How to Use PairTrade AI Finder: Reading Z-Score Signals and Backtests Step by Step

Market makers, HFT algorithms, and hedge funds often don't care which way the market moves. They profit from mathematical inefficiencies using market-neutral strategies, and one of the most important is statistical arbitrage (pair trading).

PairTrade AI Finder by PairTradingPRO takes over all the calculation routine. The algorithm analyzes thousands of asset combinations, finds pairs with high historical cointegration, and delivers ready-made Long/Short setups with exact capital allocation, Z-Score, and a quick backtest.

🤖 Quick Answer

What is PairTrade AI Finder?

It is an AI-powered screener for pair trading and statistical arbitrage on crypto futures. The algorithm searches for cointegrated pairs: assets whose spread has historically returned to its mean (mean reversion). For each pair, the screener shows:

• Log Spread — the spread formula with asset weights;
• a ready-made signal (LONG SPREAD, SHORT SPREAD, NEUTRAL);
• capital allocation between the legs of the trade;
• Z-Score — how far the spread has currently moved from its norm;
• P-Value from the Dickey-Fuller test;
• correlation and its stability;
• backtest results: Profit Factor, Win Rate, average trade duration, average PnL, and annual return.

Calculations use hourly candles over 360 days. Only pairs with a correlation of 0.7 or higher and a cointegration significance level of 0.05 make it into the results.

🎯 Correlation vs. Cointegration: The Beginner's Trap

Pair trading is based on the mean reversion hypothesis. You look for two historically linked assets, wait for an abnormal divergence in the spread between them, and bet that they will converge again.

The biggest trap is confusing correlation with cointegration. Most traders trade correlation, while quant funds trade cointegration.

🐕 The Drunk Owner and the Dog

• Correlation is like two random drunk passers-by walking down the street in the same direction. They stagger in a similar way, but either of them could turn into an alley at any moment, and the connection would be gone for good.
• Cointegration is like a drunk owner walking a dog on a retractable leash. The owner staggers, the dog chases cats, runs ahead, or falls behind. But no matter how far the dog runs, the leash eventually stretches tight, and they have to come back together. The distance between them is a stationary time series.

📋 Comparison Table

💡 How the Profit Works

When the "leash stretches" and one asset becomes unjustifiably expensive relative to the other, you short the overvalued one and buy the undervalued one. When the leash pulls back, you take the profit. The overall market direction matters much less: the two legs hedge each other's directional risk.

⚙️ Calculation Settings and Interface

📊 Top Panel

• Timeframe 1h — calculations use hourly candles.
• Window 360 days — the analysis window: 360 days of history.
• Correlation ≥ 0.7 — only pairs with a correlation of 0.7 or higher are shown.
• Coint. α 0.05 — the significance level of the cointegration test: the P-Value must not exceed 0.05.
• Updated — the time of the latest recalculation.
• Detailed guide — a link to the detailed instructions.

🎛 Filters

📲 Signals in the Telegram Bot

In the top-right corner of the screener there is a Subscribe to the signals button. Click it, and new screener signals will start arriving in the PairTradingPRO Telegram bot. You won't miss an entry point, and you won't need to keep the screener open all the time.

Once you are subscribed, the button changes to Unsubscribe from signals, which turns the notifications off.

🧩 Every Screener Column Explained

The table has four blocks: BASIC, COINTEGRATION, CORRELATION, and BACKTEST.

1️⃣ BASIC: Pair, Signal, and Capital Allocation

Log Spread — the spread formula in log prices, with asset weights. For example:

−0.5827 × WIFUSDT.FT + 0.4173 × OPUSDT.FT

Log prices let you compare assets with very different price levels, and the coefficients set the proportion in which the legs balance each other.

The .FT suffix marks a futures contract. Futures are used so you can open short positions: you can't build a delta-neutral pair without a short leg. On the spot market you can't sell an asset you don't own, while on futures a short is as easy to open as a long.

Category — the pair's market (Crypto in the screenshot).

The signal — a ready-made trading recommendation:

Capital allocation — the specific action for each leg and its share of capital. The AI takes into account the historical volatility of both assets so that the more volatile coin doesn't skew the risk and the position stays truly delta-neutral.

Example for the WIF/OP pair with a STRONG SHORT SPREAD signal:

• LONG WIFUSDT.FT — 58.3% of capital;
• SHORT OPUSDT.FT — 41.7% of capital.

The shares match the coefficients in the Log Spread formula. "Selling the spread" with a minus sign in front of WIF means buying WIF and selling OP.

2️⃣ COINTEGRATION: Entry Triggers

Z-Score — how many standard deviations the spread has currently moved from its historical norm. The indicator bar shows the direction and strength of the deviation.

• Positive Z (WIF/OP: +3.36) — the spread is above the norm; a decline is expected.
• Negative Z (1000PEPE/VET: −3.28) — the spread is below the norm; a rise is expected.
• Around 0 — the spread is in equilibrium.

