STRATEGY
Specter is a single mean-reversion strategy on liquid US ETFs. Long-only, intraday hold. Daily bars, no high-frequency data, no leverage.
The Signal
IBS (Internal Bar Strength) measures where the close sits within the day's range:
IBS = (Close - Low) / (High - Low)
A close near the day's low (IBS < 0.20) on a liquid ETF tends to bounce. The strategy enters long at the next day's open and exits at the next day's close — a one-day intraday hold (Variant A: MOO → MOC).
Universe
| TICKER | NAME | FILTER |
|---|---|---|
| SPY | S&P 500 | IBS only |
| QQQ | Nasdaq 100 | IBS only |
| IWM | Russell 2000 (Small Caps) | IBS + SMA200 + VIX<30 |
| DIA | Dow Jones 30 | IBS only |
| EEM | Emerging Markets | IBS only |
| EFA | Developed ex-US | IBS + SMA200 + VIX<30 |
| XLF | US Financials | IBS only |
| XLK | US Technology | IBS only |
| XLV | US Healthcare | IBS only |
Per-ticker filter (raw vs filt) chosen on in-sample 2015–2020 Sharpe.
Sizing
- Equal-weight per asset:
1 / 9 ≈ 11.1%of equity per name - Max simultaneous exposure:
≤100%(≤9 entries) - Average historical exposure:
~42%(capital-efficient) - No leverage, no shorting, no overnight beyond the 1-day hold
Cost Model
1 bp slippage per leg (2 bps round-trip) modeled on Alpaca/IBKR retail execution. Commissions $0.
Backtest (Honest)
v3 of the backtest claimed Sharpe 0.72, but assumed close-to-close execution (requires market-on-close orders not all retail brokers offer). v4 corrects to a realistic open→close fill model. We publish the honest number.
Honest Caveats
- 5-year out-of-sample is a small window; the next regime could break the pattern. Edge in retail mean-reversion has been decaying since 2010.
- XLE was post-hoc dropped from the universe after seeing OOS Sharpe 0.07 — that is a form of selection bias.
- The original Pagonidis IBS paper (arxiv 2306.12434, ~75% win rate cited in community discussion) is about country ETFs. Applying the pattern to US sector ETFs is our extension, not academically validated.
- Capacity is small (~$1–5M per ticker before slippage eats edge). Public visibility could shorten that further.
- Strategy is currently PAPER. Live capital activates only after a sustained green window.
Why Publish
The literature says you should hide your edge. The same literature also says most backtests fail in production. Publishing every signal, fill, and revision keeps us honest with ourselves and the few people watching. If the edge survives transparency, it's a real edge. If not, the failure is on the record too.