Imported from previous forum
Hello everyone,
I am currently researching the algorithmic trading market- the key players, what specs are most important when looking at competing products, and most importantly the size of the market.
Would anybody be able to provide some brief input on any of those? Or perhaps refer me to someone I could speak to a resource centre I can look at to learn more about the above-mentioned issues?
Thanks in advance!
regards,
Marta
[ original email was from David GUERINEAU - david.guerineau@etrali.com ]
Dear Marta,
I can recommand a good introduction book: Electronic and Algorithmic Trading Technology - The complete guide from Kendall KIM
You should find most of the answers to your questions.
Regards
david
Hello everyone,
I am currently researching the algorithmic trading market- the key
players, what specs are most important when looking at competing
products, and most importantly the size of the market.Would anybody be able to provide some brief input on any of those? Or
perhaps refer me to someone I could speak to a resource centre I can
look at to learn more about the above-mentioned issues?Thanks in advance!
regards, Marta
Marta,
For a proxy on the algorithmic trading marketplace size you might look at “Program Trading” as defined by the NYSE (using an “old” measurement and a “new” measurement yardstick.) They state: “Program Trading is defined as a wide range of portfolio trading strategies involving the purchase or sale of 15 or more stocks having a total market value of $1 million or more.” Statistics are available at: http://www.nyse.com/marketinfo/datalib/1152267398806.html
At the end of the day you find that depending on how you count it, 30-50% of the daily volume on the NYSE is driven by program trading. How much of that (or even more than that) is properly deemed “algorithmic trading” is an open issue.
I like to think of an algo trades as involving two distinct parties: 1. a Buy-Side customer that is “simply” entering a new algo order on behalf of itself (a fund or LLC/LP) or on behalf of an individual account they are advising, and 2. a Sell-Side supplier that did the up front R&D work on the algo strategy, and built out the execution infrastructure to make it run. That “pair” of entities then makes up one side of an executed algo trade.
In situations where a firm does the R&D itself, builds the execution infrastructure in-house, and is trading only that firm’s own capital – then things get a bit blurred. Are those orders to be counted as algos? Or are they just “pure” program trades? Thankfully, the trading communications standards are the same for both the in-house-entity and the more typical buy-side, sell-side two-independent-entity arrangement.
There are market research groups that occasionally cover algorithmic trading. Tabb Group (http://www.tabbgroup.com/) and Towers PR (http://towerspr.com)come to mind.
For a starter list of who’s who in algorithmic trading see the firm listing at: http://en.wikipedia.org/wiki/Algorithmic_trading#Communication_Standards
For some background on specific algo strategies search the web for: TWAP, and VWAP, and perhaps some of the following: AutoReload, Dagger, Direct To Market, Dynamic Scaling, Get me done, Iceberg, IntelliShort, Relative Scaling, Sniper, Sonar, VolPar etc.
To assess which algo might be selected to accomplish a specific trading object try searching the web on “pre trade analytics” and “post trade analysis”.
For actual XML samples that describe algo orders register for free on the FIX site at http://www.fixprotocol.org/register, then log in and look at the following directory.
http://www.fixprotocol.org/working_groups/algowg/documents. Inside the XML, for each strategy, all the custom algo parameters are listed.
Total visibility into the entire algo marketplace is impossible as firms carefully protect their very best proprietary trading strategies. However by working it from the above “angles” you can get a good idea of what is available, and what the structure of the algo market looks like.
Rick
[ original email was from John Greenan - john.greenan@alignment-systems.com ]
Do you want to invest in software houses that make Algorithmic trading software or do you want to use algos as a buy-side market participant?
Thank you everyone for responding to my post. Your help thus far has been very useful. I was particulary pleased to get to speak to some of you and learn more about the subject.
In response to John Greenan’s message, we are a VC currently considering an investment into a company that develops algorithmic trading software.
So, if i understand correctly, this market is still in its early stages, and there is room for improvement/development. However, switching costs are high, thus making it unlikely that an institutional trader would switch software easily. Is this right? Do institutions ever use more than one algorithmic software? How are these products even marketed?
Also, are any of you technical consultants in this area? And if so, are you based in London by any chance, or know someone I could speak to down here?
I sincerely appreciate and all opinions I can get on this issue.
regards,
Marta
Marta,
Some notes based on your questions.
So, if i understand correctly, this market is still in its early stages, and there is room for improvement/development.
I’d characterize it as the early, very high growth period. Virtually all, large, buy-side institutions currently trade right now, every day, using algorithmic trades. Virtually all, large sell-side institutions offer them. They are well tested and proven to add value in these environments and growing strongly.
