Banning Some HFT Strategies Helped This Market Grow: Chart(bloomberg.com)
bloomberg.com
Banning Some HFT Strategies Helped This Market Grow: Chart
http://www.bloomberg.com/news/articles/2016-03-08/banning-some-hft-strategies-helped-this-market-grow-chart
7 comments
I've also wondered the same thing. I've always been replied to with "it provides liquidity to the market" but I still don't see why anybody needs microsecond level liquidity. Seems like it provides marginal benefit over a minute or hourly quantized exchange at tremendous cost (to society).
Ordinary traders obviously don't need microsecond level liquidity. But having microsecond level liquidity allows more market makers to compete with each other to provide liquidity. This competition lowers the prices ordinary traders pay for liquidity.
TLDR: Compounding effects that add up over time, and extremely dubious use of the words "tremendous cost (to society)."
HFT is a tax on the markets (and by extension society).
I mean...that's kind of a hugely controversial point to just say as a fact. It also completely ignores the fact that HFTs employ a lot of different strategies, some of which are uncontroversially beneficial to the market (market making).
According to this older HN posted article the way to slow down HFT strategies is get rid of the subpenny rule, such as pricing securities in increments of $0.0001
https://www.chrisstucchio.com/blog/2012/hft_whats_broken.htm...
https://www.chrisstucchio.com/blog/2012/hft_whats_broken.htm...
I'm not sure that this would slow down HFT, but it might cut costs for consumers. It would definitely be a good idea to have different tick sizes for different products as it's odd that the most liquid products and the least liquid products might both have a 0.5bps tick size in the US.
If you do reduce the tick size you'll need to make some changes to internalization to reduce the toxicity of lit venues.
If you do reduce the tick size you'll need to make some changes to internalization to reduce the toxicity of lit venues.
How can that be enforced? Why wouldn't secondary markets just spring up allowing people to sell in between?
The problem isn't HFT per se it's adverse selection. Large trades are more likely to happen if the exchange discriminates in who participates in the exchange, not how they participate. Regulation NMS destroyed upstairs trading in the US and is effectively making such discrimination illegal/impossible unless you operate a dark pool which cannot openly show prices.
What if all stock market would adopt a strategy that orders are processed one per minute. Same way it works during opening/closing market.
Some advantages:
1. All assets will be as liquid as they are now. Still got opportunity to buy/sell 8 * 60 = 480 times per day.
2. Most of the HFT industry will be redundant. Instead of allocating super smart people on made up problem, they can solve more important social problems (e.g. robotics, curing cancer, aging, space exploration, high growth tech companies).
3. Less day trading, encourage long term perspective.
Some advantages:
1. All assets will be as liquid as they are now. Still got opportunity to buy/sell 8 * 60 = 480 times per day.
2. Most of the HFT industry will be redundant. Instead of allocating super smart people on made up problem, they can solve more important social problems (e.g. robotics, curing cancer, aging, space exploration, high growth tech companies).
3. Less day trading, encourage long term perspective.
I am an active participant on an exchange that functions in this way.
As others have mentioned, this structure does not negate the benefits of speed.
Someone else already asked what you would do if there are more buyers than sellers (or vice-versa) in auction. If you're going to allocate randomly instead of using time-priority this is equivalent to pro-rata allocation in expectation. There are already pro-rata markets in the US and elsewhere.
This structure does increase transaction costs because I am less willing to make a tight market if I have to hold risk for such a long period of time.
> Most of the HFT industry will be redundant. Instead of allocating super smart people on made up problem, they can solve more important social problems (e.g. robotics, curing cancer, aging, space exploration, high growth tech companies).
It seems to me that you are overestimating the profits generated by HFT and the number of people employed in this industry.
The above is just my probably incorrect opinion.
As others have mentioned, this structure does not negate the benefits of speed.
Someone else already asked what you would do if there are more buyers than sellers (or vice-versa) in auction. If you're going to allocate randomly instead of using time-priority this is equivalent to pro-rata allocation in expectation. There are already pro-rata markets in the US and elsewhere.
This structure does increase transaction costs because I am less willing to make a tight market if I have to hold risk for such a long period of time.
