Current industry algorithms are stale productivity tools with opaque heuristics and little focus on order placement or liquidity optimization. BestEx Research is analyzing every element of your trading and shaving off every excess basis point. And our unique broker-neutral model allows you to use our high-performance algorithms while you trade with the brokers you choose.
We optimize every aspect of execution for each instrument’s market structure—trade planning, short-term price prediction, order placement, and order routing. Each child order our algorithms place or cancel—size, price, and venue—is based on a rigorous quantitative framework, backtested and re-optimized over thousands of simulations.
Most algorithms have a serious design flaw—timing actions by the clock when the market’s memory is in volume. A trade can take a few minutes in a small cap stock or one second in a large cap stock. A heuristic that cancels and replaces limit orders every 10-15 seconds will create extreme information leakage for a mid or small cap stock, while achieving very few passive fills.
A simplified limit order book evolving over 25 seconds. In each 5-second interval, the right side represents the back of the queue and the left side is the front. Here, 10 seconds was not enough time for a limit order in this stock to move to the front of the queue and execute. Similarly, 10 seconds was not enough time for an order placed at midpoint to be executed.
Our algorithms make every decision—including cancel and replacement of limit orders or pausing after a fill—in stock-specific volume time, reducing information leakage and increasing passive fill rates.
This illustration shows simulated passive fill rates when BestEx Research trades portfolios of stocks selected randomly from the indices shown, with order sizes ranging from .25-5% of each stock’s average daily volume.
Want to know exactly how it works? Contact Us
Most algorithms use simple heuristics to decide when to place orders and at what price, leading to high spread costs, especially in mid and small cap stocks.
The illustration above shows the average spread in baskets of stocks selected randomly from the indices shown.
The heart of our algorithms is our multi-step fill probability model that prices each limit order as an option based on historical and real-time liquidity characteristics of the stock you’re trading, yielding higher spread savings. And it doesn’t matter whether you’re using a VWAP strategy or any other algorithm we offer.
Our simulated performance versus interval VWAP for randomly selected portfolios from each index, with orders ranging from .25-5% of each stock’s average daily volume. Performance vs. interval VWAP measures spread costs because an order’s market impact is incorporated into the VWAP price. BestEx Research algorithms save customers 75% of the total spread, on average.
Want to know exactly how it works? Contact Us
Most algorithms use aggressive logic only to catch up on a trade plan without accounting for varying depth and spread, resulting in higher costs for taking liquidity.
This image illustrates spreads widening and narrowing throughout the trading day. Overly simplistic aggressive logic executes market orders instantly as needed to get back on schedule, risking execution when spreads are at their widest and yielding high spread costs.
Our algorithms monitor every quote change and utilize forecasts of spread, depth, and short-term alpha to take liquidity opportunistically, significantly minimizing your spread costs and temporary market impact.
BestEx Research forecasts spreads and depth to decide whether waiting to take will reduce spread costs. If spreads are likely to tighten, we’ll wait to take liquidity when the time is right, creating substantial savings.
Want to know exactly how it works? Contact Us
While passive orders earn spread, they also tend to fill when prices are moving in your favor. Improved prices after your fills are referred to as “adverse selection” and the cost of adverse selection can be much higher than a half spread.
The price of this stock is moving down over time. This example shows adverse selection costs exceeding spread costs; prices are declining rapidly during your buy order, by more than the size of the half spread in the limit order book.
Our algorithms do what market makers do, canceling limit orders when our proprietary short-term alpha model points to a high probability that the price will move in your favor.
This illustration shows limit orders to buy and sell in the limit order book for a large cap stock. Our short-term alpha model detects the thinning liquidity at the NBBO and below, and our algorithms will cancel those orders and replace with new limit orders deeper in the order book.
Want to know exactly how it works? Contact Us
One stock-specific detail most algorithms ignore is relative tick size. Because the minimum tick size is one cent, highly liquid, low-priced stocks (e.g. SIRI) often have long queues. Waiting only a few seconds or trades before canceling and replacing a limit order would never allow for passive fills in these stocks. Most algorithms don’t account for this and pay the spread when their short wait times expire.
Time between trades for four example stocks. SIRI is a common example of a long-queue stock, for which traders must wait several times longer than is typical to execute limit orders. Limit orders placed at the bid for GOOG and MSFT are likely to execute more quickly.
Our algorithms handle long-queue stocks—like SIRI—by layering orders in the book intelligently to optimize passive executions. We also deviate from the schedule as needed to accommodate passive fills and optimize venue selection based on queue times.
Our algorithms place limit orders (shown in green) according to stock-specific queue lengths, layering orders to optimize passive fills. Above, we show example order books for SIRI and GOOG—with two very different behaviors. For SIRI we place multiple orders at the same price, but for GOOG we place orders that will execute quickly while earning wider spreads.
Want to know exactly how it works? Contact Us
Most algorithms pass each child order to a router that returns a fill, separating routing decisions from trade decisions and treating every child order independently. This practice eliminates opportunities to customize venue selection to the time available for waiting on a queue or optimizing distribution of limit orders across venues.
This illustration ranks venues on two dimensions, each X marking stock-specific real-time queue length and fill rate at an exchange. While a router might use this information to route a single limit order, isolating child orders does not yield globally optimal routing strategies when multiple child orders are placed.
Our algorithms control every detail of order placement through each child order’s entire lifespan, more effectively matching child orders to a wide range of venues selected according to real-time fill rates and queue lengths and other metrics like adverse selection.
Here we depict optimal routing over two dimensions (queue length and fill rate), sending a variety of orders to appropriate venues based on urgency. Our routing technology optimizes over these and other dimensions of your trading to reduce spread and adverse selection costs.
Want to know exactly how it works? Contact Us