# Investment memo — relative strength rotation

**To:** Elena Vasquez, Aurora Capital  
**From:** Aurora Capital trainee (worked example)  
**Date:** August 2026  
**Subject:** Capstone — NVDA vs SPY 20-day relative strength

## Executive summary

I tested a simple rule: **hold NVIDIA when its prior 20-day return beats SPY’s; otherwise hold SPY**. On Jan 2024–Jul 2026 data the strategy grew £1 to about **£2.26** (+126%), with annualised volatility ~**36%**, Sharpe ~**0.95**, and max drawdown ~**−38%**. Buy-and-hold NVIDIA did much better on return and Sharpe (**+318%**, Sharpe ~**1.31**) with a similar worst drawdown (~**−37%**). I would **not** recommend this rule over buy-and-hold NVDA for a growth-oriented sleeve — it cut risk a little and gave up a lot of upside.

## Hypothesis

When NVIDIA’s recent relative strength exceeds the market’s, it will continue to outperform; when it lags, rotating into SPY reduces drawdowns without destroying long-run return.

## Rules (exact)

1. Each day compute trailing 20-trading-day returns for NVDA and SPY.  
2. If NVDA’s 20-day return > SPY’s, signal = hold NVDA; else hold SPY.  
3. Trade on **yesterday’s** signal (one-day lag; no same-day look-ahead).  
4. No shorts, no leverage, no transaction costs modelled.

## Results

| Metric | Relative strength | Buy-and-hold NVDA |
| --- | ---: | ---: |
| Total return | +126% | +318% |
| Ann. volatility | 36% | 49% |
| Sharpe (rf 4%) | 0.95 | 1.31 |
| Max drawdown | −38% | −37% |
| Days in NVDA | ~59% | 100% |

Chart: `outputs/charts/module5_capstone.png`.

## Limitations

- One bullish AI regime; relative strength often meant “stay long NVDA.”  
- No costs, taxes, or capacity constraints.  
- SPY is a crude defensive proxy (still equity beta).  
- Parameters (20 days) were not stress-tested across a grid without a hold-out plan.

## Recommendation

Reject for live capital as a replacement for NVDA buy-and-hold. It is a useful **teaching** example of a clean backtest with drawdown reporting. If the committee wants lower NVDA concentration, prefer an explicit weight cap or the Module 2 Sharpe mix over this rotation rule.