Let Your Winners Ride

All top traders know the way to profit is minimizing losses and maximizing gains.  Cut your losers quick and let your winners run.

This investment philosophy is also especially true for execution algorithms. How you build your algos and/or which algorithmic provider you choose is going to have a significant impact on the overall performance of your trades. If your provider is not publishing aggregate statistics of performance, buyer beware. It is one thing for a vendor to measure results, and it is another thing altogether to publish results that outperform standard benchmarks. 

In order to optimize for performance, at very extensive cost, we built and maintained very low latency, high throughput systems to consume market data and place orders on exchange with the least possible latency. 

Our algorithms use embedded price predictive signals to get more aggressive when the market is moving adversely and less aggressive when it is moving in your favor.  

When a buy order is being executed, if it predicts the price of the asset will drop, the algorithm will be more patient and try to buy at a much lower price. If it predicts the market to run higher, it will get more aggressive.  The goal is to minimize the bad events while allowing positive outcomes the opportunity to be captured.

As a result, CoinRoutes performance is generally positive, with very low slippage costs in the aggregate.  It is very easy from the data to see the impact of the algorithms managing the tail risk of large losses while improving conditions for large gains.  

The boring but important stuff. We evaluated trades over a six week period. This report includes hundreds of different assets, in a random distribution. Please note, results are higher when limiting data analysis to major currencies only. In this case, we decided to leave alts and alt-perps included, so give a more complete sample.  The  benchmark we used is often referred to as “arrival”, where we snapped  the best price (both inside and mid) on the aggregate of many relevant exchanges at the time the parent order was submitted.  We filtered results to exclude all orders with performance in excess of a 10% deviation from the midpoint.

The results were striking:  For perpetual swaps, the algorithms returned over 10 Basis points of positive performance and for spot orders, the outperformance was 6 basis points.   The details can be seen in the following screenshots from our reporting engine:

In order to prove this effect is consistent across time periods, below we display reports for different periods. You’ll notice different dates but similar results.

The following screenshot shows performance data for January 1st through the 14th.  For perpetual swaps, the sample that included large outliers outperformed by almost 9 basis points and spot trades by a smaller amount -- less than .4 of a basis point.

Next, we look at the middle period from Jan 11 thru the 24th, where including the outliers up to 10% outperformed on perpetual swap orders by just under 6 basis points and on spot orders by 1.5 basis points.

Lastly, we look at the period from Jan 25th through February 7th and the pattern is the same.  Including all outliers up to 10% showed outperformance on perpetual swaps by over 17 basis points, while on spot it delivered roughly 9.5 basis points of outperformance.   

Such consistent results could only happen if there was a consistent bias in the data where significant positive performance is more common than negative performance.  

Despite market fluctuations, different volatility regimes, volume profiles, global macro events, and whatever changes markets can experience day to day and hour to hour in a six week period, CoinRoutes’ returned consistent, significant outperformance of large orders. 

We can attribute this consistent outperformance to one persistent factor - our award winning, proprietary algorithms. CoinRoutes’ approach to trading algorithms using our low latency technology stack, and our embedded short term price predictive signals, produces consistent and statistically significant outperformance.