A detailed exposé on how the market is rigged from a data-centric approach • Hacker News Board

Re: A detailed exposé on how the market is rigged from a data-centric approach

#61

Post by k__ » Wed, Jul 16, 2014, 11:55 AM UTC

axanoeychron wrote: post #3

You cannot defend frontrunning of a market. If I ask for X at Y. Someone else shouldn't have the facility to buy it based on my own trade signal and try sell it back to me. It is mindblowingly simple theft. The arguments for liquidity do not hold. There is some fascinating cognitive dissonance when it comes to the HFT industry.

It doesn't sound like anything different from "normal" markets is happening.

You are selling stuff, someone or some people buys a big bunch of it, you think it sells good an rise the price later, now the someone has to buy the stuff for a higher price.

The only difference is the time between the first buy and the second buy for a higher price.

Even if you send your buy-orders at the same time to two different sellers, they don't have to arrive at the same time and if the second seller has "heard" about the first sell, he has enough time to rise the price.

Re: A detailed exposé on how the market is rigged from a data-centric approach

#62

Post by downandout » Wed, Jul 16, 2014, 11:58 AM UTC

throwaway283719 wrote: post #38

What is happening here is really quite simple, and doesn't deserve an entire blog post. There are two exchanges, A and B, and a market maker Jill is quoting (say) 10,000 shares on each of those two exchanges for $17. Big institutional trader Jack sees the 20,000 shares and decides that he wants to buy 15,000 of them, so he sends two orders for 7,500 shares each to A and B. Because of various effects (network latencie…

2. The only reason that Jill has a speed advantage over Jack is because she has paid for it! She has paid to co-locate her server at the exchange, and she has paid to use high-speed connections between exchanges. Are we going to declare that paying for a competitive advantage is suddenly immoral?

Here's the problem with that: the number of available ultra-close connections to the market is finite. If you carry this out to its only possible conclusion, whomever has the closest connection always wins, and everyone else always loses. The other market participants eventually realize that it is simply not possible for them to win, and that a closer connection is not for sale at any price, so they simply stop participating. This solves one problem - people stop losing money - but also destroys the market.

Re: A detailed exposé on how the market is rigged from a data-centric approach

#63

Post by ripb » Wed, Jul 16, 2014, 12:00 PM UTC

willvarfar wrote: post #59

All exchanges should have synced clocks and all messages should have a timestamp up to 2 seconds in the future when they will be published by each exchange. The buffering would be internal to each exchange and not shared with anybody. You can only cancel after what you are cancelling is published. This would allow everyone to make all exchanges publish at once so people with fast cable between exchanges can't beat ou…

>This would allow everyone to make all exchanges publish at once so people with fast cable between exchanges can't beat out those that don't.

Why is one actor paying for an advantage that is available to anyone who should desire it, and have the means to pay for it, an issue?

Re: A detailed exposé on how the market is rigged from a data-centric approach

#64

Post by asgard1024 » Wed, Jul 16, 2014, 12:06 PM UTC

throwaway283719 wrote: post #48

rtpg wrote: post #43

Earlier quoted context omitted.

>2. The only reason that Jill has a speed advantage over Jack is because she has paid for it! She has paid to co-locate her server at the exchange, and she has paid to use high-speed connections between exchanges. Are we going to declare that paying for a competitive advantage is suddenly immoral? If we want an efficient market,we need perfect information. Information asymmetry creates inefficient markets. The moral…

If you want a reasonably efficient market, you need some participants to have close to perfect information. There is no market anywhere in the world that is 100% efficient, because the costs of getting to efficiency are prohibitively high. It's like trying to reach the speed of light - you can expend more and more effort getting closer and closer, but you can never actually reach it. I'm not saying that what we have…

> If you want a reasonably efficient market, you need some participants to have close to perfect information.

Is this proven somewhere or you just assume the optimal strategy for markets is continuous?

I mean, it's not clear that the optimal strategy for "slightly imperfect markets" is at all close to the optimal strategy for markets with perfect information. And I actually doubt it can be proven, in the general case.

