Real-World Test of PairScan on an Unpredictable Market: Positive Accumulation

PairScan's premise is simple: find pairs that mean-revert so you can buy the spread when it's wide and sell when it narrows. But how does that hold up when the market itself is unpredictable — sudden spikes, flash crashes, regime shifts? You can't trust a backtest that assumes yesterday's volatility will repeat. This article walks through a real-world test of PairScan's screener during a period of high uncertainty.

Why Most Pair Screener Tests Are Misleading

A typical test picks a pair, runs a backtest over the last year, and reports a Sharpe ratio. That's not useful. The market isn't stationary — volatility clusters, correlations break, and the same pair that worked for six months can fail the next week. PairScan avoids this by requiring a Hurst exponent below 0.5 (mean-reverting, not trending) and an ADF p-value below 0.05 (stationarity) on a rolling window. But even those filters don't guarantee future performance. Where this fails: a pair can pass both tests today and still blow up tomorrow if the underlying relationship changes due to a fundamental event (e.g., a fork, a regulatory crackdown, or a liquidity crisis).

Setting Up the Real Test

You run PairScan's screener daily, scanning a universe of crypto pairs (e.g., SOL/XRP, ETH/BTC, or tokenized equities like AAPL/MSFT on xStocks). The screener outputs pairs that meet the mean-reversion criteria. You then paper-trade the top 3 pairs by the position-in-range metric: bottom zone (spread > 1 std below mean), mid zone (spread near mean), top zone (spread > 1 std above mean). The idea: enter when the spread is extreme, exit when it reverts. No leverage, no compounding — just raw spread moves.

Observations Over a Chaotic Month

During the test period, the market experienced a flash crash (BTC dropped 15% in an hour) and a subsequent recovery. Here's what happened qualitatively:

• Pairs with high correlation broke down temporarily. For example, ETH/BTC, normally mean-reverting, became trending during the crash. PairScan's Hurst filter correctly flagged it as non-mean-reverting (Hurst > 0.5) and removed it from the list. You avoided a losing trade.
• Some pairs stayed mean-reverting. SOL/XRP maintained its spread behavior. When the spread hit the top zone (XRP outperforming SOL), you short the spread (short XRP, long SOL). The spread reverted within 2 days. Not a huge win, but consistent.
• False signals from low-liquidity pairs. A tokenized metal pair (e.g., GOLD/SILVER) showed a false bottom zone signal because of a one-time liquidity gap. The spread didn't revert; it widened further. PairScan's walk-forward backtest would have caught this if you had checked the recent out-of-sample performance, but in real time you took the signal. This is where the walk-forward component matters: it tests the strategy on data the model hasn't seen, reducing overfitting. Still, no filter is perfect.

What Positive Accumulation Means

Despite the false signal, the cumulative P&L from all signals over the month was positive. Not because every trade won, but because the winners were larger than the losers on average. The key metric: the average reversion speed (how many days until the spread returns to the mean) was shorter than the holding period for losing trades. This is a property of mean-reversion systems — they have a positive expectancy if the Hurst and ADF filters are strict enough. But don't mistake this for a free money machine. The drawdowns can be sharp, especially if you size too aggressively.

Practical Takeaways

• Use the screener daily, not weekly. Pairs can pass or fail filters overnight. PairScan's daily scan is essential.
• Check the walk-forward backtest report on pairscan.io/methodology. It shows how the pair performed on unseen data. If the out-of-sample Sharpe is negative, skip it.
• Diversify across uncorrelated pairs. Don't put all capital on one pair. If you have 3-4 pairs from different sectors (crypto, equities, metals), the portfolio is more robust.
• Position sizing matters. Use the position-in-range zones to determine entry: bottom zone for long spread, top zone for short spread. Avoid mid zone entries — they have no edge.

PairScan's methodology is sound, but it's a tool, not a crystal ball. The real test shows that positive accumulation is possible even in unpredictable markets, provided you respect the filters and manage risk. If you want to run your own test, the free tier on pairscan.io/screen gives you daily signals for a handful of pairs. Start small, track everything, and see for yourself.