Say the public market in a stock has the national best bid and offer (NBBO) priced at $50.00/$50.02. A large size buyer comes into the market and decides that they are prepared to buy up all stock available at any price up to and including $50.05. To do this, they conduct an “inter-market sweep”, where they submit ISO limit buy orders priced at $50.05 to many (if not all) trading venues “simultaneously”. I say “simultaneously” because clearly there are delays in the handling and processing of orders due to both design issues in the execution algos/order management system and the network infrastructure utilized to transmit messages to the various venues.
If you’re an Ultra Low Latency High Frequency Trading (ULLHFT) outfit geared for this, when you see on one exchange an order trading through the NBBO up to $50.05, you can assume (probabilistically) that there is a liquidity sweep being conducted and race the orders to other exchanges. Your objective is to take liquidity at some other exchange at $50.03 and $50.04 and turn around immediately and offer the acquired shares at $50.05. If you’re successful, you’ve managed to buy shares at $50.03 & $50.04 (and possibly even $50.02) and then sold them a cent or more higher with relatively little risk. The only reason that you’d be able to trade this kind of strategy is because of the higher latencies of other market participants.
The risk is if your assumption was wrong and the buyer doesn’t require that much size to fill their initial order. Now you have inventory that you cannot unload.
Long-story short, we prefer to avoid these games. That is why we trade futures markets that have one exchange instead of the fragmented stock market which has at least 13 exchanges.
Remember, not all HFT is the same; for example, we have HFT strategies but simply utilize speed to gain priority in the que (order fills). The alpha is more directional rather than game theory based on other market participants entering orders.
Thanks for the read,