> Most of the HFT industry will be redundant. Instead of allocating super smart people on made up problem, they can solve more important social problems (e.g. robotics, curing cancer, aging, space exploration, high growth tech companies).
It seems to me that you are overestimating the profits generated by HFT and the number of people employed in this industry.
The above is just my probably incorrect opinion.
This doesn't have the advantages you think it has.
1) Speed is still critically important as you have every incentive to wait until the very last microsecond before the end of the 1 minute window to submit any orders.
2) What do you do if there are more buy orders than sell orders at a given price for a given batch? Who's orders get filled?
1) Speed is still critically important as you have every incentive to wait until the very last microsecond before the end of the 1 minute window to submit any orders.
2) What do you do if there are more buy orders than sell orders at a given price for a given batch? Who's orders get filled?
2) Wouldn't that sort itself out since you now only get information once a minute?
So it would still be first come first serve. Or am I missing someting?
So it would still be first come first serve. Or am I missing someting?
You only get information directly from the market-center once a minute, but the rest of the world (and other markets which trade continuously, or at a different frequency) are generating new information constantly. Clever and/or well-capitalized participants will be taking advantage of that information in much the same ways as with a continuous market.
In that sense, you can see what the HFT firms are doing (providing liquidity at the microsecond level) as bringing the market more in-line with reality.
Another way to look at it is that the less the market resembles reality (through regulations, such as larger defined processing intervals) the more it can be taken advantage of. Rules often provide sharp edges where there was previously a continuous curve, and sharp edges are where you can game the system.
Another way to look at it is that the less the market resembles reality (through regulations, such as larger defined processing intervals) the more it can be taken advantage of. Rules often provide sharp edges where there was previously a continuous curve, and sharp edges are where you can game the system.
I am trying to understand how this reflect reality. If milliseconds gives us HFT wouldn't minutes get rid of the advantage of doing HFT? I.e. the reality is determined by it being milliseconds, seconds, minutes etc. So we can control which reality we want.
As far as I understand HFT, it's not that it's better at predicting the market but rather it's better at taking advantage of micro volatility?
Or am I completely wrong?
As far as I understand HFT, it's not that it's better at predicting the market but rather it's better at taking advantage of micro volatility?
Or am I completely wrong?
Depending on how queued orders are served at whatever interval the market coalesces at, there will always be some type of high speed trading, even if the frequency isn't as high. In reality (as in any transaction between two individuals), there's no defined interval at which we wait to complete our transactions, we just do them. Adding arbitrary times at which transactions go through creates edge cases, and edge cases can be gamed. Current markets are limited by either an arbitrary interval upon which orders are fulfilled, or as fast as a system can handle them, which is limited by the technology in use.
For a very simplistic example of this, if you look at some type of transaction that has extra regulatory requirement/burden that kick in on certain criteria (for example, over a certain price), you should see some interesting behavior when looking at transactions immediately above and below this price level. You'll also likely find interesting solutions to circumvent that rule to capitalize on the reduced competition that allows (to my understanding, this is a semi-accurate high level overview of a part of the derivative mess we found ourselves in a few years back).
> As far as I understand HFT, it's not that it's better at predicting the market but rather it's better at taking advantage of micro volatility?
My understanding is that you are correct, but in taking advantage of the volatility, they make it a micro level volatility, and both provide a more accurate value of a stock (by reducing the spread), and at a shorter time interval (which together I guess is liquidity).
Of course that's all armchair theorizing on my part (in case it came across as overly authoritative or knowledgeable). I'm not a trader, don't work in the industry, and am not an economist. This is just the working model in my mind based on the economics as I see them and my knowledge of the markets, which admittedly is mostly from discussion here...
For a very simplistic example of this, if you look at some type of transaction that has extra regulatory requirement/burden that kick in on certain criteria (for example, over a certain price), you should see some interesting behavior when looking at transactions immediately above and below this price level. You'll also likely find interesting solutions to circumvent that rule to capitalize on the reduced competition that allows (to my understanding, this is a semi-accurate high level overview of a part of the derivative mess we found ourselves in a few years back).
> As far as I understand HFT, it's not that it's better at predicting the market but rather it's better at taking advantage of micro volatility?
