orders distribution vs forward price and trade distribution

[ronin]

If you look at your picture, there is very little that has actually happened in those ten seconds. The market didn't go anywhere, the spread didn't go anywhere. Yes, there are thousands of events, but on the whole they add up to nothing.



Also, it doesn't seem to be very difficult to improve on the best market levels. There is plenty of room inside the spread, and the oscillation of touch levels leaves you enough room to scratch any unwanted fill with some room to spare.



If I was designing a market making strategy for this sort of market, that would be where I would start.

[nikol]

Indeed. It makes sense.

Here is my plain vanilla estimation (in red) based on the cross markets information (my basic algorithm). I still have to remove oscillations.



Blue line is what 'market' is doing.

[img]/User%20Files/694/photo_2018-09-24_17-41-21.jpg[/img]



Of course, I will quote +1 tick better as I see no need to cut the spread all at once.

[ronin]

It makes sense in the middle of the picture. Put in some condition for changing your price level, and calibrate the threshold. You really only want to change when there is a good reason. And good reason can not rely on things like "best bid", "best ask" and "mid". Because somebody can throw one clip inside the spread and manipulate best bid, best ask and mid until the cows come home.



But I don't understand the left and right parts of the picture.



On the left, you are buying at ask and selling a few levels above ask? Fair enough if the ask is just some artificial thing inside the real spread.



On the right, the market has gone up but you have gone down? The right end doesn't make sense.

[nikol]

> But I don't understand the left and right parts of the picture.



My bids and asks are direct function of other markets. On the right: Reds under Blues might indicate that I have an arbitrage (possible fee and wanted margin are not yet accounted). Red lines are my prices, which I do not publish with orders.



If it is an arb, I plan to execute this sequence of actions:

a. wait for hit of my sell limit-order below blue upper line at this market (let say it is market-A),

if yes, then

b. I sell market-order at another market (let's call it market-B)



However, based on your comments this situation might be a trap. While my machine is taking the decision and the signal goes through the wire, my 'faster' opponent can increase his buy limit-order on market-B and I loose money.

[ronin]

Got it.



If you can't make money by crossing two spreads and paying two taker fees, then it does become complicated and it depends a lot on how other traders react to you.



Try a few things and see what works. Also, expect "what works" to change every once in a while as other traders change how they interact with you.