ALGORITHMIC TRADING AN OVERVIEW · Algorithmic Trading – What is it? • Electronic trade...
Transcript of ALGORITHMIC TRADING AN OVERVIEW · Algorithmic Trading – What is it? • Electronic trade...
![Page 1: ALGORITHMIC TRADING AN OVERVIEW · Algorithmic Trading – What is it? • Electronic trade execution “piloted” by a model • Model tries to match a specific execution cost benchmark](https://reader030.fdocuments.us/reader030/viewer/2022040800/5e35e5905bb06531c1012e76/html5/thumbnails/1.jpg)
ALGORITHMIC TRADING
AN OVERVIEW
![Page 2: ALGORITHMIC TRADING AN OVERVIEW · Algorithmic Trading – What is it? • Electronic trade execution “piloted” by a model • Model tries to match a specific execution cost benchmark](https://reader030.fdocuments.us/reader030/viewer/2022040800/5e35e5905bb06531c1012e76/html5/thumbnails/2.jpg)
Algorithmic Trading – What is it?
• Electronic trade execution “piloted” by a model
• Model tries to match a specific execution cost benchmark
• Usually breaks up a larger trade into a series of smaller trades to be executed over time
![Page 3: ALGORITHMIC TRADING AN OVERVIEW · Algorithmic Trading – What is it? • Electronic trade execution “piloted” by a model • Model tries to match a specific execution cost benchmark](https://reader030.fdocuments.us/reader030/viewer/2022040800/5e35e5905bb06531c1012e76/html5/thumbnails/3.jpg)
Algorithmic Trading – How much is there today?
Tower Group says 7% of all Buy Side flow…
Source: “Market Structure and the Role of Exchanges, ECNs and ATSs” Tower Group (2004)
![Page 4: ALGORITHMIC TRADING AN OVERVIEW · Algorithmic Trading – What is it? • Electronic trade execution “piloted” by a model • Model tries to match a specific execution cost benchmark](https://reader030.fdocuments.us/reader030/viewer/2022040800/5e35e5905bb06531c1012e76/html5/thumbnails/4.jpg)
Algorithmic Trading – How much is there today?
..while the Tabb Group reports that over half of all firms have access to algorithms.
Source: “Institutional Equity Trading in America: A Buy-Side Perspective” The Tabb Group (2004)
![Page 5: ALGORITHMIC TRADING AN OVERVIEW · Algorithmic Trading – What is it? • Electronic trade execution “piloted” by a model • Model tries to match a specific execution cost benchmark](https://reader030.fdocuments.us/reader030/viewer/2022040800/5e35e5905bb06531c1012e76/html5/thumbnails/5.jpg)
Most Popular Algorithms
The list of most popular algorithms is dominated by the key execution benchmarks: VWAP and Arrival Price
•VWAP (27%)
•Arrival Price (19%)
• Imp. Shortfall (13%)
•EOD / Beat Close (10%)
•Guerilla (8%)
•Liquidity (8%)
•% ADV (6%)
•Other (9%)
Essentially the same algorithm (32%)
Source: “Institutional Equity Trading in America 2005 –A Buy-Side Perspective.” Tabb Group (2005)
![Page 6: ALGORITHMIC TRADING AN OVERVIEW · Algorithmic Trading – What is it? • Electronic trade execution “piloted” by a model • Model tries to match a specific execution cost benchmark](https://reader030.fdocuments.us/reader030/viewer/2022040800/5e35e5905bb06531c1012e76/html5/thumbnails/6.jpg)
Trading Cost – How Big?