My understanding is that you are correct, but in taking advantage of the volatility, they make it a micro level volatility, and both provide a more accurate value of a stock (by reducing the spread), and at a shorter time interval (which together I guess is liquidity).
Of course that's all armchair theorizing on my part (in case it came across as overly authoritative or knowledgeable). I'm not a trader, don't work in the industry, and am not an economist. This is just the working model in my mind based on the economics as I see them and my knowledge of the markets, which admittedly is mostly from discussion here...
There will always be some high speed trading but the problem is that once you take it into an territory humans can compete then it's not going to be beneficial for that same High Speed Trading to occur.
It's advantage was that it could do it at a speed humans couldn't do.
While there is no specific speed at which we make trades we do mostly get our information from the market. If the frequency of updates go down I would guess so would the randomnes of when we make our bets. The frequency of the market affects the randomness of the order.
And yes I am really just a Layman although I did do this for Nasdaq :) http://tv.adobe.com/watch/customer-stories-air/nasdaq-market...
It's advantage was that it could do it at a speed humans couldn't do.
While there is no specific speed at which we make trades we do mostly get our information from the market. If the frequency of updates go down I would guess so would the randomnes of when we make our bets. The frequency of the market affects the randomness of the order.
And yes I am really just a Layman although I did do this for Nasdaq :) http://tv.adobe.com/watch/customer-stories-air/nasdaq-market...
Humans aren't competing against HFTs on speed. That's not an accurate view of the world.
Different HFT firms are competing against each other on various bases (one if which is speed) to provide you, a human, the best service.
Different HFT firms are competing against each other on various bases (one if which is speed) to provide you, a human, the best service.
Thats not what I am saying either or at least not what I meant to say.
What I was trying to say was that humans simply can't do HFT and that the advantage of that is only possible if you have millisecond updates. If you remove that then you remove the advantages of HFT and humans would be able to compete.
Whether that is good or bad for the market is another discussion.
What I was trying to say was that humans simply can't do HFT and that the advantage of that is only possible if you have millisecond updates. If you remove that then you remove the advantages of HFT and humans would be able to compete.
Whether that is good or bad for the market is another discussion.
Humans can't compete in the specific thing HFT does, but I'm not sure why humans need to compete there. There's plenty of other areas that humans can compete in (such as analyzing data and market prediction), so it's not like humans are being squeezed out of the market.
I don't think making the market quantized (upon reviewing the Nanex story from last week[1], that's apparently the term) makes the problem of people not being able to compete in this small aspect of the market go away, it will just change its nature. Everyone will make trades immediately before and after the point at which it coalesces, otherwise you are leaking information to the market that might be used against you. This requires fast machines, and we're back where be began, with an arms race to see who can submit the most orders in the shortest time right before that point, and who can analyze the orders fastest immediately after (which is specifically the point I was originally making regarding. "sharp edges" and "continuous lines", since the quantized market is essentially a step function).
1: https://news.ycombinator.com/item?id=11211093
I don't think making the market quantized (upon reviewing the Nanex story from last week[1], that's apparently the term) makes the problem of people not being able to compete in this small aspect of the market go away, it will just change its nature. Everyone will make trades immediately before and after the point at which it coalesces, otherwise you are leaking information to the market that might be used against you. This requires fast machines, and we're back where be began, with an arms race to see who can submit the most orders in the shortest time right before that point, and who can analyze the orders fastest immediately after (which is specifically the point I was originally making regarding. "sharp edges" and "continuous lines", since the quantized market is essentially a step function).
1: https://news.ycombinator.com/item?id=11211093
I feel like the discussion keeps getting turned on it's head now.
The original claim was that if we did by the minutes updated then there would be issues with more buyers than sellers. This was the original claim that was being used as a reason for why a minute frequency would be problematic.
I haven't seen any real explanation for why that is problematic for the reasons discussed in this thread.
Less frequent updates means less frequent available information to trade on, means less advantage for the HFT algos who won't have the advantage of being able to trade in the milliseconds anyoore, something humans can't do.
If one thinks HFT is a problem then thats a perfectly fine solution which isn't from what I can se giving problems of more buyers than sellers.