Re: A detailed exposé on how the market is rigged from a data-centric approach

#65

Post by throwaway161803 » Wed, Jul 16, 2014, 12:06 PM UTC

This article establishes that two things often happen shortly after you place an order to buy shares: 1) Another trader places a similar order on another exchange. 2) A large number of outstanding sell orders are cancelled.

It's not clear to me that either of these are Bad Things, deontologically speaking. [I don't recall Jesus mentioning them.]

The key question is consequentialist: Are there regulatory changes which would improve the lot of the average investor, investing through, say, an index tracker or pension fund? For each potential change, one ought see how it fares w.r.t. this standard, considering, to the extent that it is possible, the induced second-order effects.

Talk of theft, rigging, fairness (you don't owe people like me anything), stolen goods and frontrunning is only useful to the extent that it helps us converge on an answer to this question. These words are tools that we have developed for analysing more familiar situations, where they correspond to actions which are clearly harmful.

Most changes proposed here either lose market efficiency directly (trade buffering / increased tick sizes) or just give us new games to play (if the market clears once a second, we will get our orders in last), potentially resulting in a less direct loss. The question remains.

Re: A detailed exposé on how the market is rigged from a data-centric approach

#66

Post by rahimnathwani » Wed, Jul 16, 2014, 12:26 PM UTC

throwaway283719 wrote: post #38

What is happening here is really quite simple, and doesn't deserve an entire blog post. There are two exchanges, A and B, and a market maker Jill is quoting (say) 10,000 shares on each of those two exchanges for $17. Big institutional trader Jack sees the 20,000 shares and decides that he wants to buy 15,000 of them, so he sends two orders for 7,500 shares each to A and B. Because of various effects (network latencie…

Jack is not upset because he couldn't buy the shares at the price he wanted. He is upset because someone was offering shares at a specific price, and Jack was willing to pay that price, but the order was not executed. The reason the order was not executed is not because someone else accepted the offer before him, or because Jill cancelled before he tried to accept. It was because Jill was able to see his acceptance in transit and cancel part of her offer as a result of this new information. This isn't how markets are supposed to work. If Jill had posted a single offer for half the number of shares that Jack wanted, it would be different.

Someone shouldn't need to use Thor or some other delaying mechanism to accept open offers. The latency between different exchanges (which was exploited in this example) does nothing to increase market efficiency.

You're right that I shouldn't care about this as a practical matter, as the impact on me is very small, but that doesn't mean it's right.

I really appreciate the clarity in your comments on this thread.

Re: A detailed exposé on how the market is rigged from a data-centric approach

#67

Post by throwaway283719 » Wed, Jul 16, 2014, 12:26 PM UTC

downandout wrote: post #62

throwaway283719 wrote: post #38

What is happening here is really quite simple, and doesn't deserve an entire blog post. There are two exchanges, A and B, and a market maker Jill is quoting (say) 10,000 shares on each of those two exchanges for $17. Big institutional trader Jack sees the 20,000 shares and decides that he wants to buy 15,000 of them, so he sends two orders for 7,500 shares each to A and B. Because of various effects (network latencie…

2. The only reason that Jill has a speed advantage over Jack is because she has paid for it! She has paid to co-locate her server at the exchange, and she has paid to use high-speed connections between exchanges. Are we going to declare that paying for a competitive advantage is suddenly immoral? Here's the problem with that: the number of available ultra-close connections to the market is finite. If you carry this o…

It's empirically not true that "whomever has the closest connection always wins and everyone else always loses" as is evidenced by the fact that there are multiple competing market makers who are all profitable.

Arguments of the kind "let's carry this to its logical conclusion" are almost always fallacious, because they ignore limiting factors, or alternative explanations.

If your only advantage is speed then you need to have the fastest connection to the exchange, else your business model doesn't exist. If you have other advantages, then speed is less important. Nowadays there are very few market makers whose only advantage is speed, because most of them realized that continually paying through the nose to compete on speed is a mug's game.

