AI is a bubble

Verdict: Wait — Compelling Backtest, But Thesis Mismatch and Thin Evidence

The idea's thesis is that AI is a bubble, yet the supplied rule set is entirely long-only momentum — buying pullbacks when price sits below the 50-day EMA while RSI holds above 65 and ADX exceeds 25. That internal tension is the defining feature of this trade: the 60-month NVDA backtest produced roughly 528% across just 6 trades with a 67% win rate, which is an extraordinary return, but it reflects a trend-following system riding the AI surge, not a contrarian short. The strongest argument against is that those results rest on a dangerously small sample where exits were filled on daily bars rather than intrabar data, meaning the 2% stop and 4% take-profit levels are approximate and likely overstate execution quality — particularly for a stock like NVDA that can gap several percent in a session. Additionally, no tested evidence is provided for the other four tracked symbols (PLTR, SMCI, AMD, MSFT), so the multi-ticker diversification benefit is entirely unquantified. No robust parameter setup was established; the optimization budget was exceeded without producing a recommendation. With market state disconnected, none of the four simultaneous entry conditions can be confirmed as live today. **Conviction breakdown:** - **Thesis support (35):** The narrative (AI is a bubble) is logically incompatible with the long-only momentum rules that produced the backtested returns. - **Trade readiness (15):** Live market state is disconnected, so none of the required entry conditions can be evaluated for any symbol. - **Risk quality (40):** The 2% stop offers tight per-trade control, but the 29% max drawdown, daily-bar fill approximation, and unquantified cross-ticker correlation leave material risk unaddressed. - **Backtest evidence (45):** Six trades on a single symbol over five years is too sparse for statistical confidence; the 67% win rate should be treated as an upper bound. - **Fundamentals trend (50):** No fundamental data was supplied for any of the five tracked tickers, leaving valuation and earnings context absent.

Trade now

The thesis says AI is a bubble, but the strategy rules are built to go long on overbought momentum — buying when price is below the 50-day EMA, ADX (14) exceeds 25, RSI (14) exceeds 65, and price tags resistance. That internal tension means the backtested returns (528% on NVDA over 60 months across 6 trades at a 67% win rate) reflect a trend-following framework, not a contrarian short. The two are philosophically incompatible, and you should treat the quant layer as a separate trade idea from the narrative. No parameter-sensitivity recommendation was established — the optimization timed out without producing a robust nearby-parameter setup, so no variant is being applied to the live strategy before publication. Position sizing is fixed-risk at 2% per trade with a maximum position of 20% of equity and a $100 minimum. The hard exit framework is tight: a 2% stop loss and a 4% take profit, meaning every triggered trade targets a 2:1 reward-to-risk ratio. A time-based exit also closes any position held for 60 bars (roughly three trading months). Live market state is not currently connected, so I cannot show you the real-time distance between NVDA's current price, its 50-day EMA, RSI (14), or ADX (14) readings versus each entry threshold. Without those live values, there is no actionable trigger to evaluate today. "Wait" in this context means exactly that: do nothing until the live indicator feed reconnects and you can confirm…

Scores

• Conviction score breakdown: 37
• Thesis support: 35
• Trade readiness: 15
• Risk quality: 40
• Backtest evidence: 45
• Fundamentals trend: 50

Watch items

• NVDA — RSI (14)
• NVDA — ADX (14)
• NVDA — Price vs 50-day EMA
• NVDA — RSI (14)
• PLTR — RSI (14)
• SMCI — RSI (14)
• AMD — ADX (14)
• MSFT — Price vs 50-day EMA