The Market's Queue Value
Suppose you have been standing in a line all night to get your hands on the latest ACME Corporation widget. Now there is a rumor spreading that the anvil is not as good at catching Road Runner's as initially anticipated.
Should you give up your spot in the queue on this rumor? Probably not. If it turns out to be wrong you gave up your highly desirable spot for naught. If, however, you just joined the queue things might be different. The queue position has optionality value.
The same principle holds for passive orders in a price-time priority queue. All else being equal, the market maker should be more willing to cancel orders at the back of the queue than orders in the front.
And Fig. 1 shows that this is indeed being applied by market makers. The color represents the likelihood of an order being canceled on the next tick. The x- and y-axes represent the order's queue position; they are the number of orders ahead and behind in the queue, respectively. We only consider the BBO level. The probability is normalized by the number of occurrences of each queue position.
We see that going vertically upwards (i.e., the longer the queue behind is), the less likely it becomes that an order will be canceled. Interestingly, there is much less variation in the horizontal direction. An additional order behind is more significant than an extra order ahead. If anything, an additional order ahead reduces the cancellation probability.
领英推荐
Fig. 1 is for the EURO STOXX50 future on Eurex (FESX) but the same plot for the Bund future (FGBL) is almost indistinguishable. Both instruments have similar volatility-adjusted tick sizes and hence microstructure.
Figure 2, however looks markedly different. It shows the same data for the S&P500 future (E-mini, ES) on the CME. Here we observe that the queue ahead and behind are similarly valued (gradient from top left toward bottom right).
One possible explanation for the different weight of the queue length ahead between the two instruments is the (partial) private-first principle - the first few orders being filled receive the private fill message before the trade will be received via the public feed.
Another oddity is the speck in the lower left corner which is also present e.g. for ZN, the 10-Year T-Note future). Not exactly sure what it going on there.
High-Frequency Market Making
3 个月I think you answered your own question about the high cancel probability in the lower-left when you cited the private-first principle on CME. If you're at the front of a small queue, there's not much information there to be gained by getting an early fill (or perhaps more accurately, the information isn't that valuable).
Head of Quantitative Strategies at LaSalle Trading Group, LLC | MSFM at UChicago
3 个月This is super interesting. Do you have the same visualizations for the 10 Year Futures?
Product Owner at Sentillia B.V.
3 个月The "private-first principle", is not a good thing for the overall market, correct? It necessitates the use of canary orders which adds to overall load and it might unfairly advantage participants with the rate limits to sustain these signalling orders. Would it not be preferable for the public feed to always be faster than the private feed?
Quant Researcher @ Geneva Trading
3 个月Is the bottom left indication of spoofing going on in ES?