If one doesn't think it's a problem then sure we don't need to compete with them but at least to my understanding the premise of this whole discussion was that someone suggested slowing down the frequency and the answer was that this would be problematic.
The original claim was that if we did by the minutes updated then there would be issues with more buyers than sellers. This was the original claim that was being used as a reason for why a minute frequency would be problematic.
I haven't seen any real explanation for why that is problematic for the reasons discussed in this thread.
Less frequent updates means less frequent available information to trade on, means less advantage for the HFT algos who won't have the advantage of being able to trade in the milliseconds anyoore, something humans can't do.
If one thinks HFT is a problem then thats a perfectly fine solution which isn't from what I can se giving problems of more buyers than sellers.
If one doesn't think it's a problem then sure we don't need to compete with them but at least to my understanding the premise of this whole discussion was that someone suggested slowing down the frequency and the answer was that this would be problematic.
The main advantage HFTs have over humans is that they can react quickly to new information. The frequency of new information becoming available is only an added bonus, not a requirement.
HFT don't react to the same kind of information that Humans do. HFT reacts to price at the millisecond level not to anything outside of that as far as I understand.
The problem you mention is real, but first come first served is an easy solution. That also solve the order problem.
That encourage you to submit transactions earlier in time slot (as you get more chance to complete them).
Still there is some HFT strategies possible, but window of opportunity will be by an order of magnitude or two smaller.
That encourage you to submit transactions earlier in time slot (as you get more chance to complete them).
Still there is some HFT strategies possible, but window of opportunity will be by an order of magnitude or two smaller.
If orders are filled on a first come first serve basis then you've just recreated the continuous market that we have today and changed nothing.
You removed the HFT advantage of being able to trade at the millisecond no?
No. Because if orders are filled first come first serve if my order arrives 1 millisecond before yours my order will get filled before yours. Just like today with continuous markets.
The continous market delivers continues information which can then be used by HFT companies to take advantage of doing something humans can't. buy and sell in milliseconds.
This is the problem NOT that one trader gets his or her orders in before the other. At least that is how I understand the problem.
This is the problem NOT that one trader gets his or her orders in before the other. At least that is how I understand the problem.
If my order goes in before yours then there is a good chance that my order will get filled and yours won't. The speed at which traders (automated or otherwise) get in orders is of critical importance.
If you don't understand that, then your understanding of the problem is hopelessly wrong.
If you don't understand that, then your understanding of the problem is hopelessly wrong.
That someone get an order in before another will always be a reality. It's not a problem anymore than there are winners and loosers on the stock market.
What we are talking about is specifically removing the advantages of HFT (if one consider it a problem), not removing the fact that someone can get and order in before someone else.
What we are talking about is specifically removing the advantages of HFT (if one consider it a problem), not removing the fact that someone can get and order in before someone else.
How is that different than today? The big guys also have access to information that others in the market don't.
0. Why once a minute? Why not twice? Why not once every 5 minutes? 15? 60? You can still buy or sell eight times a day so once an hour should be plenty.
1. Putting "high growth tech companies" (we know you mean startups) in the "important social problem" bucket, especially in the same group as aging, cancer, space exploration, is tenuous at best.
2. Until any of the above pay half a million a year plus huge bonus potential, I seriously doubt anyone going into HFT would give them more than a passing glance.
1. Putting "high growth tech companies" (we know you mean startups) in the "important social problem" bucket, especially in the same group as aging, cancer, space exploration, is tenuous at best.
2. Until any of the above pay half a million a year plus huge bonus potential, I seriously doubt anyone going into HFT would give them more than a passing glance.
For what it's worth, I am a trader at an HFT firm and my pay is comparable to what my friends in similarly competitive tech roles make. We are talking about the levels of compensation that are always mentioned here and generate discussion about how they do not exist.
I have not heard of similar comp levels in the aging/cancer/rocket space. My suspicion is that this is the result of cultural differences where management takes a much bigger piece of the pie. I would be happy to change fields once this is fixed.
I have not heard of similar comp levels in the aging/cancer/rocket space. My suspicion is that this is the result of cultural differences where management takes a much bigger piece of the pie. I would be happy to change fields once this is fixed.
