Unlock This Note

Same lopsided order flow, opposite trades: follow it on GOOGL, fade it on AMZN. This strategy trades the GOOGL and AMZN stock perpetual contracts on Binance and watches one number: Blave's Taker Intensity, the standardized gap between market buy and market sell orders. Half the capital goes to GOOGL, where it follows strong flow. The other half goes to AMZN, where it fades extreme flow. We re-ran the listed source code ourselves: April to October 2026 on 1-hour bars, +22.4% total return, Sharpe 4.22, and a max drawdown of only −2.9%. Holding both names half-and-half over the same period returned about +22.8%. The edge is in the drawdown, not the return. What is Taker Intensity, and why trade the two names in opposite directions? Taker Intensity measures how many traders are willing to cross the spread. Positive means market buys dominate; negative means market sells do. An extreme reading can mean someone is building a position, or just a chase, a panic, or a wave of liquidations. The seller's observation is that the same order-flow signal behaves differently on two highly correlated tech names: on GOOGL it acts like momentum, so the strategy goes with it; on AMZN extremes tend to reverse, so it goes against them. The two legs trade independently and are not a hedge. The rules: four thresholds per leg At every 1-hour close the strategy reads Taker Intensity over a 24-hour window. GOOGL: above +2.5 go long, close the long below +0.5; below −2.5 go short, cover above −0.5. AMZN: below −2.8 go long against the flow, close above −0.6; above +2.8 go short against it, cover below +0.6. Each leg gets half the capital, and most of the time it sits in cash: GOOGL is in the market about 12% of the time (mostly short), AMZN about 10%. Backtest performance (Apr – Oct 2026, Binance GOOGLUSDT / AMZNUSDT perpetuals, 1-hour bars) MetricValueTotal return+22.4%Annualized47.6%Sharpe4.22Max drawdown−2.9%Trades79Buy and hold, half each, same periodabout +22.8% Includes a 0.05% fee per side, Binance's standard taker rate on perpetuals. Backtest results, not a guarantee of future performance. What we checked before listing it We ran the quality gates on the listed code, not on numbers the seller supplied. Look-ahead: re-running at five truncation points produced identical signals, so it uses no future data. Significance: this is a two-asset portfolio, and the Monte Carlo permutation test only applies to a single instrument, so it was not run. Parameter robustness: shifting both entry thresholds two steps either way (0.25 per step) gives an average Sharpe of 3.72 across the neighbouring settings, 0.88× the chosen one. That is a plateau, not a spike. Data the listed backtest didn't use: the AMZN contract listed on February 9, 2026, almost two months before the backtest starts. Running the AMZN leg alone over that stretch with a simplified calculation gives −3.1%, while holding AMZN returned +2.2%. Four things to know before you run it First, the sample is six months, and the only out-of-sample stretch lost money. The thresholds were set on the same data they are tested on; the two earlier AMZN months are the only check outside it, and they were negative. Expect live results well below a 4.22 Sharpe. Second, it did not beat buy and hold: +22.4% against about +22.8%. What you get is a small drawdown and a lot of time in cash, not extra return. Third, the gains are concentrated: July alone made +12.3%, about half the total, and it has been roughly flat since September. Fourth, these are Binance perpetuals, not shares. They trade around the clock, weekends included, and you need a Binance futures account; a brokerage account won't work. Taker Intensity here is the order flow on these contracts, not Nasdaq trading in the stocks. Holding a perpetual means paying or receiving funding; using Binance's historical rates, the effect over this period was about −0.3%. The rules are simple. Running both legs every hour is the work. The logic is all above. What takes time is pulling Taker Intensity for both names every hour, checking each leg's thresholds and placing orders without missing a bar. Blave Agent runs that chain for you: buy this strategy in the library for 3 TWD, then tell your Agent to deploy it. See the GOOGL × AMZN Taker Intensity strategy in the library →