˜ 20bps Implementation Shortfall for Average Trade per Side*
* Plexus BrokerEDGEMonitor (2004)
At 1 x Portfolio Turnover, 40bps Drag on Annual Portfolio Performance
Assume 6% Annual Return
40bps / 6 % = 6.7% negative return contribution
![Page 7: ALGORITHMIC TRADING AN OVERVIEW · Algorithmic Trading – What is it? • Electronic trade execution “piloted” by a model • Model tries to match a specific execution cost benchmark](https://reader030.fdocuments.us/reader030/viewer/2022040800/5e35e5905bb06531c1012e76/html5/thumbnails/7.jpg)
Implicit Trading Cost Attribution
•Market Impact
•Timing / Trend Cost
•Unexecuted orders/Opportunity Cost
•Spread Cost
Strategic / Scheduling
Tactical / Implementation
![Page 8: ALGORITHMIC TRADING AN OVERVIEW · Algorithmic Trading – What is it? • Electronic trade execution “piloted” by a model • Model tries to match a specific execution cost benchmark](https://reader030.fdocuments.us/reader030/viewer/2022040800/5e35e5905bb06531c1012e76/html5/thumbnails/8.jpg)
Bid
Ask
Spread midpoint
Execution Price A
17.30
17.28
17.29
17.31
1 Order –3 Executions Execution A Execution B
10:40am:Order arrives to Buy 60,000
shares
10:40am: Buying 20,000 shares @ 17.31
Cost Attribution Example: Order to Buy 60,000 shares, filled via 3 Executions*
Timing Cost is Critical, but Spread Costs Matter Too!
Spread Cost: 1c
Bid
Ask
Execution Price B
17.35
17.365
17.38
Spread midpoint
Timing Cost 7.5 cents
11:22am: Buying 20,000 shares @ 17.38
Bid
Ask
Execution Price C
17.38
17.34
17.36
17.39
Spread midpoint
Timing Cost 7 cents
Execution C
11:48am: Buying 20,000 shares @ 17.39
Impact Cost: 1c
Spread Cost: 1.5c
Spread Cost: 2c
Impact Cost: 1c
Total Cost: 2c Total Cost: 9c Total Cost: 10c
Order Trading Cost is Sum Total of Executions A/B/C* Hypothetical Executions
![Page 9: ALGORITHMIC TRADING AN OVERVIEW · Algorithmic Trading – What is it? • Electronic trade execution “piloted” by a model • Model tries to match a specific execution cost benchmark](https://reader030.fdocuments.us/reader030/viewer/2022040800/5e35e5905bb06531c1012e76/html5/thumbnails/9.jpg)
Two Common Benchmarks
Volume Weighted Average Price (VWAP)
• A weighted average price
• Trading volumes are the weights
• Intuitive & easy to use for traders > commonly used in the market
• VWAP does not measure Trend/Timing Cost!
Arrival Price
• The portfolio decision price
• Difference between Arrival Price and execution price, plus forgone returns on unexecuted shares is called “Implementation Shortfall”
• There is little agreement in the market on what an “appropriate”implementation shortfall ought to be
![Page 10: ALGORITHMIC TRADING AN OVERVIEW · Algorithmic Trading – What is it? • Electronic trade execution “piloted” by a model • Model tries to match a specific execution cost benchmark](https://reader030.fdocuments.us/reader030/viewer/2022040800/5e35e5905bb06531c1012e76/html5/thumbnails/10.jpg)
VWAP Advantages• Easy-to-use• Direction-neutral
VWAP Disadvantages• “Becoming” VWAP• “Waiting out” VWAP• “Stretching out” the order• “Saying No” to a block
Pros and Cons: VWAP vs Arrival Price
Arrival Price Advantages• Directly related to portfolio
returns (“capturing alpha”)• Can’t be fooled
Arrival Price Disadvantages• Harder to implement• Law of large numbers• Direction-bias
![Page 11: ALGORITHMIC TRADING AN OVERVIEW · Algorithmic Trading – What is it? • Electronic trade execution “piloted” by a model • Model tries to match a specific execution cost benchmark](https://reader030.fdocuments.us/reader030/viewer/2022040800/5e35e5905bb06531c1012e76/html5/thumbnails/11.jpg)