However, switching costs are high, thus making it unlikely that an institutional trader would switch software easily. Is this right?
Most large, institutional Buy-Side traders sit in front of an order management system – a software package that they license - that allows them to pick from say 20 to hundreds of different algos, offered by say 10-20 different Sell-Side institutions. When that trader goes to trade, he makes the algo purchase decision. Switching from one strategy to the next, one sell-side institution to the next one. It’s exceedingly easy. The market for algo execution is almost pure competition.
That said, there are some “stickiness” factors on the consume side. Things like:
Sales and marketing efforts by Sell-Side to reach potential Buy-Side customers to introduce the new strategies is very expensive and limited.
Getting Buy-Side traders to trial new strategies and use them enough to trust them and be comfortable with them takes time.
Service levels – how responsive is the Sell-Side to the service needs of the Buy-Side? Seminars presented? Tutorials given? Hand holding? Questions answered? Errors in usage resolved?
Commission arrangements (pricing, terms, volume discounts, etc).
Buy-Side preferences to limit trading to fewer firms vs many, to simplify life and to leverage purchasing power.
Human resistance to change – particularly since trading can be high stress. Constant change and “continuous innovation” is tantamount to total chaos in trading.
On the supply side, (those that construct algos and allow others to use them) there are more constraints:
Heavy R&D investment to write a good algo. It sounds easy, but it really isn’t. It takes a ton of market experience plus excellent coding skills to write a viable algo. Often that means a team-programming environment, over a long period of time, with the entire associated overhead that brings.
If the algo is so great, why market it at all? Just hit the “make me lots of money” button on the algo and let the machine do the work. Realistically, if the algo actually generates a huge incremental return why not just trade it in-house? Here you are back to “program trading”. If it makes a ton of money all by itself there is no need to license it out, let others trade with it, or do anything with it other than keep it tightly locked up, just churning out fantastic returns. Write it once, set it in motion, keep leveraging, and retire early. Game over. No VC needed.
Of course it’s typically not that easy. Virtually all, successful algos out there now are merely tools, which are basically worthless unless deployed by a human trader under the right circumstances. They don’t add a huge amount of value in all situations, yet they do add some solid value in the right situation. I.e. they are helpful to traders and portfolio managers because they reliably do something of value in a specific situation. The Buy-Side trader is then often willing to delegate work to the algo, and pay a higher commission for the convenience and results that it can bring.
The economics are as follows: very high-up front R&D cost, very high up-front execution infrastructure cost, and very high up-front sales, marketing and service costs. Those substantial up-front costs are then recovered over hundreds to thousands (or more) clients, over time, and over many, many individual trades involving plenty of shares. The incremental cost charged to use the algo is typically pennies or less per share. Yet, over time, those pennies add up to a handsome return for the algo provider, if the algo is popular and is utilized.
Selling advanced trading strategies is very much like gun running. Ideally you sell to both sides and they don’t ultimately kill each other with your products. The “killer” ago strategy you sell today is likely to engender the “counter” algo strategy of tomorrow. This new development will greatly reduce the original algos effectiveness and popularity in use. So most algos will require constant maintenance and updating. Likewise, competitors or in-house programming teams will fairly quickly knock off an effective algo that remains static for a sufficient period. The revenue potential will diminish fairly rapidly at that point.
Do institutions ever use more than one algorithmic software?
Large Buy-Side firms typically have access to all Sell-Side algos that are released to the market. (A Sell-Side firm may also have algo strategies that are not released and are only used internally, on proprietary trading desks.)
Buy-Side firms use an Order Management System (OMS) that is infrequently changed out in its entirety, yet frequently enhanced and updated.
Buy-Side firms sometimes have an EMS in house that is infrequently changed out in its entirety, yet frequently enhanced and updated. More often execution management is left entirely to the Sell-Side.
Sell-Side firms use an Order Management System to receive customer orders, including algo orders. This system is infrequently changed out in its entirety, yet frequently enhanced and updated. They also use an OMS on their own proprietary trading desks.
Virtually all Buy-Side firms have an Execution Management System in house that is infrequently changed out in its entirety, yet frequently enhanced and updated.
Algos get displayed on the OMS and executed on the EMS. The parameters that drive a particular algo order are entered through the OMS. The execution logic part of the algo receives those parameters and executes on the EMS. Think of an EMS like a mainframe – its large and changes infrequently. The applications that run on it (the algos) however can change quite frequently.
How are these products even marketed?
OMS and EMS packages are marked just like any other piece of capital equipment – with lots of sales people, sales calls, extensive presentations, long selling cycles, lengthy negotiations, etc.