0. Just to be fair I think any discrete time auction would reduce HFT. No matters less if it will be 100ms or 15 minutes. Just a parameter that you can optimize based on your market.
1. Not all startups are equal, but quite a few will create some lasting value.
2. Unsure if most of this career alternatives will ever give that sort of gains. The question is where the gains come from? Inefficiencies in trading strategies of your retirement fund? Maybe those inefficiencies are always present if humans type orders? Maybe we should design stock market with human first approach, not HFT exploitable?
1. Not all startups are equal, but quite a few will create some lasting value.
2. Unsure if most of this career alternatives will ever give that sort of gains. The question is where the gains come from? Inefficiencies in trading strategies of your retirement fund? Maybe those inefficiencies are always present if humans type orders? Maybe we should design stock market with human first approach, not HFT exploitable?
> Just to be fair I think any discrete time auction would reduce HFT.
But is the reduction of HFT qua HFT really the goal? Making a market low-frequency will by definition reduce/eliminate HFT as a strategy but you will still have a lot of trading activity done by well capitalized companies using lots of expensive equipment. They will just use different strategies.
But is the reduction of HFT qua HFT really the goal? Making a market low-frequency will by definition reduce/eliminate HFT as a strategy but you will still have a lot of trading activity done by well capitalized companies using lots of expensive equipment. They will just use different strategies.
Yeah yeah, its called frequent batch auctions. obligatory mention in every hft discussion: http://faculty.chicagobooth.edu/eric.budish/research/HFT-Fre...
What if every armchair market structure analyst focused their efforts on curing cancer, instead of posting the same poorly researched suggestions in every HFT thread?
I really wish someone would make a version of this form letter for HFT: https://craphound.com/spamsolutions.txt
Why the snarky tone? Why not show how he is wrong so we all can learn something?
Why reward laziness and sloppy thinking? Anyone interested in HFT can peruse a few dozen threads on HN, where these questions have already been answered at length.
Because if you really are annoyed about it perhaps you should explain why it's wrong so it will eventually stop being proposed. People are wrong about all sorts of things, that doesn't mean we need to belittle them or be snarky.
Regarding the snark:
The OP suggests making major alterations to the functioning of equity markets, without any evidence that he's done any research or even thought carefully about the costs and benefits involved. He suggests that the proposed changes are highly beneficial (we might even cure cancer!). How seriously are we supposed to take this sort of thing? Should we also welcome the kind of amateur mathematician who has a simple proof of Fermat's Last Theorem?
Personally, I think a bit of snark is warranted.
The OP suggests making major alterations to the functioning of equity markets, without any evidence that he's done any research or even thought carefully about the costs and benefits involved. He suggests that the proposed changes are highly beneficial (we might even cure cancer!). How seriously are we supposed to take this sort of thing? Should we also welcome the kind of amateur mathematician who has a simple proof of Fermat's Last Theorem?
Personally, I think a bit of snark is warranted.
4. Competing stock markets grow (maybe in another jurisdiction) that allows faster order processing
But faster (continuous-time) order processing results in wider bid/ask spreads; slower (discrete-time) batch processing results in narrower spreads. At least that's what the models show. Follow-up papers even analyze a multi-exchange model (competing continuous-time and discrete-time exchanges), and concluded that investors and retail traders would prefer the discrete-time exchange because it would have narrower spreads (i.e. lower transaction costs).
> At least that's what the models show.
What models? Increasing the amount of risk a market maker must bear will widen the quoted spreads not narrow them since that increased risk needs to be priced in.
What models? Increasing the amount of risk a market maker must bear will widen the quoted spreads not narrow them since that increased risk needs to be priced in.
The models in the "Frequent batch auctions as a market design response to HFT" paper by Budish et. al.; i linked the slides in another comment.
Since batch auctions use single-price clearing, market makers can post bids and asks without fear of being sniped (so its less risk, not more). And since you can't snipe, they have to compete on price rather than investing in faster connections, resulting in narrower spreads.
Since batch auctions use single-price clearing, market makers can post bids and asks without fear of being sniped (so its less risk, not more). And since you can't snipe, they have to compete on price rather than investing in faster connections, resulting in narrower spreads.
