Markouts and reaction times are two methods to measure how competitive the trading in an instrument is. Here I want to introduce two others which are closely linked to latencies but add a bit more context.
The first is the "missed" quantity as a fraction of the traded volume. Suppose there are 100 lots/contracts/shares on the top level and multiple participants fire off a 100 lot IOC (or FOK) order. Only the first one will get filled. All others "miss". This can be interpreted as the aggressive competitiveness.
The second is the fraction of unintended fills. Suppose you have a resting order in the book. The market changes and you send off a cancel (or modify) request. But while this request is in flight (or possible even before you sent off the request) your order was filled by the matching engine. You cancel request was hence "late" and your fill was not intended. I refer to this as the passive competitiveness.
Fig. 1 below shows these KPIs for (a) the DAX40 constituents, (b) a very liquid DAX ETF, (c) the DAX future trading on Eurex (FDAX), and (d) the DAX index options on Eurex (ODAX).
- The 40 index constituents form a tight cluster. They are "oversubscribed" by only 20% on average and late cancels only comprise a tiny fraction of 2% of the traded volume. Especially the latter number is a bit surprising given that the passive markouts for the cash equities are more negative than those for the future. This probably reflects a compositional difference between participants passively trading cash equites vs. those that provide liquidity in the future.
- Note the marker for the options in the very bottom right corner. The complete absence of late cancels has two reasons: First, there is a "speed bump" for aggressive orders on Eurex for these options. Reasonably sophisticated (i.e., fast) participants should always be able to pull their orders before any aggressor can pick them off. Secondly, for technical reasons most market makers use "quotes" for options (vs. normal "orders" for other security types). It is impossible for the exchange to unambiguously determine whether or not a fill for a quote was intended.
- Despite the speed bump the DAX options exhibit the highest aggressive competitiveness - even more than the future. One hypothesis is that, counterintuitively, the speed bump could even be helpful for low-latency participants; since other HFTs will always be able to pull their quotes, whatever volume is left must be from a non-HFT.
- Other than the special case of the options, there is a progression of competitiveness from cash equities over the ETF to the future.
- The distribution of IOC misses and late cancels is very inhomogeneous. The median trade has zero IOC missed quantity and no late cancels.
- There is no double counting for the IOC misses. I follow the methodology used for the IOC Liquidity Indicator for Options. For example, if the same participant has multiple misses for the same opportunity then only the largest missed quantity is counted (and not the aggregate).
COO Orionx | Quant Trader | Complexity Junkie
1 年Hey Stefan, great article. Quick question, how can you measure passive competitiveness without knowing when other players sent cancel requests? Is this data even accesible?