Measured performance: how accurate are Cashora's stock signals?
Across 30 US stocks and ETFs over ten years of daily data, Cashora's actionable signals were directionally correct 58.8% of the time over 2,289 positions, returning an average of +1.52% per 20-trading-day position with a mean Sharpe ratio of 0.59. On the 15 tickers held back from all tuning, the hit rate was 57.3% with a mean Sharpe of 0.38 and 12 of 15 tickers profitable. Measured 2026-09-18.
How these numbers were produced
Every signal is generated walk-forward: on each historical bar the engine sees only the data available up to that bar, never future prices. A position is entered at that day's official closing price and exited at the close 20 trading days later. Only one position is held at a time and it is fully allocated, so compounded returns cannot double-count overlapping trades.
Only signals the live app marks actionable are counted: long calls only, taken only while the price is above its 200-day average, and only at a confidence score of 60 or higher. Short calls were measured at roughly 45% accuracy and are shown in the app as information only, never as actionable.
Fifteen tickers (AAPL, MSFT, TSLA, NVDA, AMZN, GOOGL, META, JPM, XOM, KO, WMT, AMD, NFLX, COST, UNH) were used to choose those rules. The other fifteen were held back entirely and measured once, afterwards — those are the honest numbers, and they are lower.
Price history: daily split-adjusted opens, highs, lows and closes from Yahoo Finance, 2016-09-19 to 2026-09-18, 2,514 bars per ticker.
Held-back tickers (out of sample)
| Ticker | Positions | Hit rate | Avg / position | Sharpe | Max drawdown | Compounded | Buy & hold |
|---|---|---|---|---|---|---|---|
| QQQ | 92 | 70.7% | +1.18% | 0.75 | -27.1% | +153.80% | +495.70% |
| SPY | 92 | 69.6% | +0.56% | 0.52 | -18.2% | +56.70% | +233.60% |
| CRM | 71 | 63.4% | +1.57% | 0.62 | -35.8% | +130.50% | +233.10% |
| ORCL | 78 | 60.3% | +1.02% | 0.39 | -52.4% | +54.40% | +258.90% |
| V | 90 | 60.0% | +0.48% | 0.31 | -45.2% | +33.40% | +366.10% |
| MA | 93 | 57.0% | +1.01% | 0.72 | -38.0% | +126.90% | +447.10% |
| GE | 65 | 56.9% | +2.65% | 1.04 | -32.5% | +330.40% | +105.50% |
| HD | 75 | 56.0% | -0.21% | -0.10 | -50.6% | -34.00% | +119.80% |
| SOFI | 34 | 55.9% | +3.92% | 0.75 | -41.0% | +121.20% | -0.60% |
| PFE | 62 | 53.2% | +0.30% | 0.16 | -39.9% | +4.90% | -11.60% |
| BA | 65 | 52.3% | +0.87% | 0.33 | -39.5% | +33.90% | +26.10% |
| DIS | 56 | 50.0% | +0.44% | 0.21 | -29.0% | +9.30% | -1.00% |
| INTC | 65 | 47.7% | +1.16% | 0.25 | -71.5% | +5.00% | +191.10% |
| T | 53 | 47.2% | -0.16% | -0.10 | -43.9% | -15.00% | -18.30% |
| CVX | 76 | 44.7% | -0.18% | -0.09 | -46.8% | -28.80% | +79.60% |
Tuning tickers (in sample — expect these to look better)
| Ticker | Positions | Hit rate | Avg / position | Sharpe | Max drawdown | Compounded | Buy & hold |
|---|---|---|---|---|---|---|---|
| AAPL | 84 | 67.9% | +2.43% | 1.31 | -24.1% | +531.80% | +1064.50% |
| MSFT | 83 | 62.7% | +1.56% | 0.91 | -30.3% | +210.30% | +686.50% |
| TSLA | 77 | 57.1% | +4.70% | 0.91 | -56.5% | +1048.90% | +2652.10% |
| NVDA | 89 | 68.5% | +5.34% | 1.53 | -50.4% | +5520.40% | +9541.80% |
| AMZN | 84 | 64.3% | +1.37% | 0.65 | -33.3% | +148.30% | +553.40% |
| GOOGL | 86 | 65.1% | +2.33% | 1.17 | -34.7% | +492.80% | +758.00% |
| META | 83 | 54.2% | +1.02% | 0.38 | -56.2% | +59.50% | +457.40% |
| JPM | 82 | 63.4% | +1.43% | 0.88 | -29.4% | +179.30% | +309.50% |
| XOM | 69 | 50.7% | +0.86% | 0.41 | -32.9% | +50.10% | +75.80% |
| KO | 78 | 56.4% | +0.55% | 0.52 | -21.9% | +44.90% | +110.30% |
| WMT | 83 | 59.0% | +1.15% | 0.75 | -18.9% | +129.00% | +348.20% |
| AMD | 84 | 57.1% | +4.84% | 0.99 | -49.0% | +1715.50% | +5082.00% |
| NFLX | 81 | 59.3% | +1.69% | 0.55 | -45.8% | +138.70% | +480.30% |
| COST | 86 | 60.5% | +1.21% | 0.76 | -21.9% | +145.40% | +460.40% |
| UNH | 73 | 52.1% | +0.32% | 0.15 | -56.2% | +1.40% | +133.20% |
Where the model underperforms
CVX, T and HD lost money per position out of sample. INTC and T finished below a coin flip on hit rate. On most large-cap names, simply buying and holding the stock over the same decade beat the signal portfolio — NVDA is the extreme case (+9,542% buy-and-hold versus +5,520% traded). Signals are a probability tool for timing research, not a substitute for owning good businesses.
What these numbers are not
Hypothetical, back-tested results. They do not include commissions, slippage, taxes, dividends or financing costs, and no Cashora user achieved them. Past performance does not indicate future results. Cashora is a market probability signal predictor operated by Thomas Solutions LLC — it is not a broker, it executes no trades, and nothing here is investment advice.