI Portfolio Positions

Prices below are reference prices from the close on 6 November. The positions open at market at the US open on 9 November, so the actual entries can differ. Take Profit and Stop Loss are recalculated from the actual entries with the same percentages, and both go up here after the open. Charts for every stock are on the RoboForex stock charts page.

Percentages show the change in the stock price from the reference entry and are rounded.

Why the AI Picked Each Stock

AMD. Revenue for the third quarter rose 45% to 13.4 billion USD, and data centre revenue rose 97% to 8.5 billion USD. For the fourth quarter AMD expects revenue of about 17 billion USD, 66% more than a year earlier. The buy counts on the stock recovering after its sharp fall while demand for server processors and AI accelerators stays strong. The main risk is a further repricing of the technology sector, whatever the growth of the business.

ABNB. Airbnb revenue for the third quarter rose 19% to 4.9 billion USD, and the value of bookings rose 21%. The company announced an extra share buyback of up to 12 billion USD. The trade counts on the rise after the report continuing.

NET. Cloudflare revenue rose 39% to 782.5 million USD, which supports the buy. Its GAAP gross margin fell from 74.0% to 71.2%, and on 6 November the stock closed well below its opening price. That is why the AI limited the size to six shares.

AXON. Axon lowered its 2026 revenue growth forecast from between 32% and 34% to about 30%, and its free cash flow forecast from 450 million USD to 350 to 400 million USD. The sell counts on pressure holding after the revision. After a sharp fall a strong rebound is possible, and positive news about deliveries could change the picture.

DKNG. DraftKings expects adjusted EBITDA of 175 to 275 million USD for 2026, held back by spending on Predictions. Its core business stays profitable on this measure, and for 2027 the company expects 800 million to 1 billion USD of total adjusted EBITDA. The sell counts on the current spending weighing on the share price, while the rest of the business holds up.

Portfolio Targets and Stop Levels

The total size of the five trades is about 10,081 USD at the reference prices, well within the 20,000 USD the account can hold at the maximum leverage of 1:2.

If all five Take Profit levels are reached and the portfolio is not closed early, the profit comes to about 1,007 USD, or 10.1% of the deposit, before trading costs. In practice the early exit at 850 USD would close the portfolio first. If all five Stop Loss levels are hit exactly at their prices, the loss comes to about 319 USD, or 3.2%. After company reports and major news the price can jump past a stop, so the actual loss can be larger.

Four dates carry most of the risk. Every release is on the economic calendar.

Target-to-Stop Ratio of the Trades

The ratio compares the distance to the target with the distance to the stop. AMD leads at about 3.6 to 1. The other four trades sit close to 3 to 1: DKNG 3.1, ABNB 3.0, NET 3.0 and AXON 3.0. Three of them in detail:

• AMD, about 3.6 to 1. From the reference entry of 449.04 USD the target at 485.04 USD is 36 USD away and the stop at 439.04 USD is 10 USD away. Seven shares give 252 USD of potential gross profit against 70 USD of planned risk. This trade depends most on the mood in chip stocks.
• DKNG, about 3.1 to 1. From the reference sell price of 18.51 USD the target at 16.21 USD is 2.30 USD away and the stop at 19.26 USD is 0.75 USD away. For 80 shares that is 184 USD of potential profit against 60 USD of risk. After a 21.3% fall in one day a strong rebound is possible, and the short distance to the stop makes an early exit more likely.
• AXON, about 3 to 1. From the reference sell price of 313 USD the target at 271 USD is 42 USD away and the stop at 327 USD is 14 USD away. For five shares that is 210 USD of potential profit against 70 USD of risk. The result depends on the decline after the report continuing.

AI Portfolio vs the US 500

To check how well the AI built the portfolio, we compare it with a simple alternative: about the same total size in the US 500, 1.4 lots bought at the same moment as the portfolio. The entry price goes up here after the open. The US 500 position closes together with the portfolio. If the AI portfolio closes early at its 850 USD limit, the US 500 result is taken at the same moment, when both are open for trading.

Dividends count on both sides: dividend adjustments on the US 500 and any dividends on the stocks in the portfolio.

AI Portfolio Result After 10 Trading Days

The result is published on this page when the portfolio closes, on 20 November 2026 or earlier if the closing rule is triggered, together with a screenshot of the account history and the outcome of every position.

What This AI Portfolio Tests

The fourth AI portfolio is the first to trade in both directions. It buys AMD, Cloudflare and Airbnb and sells Axon and DraftKings, betting that the market keeps sorting companies by their latest reports. The main tests are US inflation data on 10 and 13 November and the NVIDIA report on 17 November.

After three rounds the AI portfolios are up 61.58 USD in total, and the US 500 positions are down 613.31 USD. On 20 November this page will show whether single stocks in both directions keep the AI ahead of the index, and the score after four rounds.

FAQ