Algos are marked by the Sell-Side through their regular institutional sales staff (dedicated sales staff calling on assigned regular customers and prospects). There are in person presentations, demonstrations, seminars, brochures, sales calls, etc. Since these are not “large” purchases “up front” they have a much shorter sales cycle. Also, they can go through an innovation/release cycle fairly quickly. Remember, the customer sits in front of a CRT offering him 100’s of algo choices, when he selects a particular strategy, that particular algo provider wins.
Algos that are just run in-house are not marketed at all.
Also, are any of you technical consultants in this area? And if so, are you based in London by any chance, or know someone I could speak to down here?
London & NYC are good hubs of Algo programming activity. One suggestion would be to find headhunters that supply the various well known entities with algo programmer talent. There are several specialists. I have seen recently a headhunter looking for no less than five top-notch algo programmers (at $250k/year each) for a single entity, all at once, with great urgency. Obviously that entity is looking to quickly build some new algos. So the market for talent is exceedingly strong. And the headhunters are very visible searching for talent to supply it.
A good headhunter consultant might be able to assess the talent assembled by your potential VC company and let you know how they might hold up against the likely competition.
Rick
[ original email was from Barry Marshall - Barry.Marshall@bidroute.com ]
> Hello everyone,
I am currently researching the algorithmic trading market- the key
players, what specs are most important when looking at competing
products, and most importantly the size of the market.Would anybody be able to provide some brief input on any of those? Or
perhaps refer me to someone I could speak to a resource centre I can
look at to learn more about the above-mentioned issues?Thanks in advance!
regards, Marta
A priori one can not say what is the best way to trade a share or program. This can only be decided at the time. For example at 10 am it might be sensible to trade agency a normal market size of Vodafone by a simple telephone call with a broker, but at 11 am if BT has just announced a profit warning it might not be sensible to trade Vodafone at all.
Thus what is the size of the algo market might be framing with the wrong question. To effect Best Execution (no need to get hung up on the term, just use its common sense interpretation) a fund manager needs to have the tools that enable it to do its job for its clients. A stat arb manager is likely to need something different to a stock picker at a small mutual fund. The manager must equip itself with the tools that make sense to those clients. The normal “make or buy” decision comes up as tools cost money.
A fund manager needs to be in control of its trading and would not want to be beholden to its brokers if they supply products for “free”. Given the expense of writing algos and the plumbing to use them this might suggest that tools that enable the buyside to be in definite control of its own destiny should, all other things equal, be favoured.
[ original email was from John Greenan - john.greenan@alignment-systems.com ]
“A fund manager needs to be in control of its trading and would not want
to be beholden to its brokers if they supply products for “free”. Given
the expense of writing algos and the plumbing to use them this might
suggest that tools that enable the buyside to be in definite control of
its own destiny should, all other things equal, be favoured.”
Are you concluding that a buy-side should or should not use broker algorithms? I just cannot seem to figure out what you are recommending here.
[ original email was from Barry Marshall - Barry.Marshall@bidroute.com ]
> "A fund manager needs to be in control of its trading and would not want
to be beholden to its brokers if they supply products for “free”. Given
the expense of writing algos and the plumbing to use them this might
suggest that tools that enable the buyside to be in definite control of
its own destiny should, all other things equal, be favoured."Are you concluding that a buy-side should or should not use broker
algorithms? I just cannot seem to figure out what you are
recommending here.
John,
I am not concluding rather contributing to the discussion but my points are:
- You can not say how a trade should be effected until the time to trade comes
- a buy side manager has a fiduciary duty to its clients to Best Execute. It should therefore have access to the tools that enable it to do so for the type of funds it manages
- if it does not have access to the right tools then it can not effect Best Execution - if you like it is trying to trade with one hand behind its back. This might be acceptable if it is your own money, not so if you are managing others
- Algos - broker generated or otherwise - are a tool, so is crossing so is DMA, so is program trading, so is a dark pool etc etc
- If you rely on a broker generated algo you have decided to “buy” rather than “make” it yourself. This might be very sensible but you have to consider the ramifications
- An analogy. You buy a hut at the top of a mountain and decide to heat it only with gas supplied by one pipeline from one gas company. Your choice. It might be sensible, but what about the ramifications if the gas company turns off the supply, deliberately or by accident. It might be sensible for you to take control and take up some wood and a match, or take a gas cylinder with you.
- And further with the analogy. You can make your own decision but if you expect my kids to come and stay with you in the hut with you your “fiduciary duty” should mean you take wood, matches and a gas cylinder with you !!!
Barry
PS To open, the commercial angle from me is that BidRoute allows the buyside to take total control of program trading but my points above stand even if BidRoute is taken out of the equation.