Live paper portfolio
RSI(14) mean reversion, long only
Across 763 S&P 500 and FTSE 350 companies, RSI was the only one of the four indicators to post a positive out-of-sample Sharpe ratio after costs (0.48). The other three are trend-following, and all three went negative net of what it cost to trade them. RSI still lost to buy & hold, which is why this portfolio publishes both side by side. It is being traded live because a rule that survived one honest test deserves a second one that cannot be re-run.
Read this before the numbers. Every session shown so far was reconstructed when this book was opened on 23 August 2026, running back to 20 February 2026. Each ran through the identical daily sequence — decide on the close, fill at the next open — so no future information touched any decision. But none of it was published in advance, so treat it as a backtest until forward recording accumulates. Days from here are appended one at a time and are never recomputed.
This is a paper portfolio. No real money is at risk, and nothing here is investment advice. Orders are decided on the close and filled at the next session’s open, paying 8 basis points per side — the same cost model as the research.
Read the exposure figure alongside the return. The portfolio finished 0.5% behind buy & hold while holding an average of just 32% of the book in the market. Matching a fully-invested benchmark from a third of the exposure is a different result from matching it fully invested — better on a risk-adjusted basis, and worse if what you want is compounding.
Track record
Portfolio value against buy & hold
Committed, not yet filled
Orders standing for the next open
Published before they execute. This is the part of a paper track record that cannot be constructed after the fact — whatever is listed here today either fills at tomorrow’s open or is withdrawn on the record.
No orders standing. The strategy is holding its current book into the next session — which for a mean-reversion rule that spends most of its time in cash is the normal state, not an omission.
Open positions
What the book holds right now
| Ticker | Shares | Avg entry (USD) | Cost (USD) | Opened |
|---|---|---|---|---|
| WMT | 209 | 103.69 | 21671 | 2026-08-21 |
Closed trades
Every completed round trip, with the reason it was taken
13 completed round trips, 9 of the 13 shown profitable. USD 437.71 paid in trading costs across the life of the book.
| Ticker | Opened | Closed | Entry | Exit | P&L | Return | Why it closed | Record |
|---|---|---|---|---|---|---|---|---|
| WMT | 2026-06-01 | 2026-07-30 | 114.46 | 111.34 | -596 | -2.9% | Closing position — RSI 51.5 — no oversold signal | reconstructed |
| CVX | 2026-07-01 | 2026-07-14 | 163.35 | 181.17 | 2263 | +10.7% | Closing position — RSI 57.6 — no oversold signal | reconstructed |
| MSFT | 2026-06-26 | 2026-07-14 | 356.48 | 382.10 | 1477 | +7.0% | Closing position — RSI 51.0 — no oversold signal | reconstructed |
| MCD | 2026-05-11 | 2026-06-10 | 273.97 | 284.32 | 743 | +3.6% | Closing position — RSI 51.9 — no oversold signal | reconstructed |
| JNJ | 2026-04-22 | 2026-05-14 | 224.89 | 229.68 | 407 | +2.0% | Closing position — RSI 52.0 — no oversold signal | reconstructed |
| MCD | 2026-03-25 | 2026-04-20 | 307.97 | 310.05 | 101 | +0.5% | Closing position — RSI 50.2 — no oversold signal | reconstructed |
| PG | 2026-03-25 | 2026-04-20 | 141.61 | 143.89 | 288 | +1.5% | Closing position — RSI 51.4 — no oversold signal | reconstructed |
| MSFT | 2026-02-23 | 2026-04-14 | 393.41 | 386.35 | -384 | -1.9% | Closing position — RSI 51.5 — no oversold signal | reconstructed |
| HD | 2026-03-13 | 2026-04-10 | 339.15 | 338.50 | -71 | -0.4% | Closing position — RSI 50.8 — no oversold signal | reconstructed |
| META | 2026-03-27 | 2026-04-09 | 563.39 | 626.39 | 2110 | +11.0% | Closing position — RSI 53.3 — no oversold signal | reconstructed |
| GOOGL | 2026-03-27 | 2026-04-07 | 277.11 | 302.55 | 1448 | +9.0% | Closing position — RSI 51.0 — no oversold signal | reconstructed |
| AMZN | 2026-02-23 | 2026-03-05 | 208.10 | 215.98 | 716 | +3.6% | Closing position — RSI 51.7 — no oversold signal | reconstructed |
| UNH | 2026-02-23 | 2026-03-03 | 284.70 | 285.13 | -2 | -0.0% | Closing position — RSI 50.2 — no oversold signal | reconstructed |
The rules
Exactly what this portfolio does
- Buy when 14-day RSI closes below 30 (oversold).
- Hold until RSI recovers through 50, then sell.
- Never short. The research tests long/short; this book trades the long leg only, because stock borrow costs are real and this project does not model them.
- Equal weight across every live signal, capped at 20% of the book in any one name.
- Orders are decided on the close and filled at the next session's open.
Universe
20 US large caps, fixed at inception and never changed: AAPL, AMZN, COST, CSCO, CVX, GOOGL, HD, JNJ, JPM, KO, MCD, META, MSFT, NVDA, PEP, PG, UNH, V, WMT, XOM. US only — FTSE names are quoted in pence, and mixing currencies without an FX model would make the portfolio value wrong.
Why long only
The research tests every indicator long and short, because that is how the rule is defined. This book trades the long leg only. Shorting incurs a stock borrow fee that varies by name and by day, and this project does not model it — so a live short book would report a return nobody could actually have earned.
What would falsify this
If the portfolio finishes behind buy & hold on a risk-adjusted basis over a full year of forward recording, the honest conclusion is that the Week 1 result was noise. That conclusion will be published here in the same place as everything else.
Audit it. The complete order book is published as raw data: every fill, every daily mark, and the derived figures shown on this page. Positions are never stored — they are replayed from the fills, so the trade log is the only source of truth.