forex-scalping.com

Fill quality at the London open versus fill quality at the New York open

Both are busy, both are liquid, and both attract scalpers. The execution character is not the same. London open flow arrives from a wide base of European banks and funds. The New York open mixes US data releases, equity market influence, and a shorter, sharper burst of activity.

In practice, spreads often tighten fastest in London and stay stable for several hours. New York can spike more suddenly, especially around 8:30 data releases, with abrupt liquidity pulls and thin books for seconds at a time.

Fills at the New York open tend to show more variance. The average slippage might be similar to London, yet the tail is fatter, with occasional large misses.

The way to see this is splitting the execution log by session and comparing not just averages but the 95th percentile of slippage. Many scalpers discover that one session fits their tolerance far better than the other.

Choosing the session with the better fit costs nothing and can lift net results.

Which open produces cleaner fills for the pairs traded, and how large was the tail difference?

I split my execution log by session last spring and the result was clearer than I expected. London fills on EURUSD averaged around 0.2 pips of slippage with a 95th percentile of about 0.7, while New York averaged 0.3 but the 95th percentile was nearly 1.6. The averages hardly differed, and the tail was what mattered, as you say.

The worst cases in New York were nearly all within a minute of the 8:30 release, so I tried a simple rule and stopped entering in the five minutes either side of data. After that, the New York tail dropped to about 0.9, which is close to London. I then kept both sessions in my plan but with different handling, with entries at market in London and limit orders in New York.

For GBPUSD the difference between sessions was smaller than I thought, while on USDJPY New York was actually the cleaner one for me. It is worth testing each pair on its own rather than assuming the same answer.