Hacker News Reader

Comment by anonymoushn | Hacker News Reader

It seems like the information being disseminated by the article is just the mapping between the name and the concept, since every participant already knows that they want to be at the front of big levels and can take from the orders behind them if they don't like the fill.

View on HN

Yeah I was hired by a prop trading firm of old Chicago floor traders in 2009 and then shortly asked to come up with an algo strategy in Kospi 200 options, with no help from any experienced algo traders. I worked something out completely on my own which was very similar to 0+, it just follows naturally once you understand the dynamics of a microstructure with low fees and FIFO queues.

They really gloss over "getting to the front of a strong queue," which generally requires stacking to many levels of depth in the market (which adds risk), or finding some way to join or remain in weak queues before they become strong, which requires some other method of estimating a theoretical value. Estimating and tracking depth in the queue at each level was definitely a significant part of the more sophisticated code I wrote.

The article may overstate the importance of lower latency as well. I definitely was not the fastest at the time, running some fairly non-optimized Java when I knew there were some faster C++ players in the market, and I was still able to do relatively well.

Can't you start off by spreading your OCO limit orders 10-11 ticks away from the last price and then just wait it out? Those queues are usually very thin.

Right, so that's "stacking to many levels of depth." To add some color to my point about how it "adds risk," the day Kim Jong-Il died, the market spiked lower and I was instantly lifted on every stacked offer I had in every put. I managed to recover a bit but it still took out about a quarter of a year's PNL in one fell swoop.

Nobody is going to start the next rentec after reading this article, but for people new to the subject I think it is quite an approachable intro to some of the practicalities of market microstructure and how they affect trading strategies.

It's helpful, but even after all the work to be on all the right levels one still has a difficult prediction problem to decide when they don't like a fill. I guess the point is that this prediction problem (conditional on fills at front of big queues, should I dump it?) may be easier to make into a profit because your predictions can be of only moderate quality if your downside is just fees. In crypto d1, the fees are maybe small enough to support this, but other concerns make it difficult. As mentioned elsewhere in the comment section, one may learn that their downside is often not just fees.

Yeah, totally, it really just gives a flavour of how the structure of price levels and price/time priority, tick sizes, etc. can create exploitable probabilistic advantages, and I think many folk are not aware of even this.

The real environment is absolutely messier and more complex than described by this article. Still, it's a layer deeper than "speed is good", to paraphrase Gordon Gekko.

Reply on news.ycombinator.com