Max Z↑ / Z↓ — historical maximum deviations upward and downward. They show how far the spread can go before it returns. For example, 1000PEPE/VET has a Max Z↓ of −8.33: the current Z of −3.28 is far from the historical extreme, so the spread could theoretically diverge further. This matters for risk calculation and margin buffer.

P-Value — the confidence indicator from the Dickey-Fuller test. A value below 0.05 means the relationship is statistically significant, not random. The lower the P-Value, the more confident the test: FIL/DOT shows 0.0037, while OP/SEI sits at exactly 0.0500, right on the threshold.

Status — the test result: Cointegrated means the pair passed the cointegration test.

3️⃣ CORRELATION: Protection Against Decoupling

Why not just trust the current correlation? It may have formed yesterday on a single piece of news. So the screener digs deeper:

• Correlation — how closely the assets are moving together right now.
• Average — the average correlation over a longer period.
• Min (labeled "Mines." in the interface) — the minimum correlation over the period. This is a key defensive metric: a deep drop in correlation in the past shows the pair is prone to chaotic decoupling. For 1000PEPE/VET the minimum fell to 0.4179, while SOL/BNB held at 0.7327.
• Stability — the overall verdict based on the three metrics above: Stable or Moderate. A green Stable badge is an extra filter against false signals.

4️⃣ BACKTEST: Mathematical Proof of the Strategy

You don't have to rely on blind hypotheses: the system has already tested how the pair would have traded in the past.

• Deals — the number of trades in the backtest. The more trades, the more statistically significant the sample.
• Profit Factor — gross profit divided by gross loss. A value above 1.5 is considered excellent: with a Profit Factor of 2.0, the algorithm earned $2 for every $1 it lost.
• Win Rate — the share of profitable trades. For mean reversion strategies, it often exceeds 70–80%.
• Transaction time — the average time capital is locked in the pair, from the Z-Score breaking the threshold to its return to zero (about 6.5 to 12.5 days in the screenshot). It helps manage portfolio liquidity.
• Avg PnL — the average return per closed trade, in percent.
• 1Y Return — the strategy's total return for the pair over a year. Based on the screenshot data, it roughly equals the average trade profit compounded over the number of trades: for WIF/OP, +2.10% per trade over 25 trades gives 53.14% per year.

📈 How to Read a Signal: An Example

Let's look at the VETUSDT.FT / XRPUSDT.FT row:

This is an example of a signal where several factors line up at once: an active deviation, a low P-Value, and a strong backtest.

⚠️ What to Consider Before Entering

• Small trade samples. The backtests in the screenshot have 17 to 35 trades per year. Very high values on such samples (for example, a Profit Factor of 17.86 for FIL/DOT) should be treated with caution: the result may not repeat.
• Multiple testing effect. When an algorithm tests thousands of pairs at α = 0.05, some pass the cointegration test purely by chance. Pairs with a P-Value right at the 0.05 threshold need extra checks.
• Economic logic. Statistics can find a relationship but can't explain it. Pairs with a clear fundamental basis (for example, SOL/BNB, two major Layer 1 ecosystems) are usually more reliable than random combinations.
• Structural breaks. Cointegration can break after news, listings, delistings, or tokenomics changes. That's exactly why the Min and Stability metrics exist.
• Futures costs. Account for trading fees on both legs and funding rates for the entire holding period.
• Margin buffer. The historical Max Z shows the spread can move much further than the entry point. Position size and leverage must be able to handle that divergence.

🛠 Step-by-Step: How to Use PairTrade AI Finder

1. Find the anomaly. Filter for active signals (LONG or SHORT SPREAD) and pick pairs with the best backtest, stable correlation, and a strong current Z-Score deviation.
2. Check the quality. Benchmarks: a P-Value well below 0.05, a Profit Factor above 1.5, enough trades, and a high minimum correlation.
3. Evaluate the pair in Spread Builder. Look at the spread chart: how evenly it oscillates around the mean and how quickly it returns after divergences. Then check the equity curve: the strategy's returns should grow smoothly, without deep drawdowns or sharp drops.
4. Open the hedge position. Use the Capital allocation column and open Long and Short orders at the same time, sticking strictly to the specified capital shares.
5. Take profit. Wait for the close signal, when the Z-Score returns to zero: this means mean reversion has occurred. Close both legs at the same time.
6. Automate. Set the pair and your entry and exit conditions in the PairTradingPRO desktop terminal so the robot watches the Z-Score and executes trades on its own.

🧭 Where AI Finder Fits in the PairTradingPRO Ecosystem

🏁 Conclusion

PairTrade AI Finder gives traders analytics and backtesting tools that used to be available mainly to hedge funds, in an intuitive interface:

• finding cointegrated pairs among thousands of combinations;
• ready-made signals with exact capital allocation;
• Z-Score and P-Value to judge entry timing and reliability;
• monitoring of correlation stability;
• backtests with Profit Factor, Win Rate, and annual return;
• signals delivered straight to a Telegram bot.

Stop fighting the market and getting stopped out by unpredictable Bitcoin moves. Sign up, launch the screener from your account, and find your first math-driven trade.