It links to a slightly longer article [1], which provides a bit of detail, but doesn't elucidate.
First of all, how do you only allow passive orders? If an order is marketable when it arrives at the exchange, then is it rejected?
Second, how does this resemble IEX's speed bump? I'm very confused.
[1] http://www.bloomberg.com/news/articles/2016-02-03/this-upsta...
First of all, how do you only allow passive orders? If an order is marketable when it arrives at the exchange, then is it rejected?
Second, how does this resemble IEX's speed bump? I'm very confused.
[1] http://www.bloomberg.com/news/articles/2016-02-03/this-upsta...
Basically in their rulebook it says that all prop (as opposed to agency) flow must be passive, and if they find aggressive prop flow from a member they can terminate the membership.
The tie in to IEX is tenuous at best.
The tie in to IEX is tenuous at best.
Good explanation of IEX here: https://www.fastcompany.com/3031242/a-no-nonsense-explanatio...
Well for one it explains why some HFT actors are so vehemently opposed to IEX's HFT-limiting exchange application, as investors leave any venue that still permits HFTs, and go to alternatives such as Aquis and, if its application is granted, IEX.
Comments like these... they just show you have no idea what you're talking about. Investors are perfectly free to trade on the IEX ATS all day long in its current incarnation.
Really, the article is one paragraph.
It quotes Doug Cifu, the head of the largest HFT firm in the world.
I don't know much about Aquis strategy for growing their exchange but if its anything like the aequitas exchange that has just opened in Canada then they get inflated numbers by begin the exchange off record where some banks get paid to report their large block trades.
It quotes Doug Cifu, the head of the largest HFT firm in the world.
I don't know much about Aquis strategy for growing their exchange but if its anything like the aequitas exchange that has just opened in Canada then they get inflated numbers by begin the exchange off record where some banks get paid to report their large block trades.
The article argues it doubled its share of public European stock trading since Feb. 8, when it banned what it considers a problematic high-frequency-trading strategy, just wanting to limit proprietary traders to passively providing price quotes, i.e. banning HFT firms which are parasitic orderflow frontrunners not market makers, and take zero risk which is about 99% of them.
As I commented in another related article, I think it in terms of "nuclear reactors", whose reaction times are crafted to match human reaction times (tens of seconds). Trading could follow the same principle.
Compounding with that, some studies say that trading itself causes most (2/3) of the volatility, so it is expected that µs-scale trading creates even more volatility.
Compounding with that, some studies say that trading itself causes most (2/3) of the volatility, so it is expected that µs-scale trading creates even more volatility.
> so it is expected that µs-scale trading creates even more volatility.
Could you help me understand why this might be expected?
Could you help me understand why this might be expected?
great chart bloomberg
http://tylervigen.com/spurious-correlations
http://tylervigen.com/spurious-correlations
The whole aspect of making updates faster, transactions faster will always lead to those who can afford it looking for an edge in speed and in most cases - getting it, aided or unaided.
But with such dysnimic speed any misdirection in price rise or fall can and has been in the past exhausebated and when you look at the impact again, wonder why so fast.
I can appreciete the fater exchanges operate the mroe transacitons and brokers just love commision, more transactions more commision and it was and always is a crux in any fiscal market.
Indeed the Llyods crash was due to many brokers factoring (watering down the pools of money in effect) risks (groups of insurance poiceys) so many times that when there was a big claim (asbestos) then the whole thing crashed as the pools of money to cover the insurance were eaten away so much by all this over factoring of the risks about with brokers taking a slice everytime it got passed about and reinsurance became a rererererereinsurance style affair that with so much commision taken out it was a financial timebomb awaiting to happen. Again can look more recently at mortgages and how companies polled those together as they do with insurance and reinsurance in much the same way. Though more case of hiding potentual asbestos risks in with less risker mortgages and the rest we all remember.
So can see why I do question the need for faster and faster stock markets when, too me daily would work and indeed even hourly. Would eliminate HFT and indeed make the whole market more stable and focused upon the companies and not what todays market become based upon somebody famous doing or saying something on twitter more then actual results.