Thanks for your effort. I think it's a valuable attempt, but I am skeptical:adamhg wrote: ↑Fri Dec 03, 2021 2:28 pmThis has been bugging me until I realized that I'm actually storing off the current bid/ask for each leg of the box spread at the time the trade is made. So in an attempt to determine which side of the trade the market market is on, I thought maybe we could check whether each leg is closer to the bid or ask and aggregate it into an indicator.comeinvest wrote: ↑Wed Nov 03, 2021 11:43 pmBut with your method of identifying box spreads, you never know if the retail investor SOLD or BOUGHT the spread, do you? My understanding is that in the options market, at least for multi-legged options spreads, opposing client orders are rarely matched, but dealers compete for incoming client orders, and the dealers' bots are programmed to required some spread. One time when I bought one of my boxes back, there was a ca. 0.1% difference in rates (ca. 0.45% vs. 0.55%) when I bought it back vs. the rate that I paid. What I'm saying is, we cannot expect to get a rate similar to a rate from the trades history, as long as we don't know on which side the retail investor vs. the dealer was, can we?adamhg wrote: ↑Wed Nov 03, 2021 9:11 pmLowest somebody got today for Dec 23 was 0.9% ($981 for $1000 wide legs) so that could be a good place to start or just under theretananaev wrote: ↑Wed Nov 03, 2021 7:44 pm I'm also having hard time getting fills. Could be that market rates are going higher because of upcoming fed tapering. Or could be just too many people found out about this trick and that's pushing prices higher.
When you say "current treasury rate", do you look at the 2 year or 3 year treasury? I'm selling Dec 23 options. I'm assuming 2 years, right?
You can kind of see the pattern. The blue (retail investor possibly BOUGHT the spread) is generally under the EMA line whereas the purple (mid point/unknown or retail investor possibly SOLD the spread) is all over the place.
I've read a number of papers that present methods to do this for individual trades, but does this work for spreads as well? Would publishing this change be overall helpful or add too much noise to the chart?
I also took the opportunity to filter out data points that were 2 sigma higher or lower than all previous points. So that should also clean up the charts a bit more. Thanks @parval for calling that issue out
- If I'm honest to myself, I don't see a consistent pattern on the charts of the blue vs the purple circles.
- "So in an attempt to determine which side of the trade the market market is on, I thought maybe we could check whether each leg is closer to the bid or ask and aggregate it into an indicator." - Can you please elaborate how exactly you make the determination? I'm a beginner in this area, but my understanding is that the "implied" trade prices of the legs of exchange-traded complex orders are assigned by the exchange based on some heuristic / theoretical fair values. So looking at the trade prices of individual legs to imply anything makes little sense, does it? I assume you looked at the average proximity or the traded price to the bid or ask over all 4 legs? If so, isn't it a self-fulfilling prophecy that the blue circles are at slightly lower yields on average?
- For most larger spreads, there is no meaningful quoted market of the individual legs.
- Even for smaller spreads with legs closer to the market price and that have higher volume, the traded price of the box is usually much closer to the midpoint / model implied fair value, than to the leg-implied bid/ask of the box. I think we might still extract information, but it's not that easy. Also, consider that often times a customer limit order for a specific leg might skew the midpoint vs if the bid and ask are both dealer quoted. I think pros and algos use interpolated data from a range of strike prices to arrive at a model fair value / midpoint / mark price, not from the bid/ask of one specific strike price.
- In summary, I'm not sure, especially for larger boxes with large leg-implied bid/ask spreads and low volume, if we can assume that the midpoint is the reference point, and/or that the proximity to the bid or ask of the box is meaningful. The actual algo that the bots are programmed with might be asymmetric. Maybe it's more realistic to assume that the clients got the more unfavorable side and the dealers the more favorable side of the trade in a client-dealer trade, i.e. split the circles into two parts. I don't know how dealer-to-dealer trades would be quoted, facilitated, or represented.