Re: A detailed exposé on how the market is rigged from a data-centric approach

#68

Post by growse » Wed, Jul 16, 2014, 12:29 PM UTC

downandout wrote: post #62

throwaway283719 wrote: post #38

What is happening here is really quite simple, and doesn't deserve an entire blog post. There are two exchanges, A and B, and a market maker Jill is quoting (say) 10,000 shares on each of those two exchanges for $17. Big institutional trader Jack sees the 20,000 shares and decides that he wants to buy 15,000 of them, so he sends two orders for 7,500 shares each to A and B. Because of various effects (network latencie…

2. The only reason that Jill has a speed advantage over Jack is because she has paid for it! She has paid to co-locate her server at the exchange, and she has paid to use high-speed connections between exchanges. Are we going to declare that paying for a competitive advantage is suddenly immoral? Here's the problem with that: the number of available ultra-close connections to the market is finite. If you carry this o…

This is obvious for any market that has a single physical location. People who stand next to the apple seller get local apple price information faster than those standing in the next town.

You've also got a very peculiar definition of winning. A person who wishes to buy 10,000 Ford shares who places an order at $17 only to find that in the meantime the market has shifted to $17.01 and therefore purchases at that price hasn't "lost". They set out to buy Ford stock at market rate, and that's what they ended up doing.

Re: A detailed exposé on how the market is rigged from a data-centric approach

#69

Post by rahimnathwani » Wed, Jul 16, 2014, 12:29 PM UTC

ryanjshaw wrote: post #55

throwaway283719 wrote: post #38

What is happening here is really quite simple, and doesn't deserve an entire blog post. There are two exchanges, A and B, and a market maker Jill is quoting (say) 10,000 shares on each of those two exchanges for $17. Big institutional trader Jack sees the 20,000 shares and decides that he wants to buy 15,000 of them, so he sends two orders for 7,500 shares each to A and B. Because of various effects (network latencie…

> But Jack has no god-given right to be able to buy shares at the price he likes best It seems like the disagreement really lies here. I'm not a finance expert so I'll probably get a few things wrong but is it fair to summarize the two perspectives as follows? 1. Jill is merely quoting a price for independent blocks of shares on independent exchanges. If a buy order is placed against that quoted price, she has the ri…

Whether she has to honour the quote is irrelevant, as she yanks the quote at the second exchange before it is hit.

Re: A detailed exposé on how the market is rigged from a data-centric approach

#70

Post by throwaway283719 » Wed, Jul 16, 2014, 12:33 PM UTC

rahimnathwani wrote: post #66

throwaway283719 wrote: post #38

What is happening here is really quite simple, and doesn't deserve an entire blog post. There are two exchanges, A and B, and a market maker Jill is quoting (say) 10,000 shares on each of those two exchanges for $17. Big institutional trader Jack sees the 20,000 shares and decides that he wants to buy 15,000 of them, so he sends two orders for 7,500 shares each to A and B. Because of various effects (network latencie…

Jack is not upset because he couldn't buy the shares at the price he wanted. He is upset because someone was offering shares at a specific price, and Jack was willing to pay that price, but the order was not executed. The reason the order was not executed is not because someone else accepted the offer before him, or because Jill cancelled before he tried to accept. It was because Jill was able to see his acceptance i…

Here is our point of disagreement, then - I think that this is exactly how markets are supposed to work (in the presence of multiple exchanges).

The job of a market maker is to supply liquidity at a price/risk tradeoff that is reasonable to them, subject to the information available to them. If there are multiple exchanges, and someone trades with them on one exchange, then the set of information available to them has changed (specifically, their knowledge of the supply/demand balance for a particular stock has changed). It's only natural that they will want to change their prices in response.

Now, we could change legislation to either (a) go back to having a single exchange or (b) restricting the ability of market makers to move their quotes on one exchange if they trade on another. But that won't necessarily result in a better deal for non-market makers, because instead of quoting 20,000 shares split across 4 exchanges, the market maker now quotes 5,000 shares on 1 exchange.

The benefit is that all market participants have a more accurate idea of the true liquidity available in the market. The disadvantage is that you have removed the element of competition between exchanges, so the exchange is no longer incentivized to offer low fees and keep improving its service.