“Becoming VWAP” Threat – Recent Data
VWAP and Arrival Price Cost vs %ADV
0
0.5
1
1.5
2
2.5
33.5
4
4.5
< 1%
1 -
5%
5 - 1
0%
10 -
25%
25 -
50%
50%
>
%ADV
Bps
cos
t vs
VW
AP
05101520253035404550
Bps
cos
t vs
Arr
ival
Pric
e
Cost vs VWAP
Cost vs Arrival Price(diff. adjusted)
Std Dev VWAP and Arrival Price Cost vs %ADV
0
10
20
30
40
50
60
70
80
< 1%
1 -
5%
5 - 1
0%
10 -
25%
25 -
50%
50%
>
%ADV
Std
Dev
of B
ps c
ost v
s V
WA
P
0
20
40
60
80
100
120
140
Str
dDev
of B
ps c
ost v
s A
rriv
al P
rice
StdDev Cost vs VWAP
StdDev Cost vs ArrivalPrice (diff. adjusted)
Data Source: ITG “The Cost of Algorithmic Trading: A First Look at Comparative Performance” (March 2005)
Charts: Instinet
ITG Study Shows Empirical Evidence of ‘Becoming VWAP’
![Page 12: ALGORITHMIC TRADING AN OVERVIEW · Algorithmic Trading – What is it? • Electronic trade execution “piloted” by a model • Model tries to match a specific execution cost benchmark](https://reader030.fdocuments.us/reader030/viewer/2022040800/5e35e5905bb06531c1012e76/html5/thumbnails/12.jpg)
Instinet & Northfield
Instinet already had a robust VWAP rule
We wanted to develop an Arrival Price rule that would
• Factor in an effective measure of timing risk
• Work for portfolios & pairs, not just single stocks
• Factor in short-term stock correlations for portfolios & pairs
So we turned to someone who
• Had expertise in risk management models
• Had a proven short-term risk model
![Page 13: ALGORITHMIC TRADING AN OVERVIEW · Algorithmic Trading – What is it? • Electronic trade execution “piloted” by a model • Model tries to match a specific execution cost benchmark](https://reader030.fdocuments.us/reader030/viewer/2022040800/5e35e5905bb06531c1012e76/html5/thumbnails/13.jpg)
The Instinet & Northfield Partnership
Benefits & Synergies
• Be able to execute portfolios against Arrival Price
• React to real-time market data
• Optimize and re-balance real-time, during the execution process
![Page 14: ALGORITHMIC TRADING AN OVERVIEW · Algorithmic Trading – What is it? • Electronic trade execution “piloted” by a model • Model tries to match a specific execution cost benchmark](https://reader030.fdocuments.us/reader030/viewer/2022040800/5e35e5905bb06531c1012e76/html5/thumbnails/14.jpg)
TRADING TACTICS FOR ALGORITHMIC TRADING
![Page 15: ALGORITHMIC TRADING AN OVERVIEW · Algorithmic Trading – What is it? • Electronic trade execution “piloted” by a model • Model tries to match a specific execution cost benchmark](https://reader030.fdocuments.us/reader030/viewer/2022040800/5e35e5905bb06531c1012e76/html5/thumbnails/15.jpg)
Implementation
Northfield Risk Model
Instinet Trading Engine
Feedback Loop
FIX NEWPORTTM Execution Management System
Instinet Order Handling
Newport TM is patent-pending
![Page 16: ALGORITHMIC TRADING AN OVERVIEW · Algorithmic Trading – What is it? • Electronic trade execution “piloted” by a model • Model tries to match a specific execution cost benchmark](https://reader030.fdocuments.us/reader030/viewer/2022040800/5e35e5905bb06531c1012e76/html5/thumbnails/16.jpg)
Scenario: Buying 50,000 of NYSE-listed mid-cap, Limit 19c, Current Market 16c x 20c. BlockPeg Execution Style “Aggressive”
1
BlockPegPost
at mid-spread
CBX DOT
Post at bid
Market: 16c x 20c Executed: 0 Left: 50,000
16c
20c
16c
20c
18c
21c
2
BlockPegPost
at mid-spread
CBX DOT
Post at bid
Market ?: 17c x 21c Executed: 0 Left: 50,000
17c
21c
17c
19c
CBX DOT
BlockPeg
Post at bid
Market ?: 16c x 20c Executed: 10,000 Left: 40,000
16c
20c
16c
20cPeer institution
takes block @ 18c
3
BlockPegNot
posting
Market ?: 22c x 25c Executed: 0 Left: 40,000
22c 22c
25c25c
CBX DOT
Not posting
Hard limit @ 19c
Hard limit exceeded
4
BlockPegPost
at mid-spread
CBX DOT
Post at bid
Market ?: 15c x 19c Executed: 2,000 Left: 38,000
15c
19c
15c
19c
17c
2000 shares marketable @
19c
5
BlockPegPost
at mid-spread
CBX DOT
Post at bid
Market ?: 14c x 18c Executed: 1,000 Left: 37,000
14c
18c
14c
18c
16cSomeone hits the bid: 1,000 shares @ 14c
6
DOT slice moves with the market
NYSE Spread-cost Minimization Tactic: BlockPegSM
![Page 17: ALGORITHMIC TRADING AN OVERVIEW · Algorithmic Trading – What is it? • Electronic trade execution “piloted” by a model • Model tries to match a specific execution cost benchmark](https://reader030.fdocuments.us/reader030/viewer/2022040800/5e35e5905bb06531c1012e76/html5/thumbnails/17.jpg)
Time9:30 am 4:00 pm
Rule Detects Intra-Day Volatility “Noise”- Pricing Buy Order Passively -
Price
Passive Limit Price
Time9:30 am 4:00 pm
Rule Detects Adverse Momentum- Pricing Buy Order Aggressively -
Price
Floating Aggressive Limit Price
Another Spread-Cost Minimization Tactic: Momentum detection & Mean reversion
![Page 18: ALGORITHMIC TRADING AN OVERVIEW · Algorithmic Trading – What is it? • Electronic trade execution “piloted” by a model • Model tries to match a specific execution cost benchmark](https://reader030.fdocuments.us/reader030/viewer/2022040800/5e35e5905bb06531c1012e76/html5/thumbnails/18.jpg)
Arrival Price
Cost vs Arrival Price of unexecuted
residuals
Tracking Transaction Cost in Newport
Cost vs Arrival Price I have realized (executed
portion of the order)
What the entire order’s realized Cost vs Arrival would be if I executed
all residuals now
![Page 19: ALGORITHMIC TRADING AN OVERVIEW · Algorithmic Trading – What is it? • Electronic trade execution “piloted” by a model • Model tries to match a specific execution cost benchmark](https://reader030.fdocuments.us/reader030/viewer/2022040800/5e35e5905bb06531c1012e76/html5/thumbnails/19.jpg)
Best Execution – What else can you do?
• Use a great model
• Use great technology
• Trade through a broker who doesn’t have a dog in the fight
• 100% Pure Agency
• 100% Anonymous
• Always trading for the customer
• No firm proprietary account
![Page 20: ALGORITHMIC TRADING AN OVERVIEW · Algorithmic Trading – What is it? • Electronic trade execution “piloted” by a model • Model tries to match a specific execution cost benchmark](https://reader030.fdocuments.us/reader030/viewer/2022040800/5e35e5905bb06531c1012e76/html5/thumbnails/20.jpg)
© 2005 Instinet, LLC. All rights reserved. INSTINET and the tagline “The Institutional Broker” are a registered mark in the United States and in other countries throughout the world. Instinet, LLC, member NASD/SIPC, branded as Instinet, The Institutional Broker, is a subsidiary of Instinet Group Incorporated which is part of the Reuters family of companies.
This presentation is provided for informational purposes only. It does not take into account the particular investment objectives, financial situation or needs of any individual or entity. Under no circumstances is it to be used or considered an offer to purchase or sell any security, or a solicitation or recommendation of the purchase, sale or offer to purchase or sell any security.
Reproduction or redistribution of this information in any form in whole or in part is prohibited except with the written permission of Instinet.