Chris Ryder «Near Silent Trading»

Trading has evolved from a noisy chaotic environment to the near silent running of algorithms. In a fast paced environment where making money is the primary objective where does a developer fit?

The presentation will start with a Trading 101 before diving into how a collaboration between traders and .NET developers that leverages each other’s domain knowledge and experience to deliver profit. The presentation will conclude with a closer look at the technologies, platforms and practices that Gazprom Marketing & Trading use to gain a competitive advantage.

Transcript

1. 0 Near Silent Trading Chris Ryder 2019-09-10
2. 1 Abstract Trading has evolved from a noisy chaotic environment to the near silent running of algorithms. In a fast paced environment where making money is the primary objective where does a developer fit?
3. 2 About me… Name: Chris Ryder Team: Digital Trading Role: Development Team Lead Company: Gazprom Marketing & Trading (GM&T) Prior to GM&T: Energy Trading, Civil Aviation, Marketing, Environmental Consultancy
4. 3 Agenda • Trading 101 • Decision Making • A Developers Approach • Evaluating Strategy Ideas • Trading Strategy Lifecycle • Pipelines • Operational concerns • Deployment Philosophy
5. 4 Trading 101 Trading is: • Buying & selling • Decision making • Trading signals • Taking a position in the market • Accepting risk • Making & losing money
6. 5 Trading 101 Day trading is a zero-sum game…* I Win You Lose * What defines winning and losing is dependent on a background agenda which is specific to the scenario and may not be visible to all parties. In some scenarios both parties may consider themselves to be winners as they have both successfully achieved their end goals.
7. 6 Trading 101 Good Decisions: • Make money ££££££ • Positive Profit & Loss Impact Bad Decisions: • Lose Money ££££££ • Negative Profit & Lose Impact
8. 7 Decision Making If you can make more good decisions then you can make more money
9. 8 Increasing Decisions: Option 1 Increase Capacity Utilization • Ask existing traders to trade more frequently Outcomes: • Lower quality decisions • Lose money more frequently • Burn out
10. 9 Increasing Decisions: Option 2 Scale Out • Hire more traders Outcomes: • Linear Growth • Unchanged decision making process
11. 10 Competitive Advantage Do we gain a competitive advantage by: • Option 1: Increasing Utilization • Option 2: Scale Out No • We're not doing anything differently • We have the same % return
12. 11 The BIG Question Can we make timely good decisions more frequently and trade off the back of them? ? Is there a 3rd option?
13. 12 What does it take to make a good decision? To answer that question lets look at the different decision types ?
14. 13 Decisions: Types Types of decision: • Simple decisions • Complex decisions How do you know there is a decision to be made? The Decision Making Process: Step 1: Identify the decision Step 2: Gather relevant information Step 3: Weigh the evidence Step 4: Take action To support a decision we look for signals Simple Decisions: few/simple signals Complex decisions: many/complex signals
15. 14 Trade Signals – TO DO What is a Trade Signal? "A trade signal is a trigger for action, either to buy or sell a security or other asset, generated by analysis. That analysis can be human generated using technical indicators, or it can be generated using mathematical algorithms based on market action, possibly in combination with other market factors such as economic indicators." https://www.investopedia.com/terms/t/trade-signal.asp
16. 15 Decisions: Timeliness Trading opportunities exist for a finite amount of time • when they pass, the opportunity to make money passes (sorry to disappoint - time travel hasn’t been invented, we can’t rewind time….)
17. 16 Requirements • Consume data • Analyse data • Generate a signals • Make decisions • Continuous operation
18. 17 Collaboration Trading automation requires introducing 2 groups of people to each other: • Traders - market & domain knowledge • Developers - programming & technical expertise
19. 18 Measuring Success Neither group measures success in the same way: • Traders – Profit & Loss* • Developers – Stability & Correctness* * Oversimplified
20. 19 Automation for all? There are many different flavours of trading For example: Discretionary/Speculative Trading • Automation Suitability: Poor Systematic Trading • Automation Suitability: Good
21. 20 If This Then That Systematic Trading: Applies IFTTT to aid decision making using evidence-based domain knowledge "If it’s raining outside, then I will take an umbrella.“ "If it gets unexpectedly cold, then the day ahead price of gas will go up“
22. 21 IFTTT: Deep dive “If it gets unexpectedly cold, then the day ahead price of gas will go up” On the surface this conditional statement seems quite simple. But is it simple?
23. 22 IFTTT: Inner Questions If it gets unexpectedly cold then the day ahead price of gas will go up Where will it get cold? What is the within day price? Is the cold spell localised or widespread? What is the day ahead price curve? What time of year is it? spring/summer/autumn/winter How quickly do we expect the price to increase/fall? What is meant by unexpectedly cold? Has market already priced in to the cold weather forecast? Are our predictions ahead/behind or different/same to our competitors? Are we in a position to trade? …. ….
24. 23 IFTTT: Balancing Effort & Reward “If it gets unexpectedly cold, then the day ahead price of gas will go up” This statement is Complex Not every IFTTT statement is a good candidate for a trading strategy!
25. 24 IFTTT: Realistic Example “If it is economically viable, then we can trade gas around a pipeline” Buy £ Sell £+ Natural Gas Country A Country B Costs
26. 25 Trading Strategy Lifecycle Primary Principle: Fail Fast Idea Evaluate - Historic Data - Jupyter Hub Prioritise Signal Weak (with potential) Stop Weak (no potential) Prioritise Develop Strategy Strong Back test Profitable Stop Profitable: Refine Not Profitable Release Profitable Monitor Not Profitable Refine Stop (Start)
27. 26 Slow Ring Fast Ring Ring Theory Supporting the fail fast principle
28. 27 Timely Decision Platform ? 0.1 seconds → 4 seconds Decision Engine Execution Platform Data Ingestion
29. 28 Data Pipeline * Receive Process Store * Several architecturally different data pipelines coexist. Non-functional requirements determine the architecture employed.
30. 29 Data Inconsistency “Inconsistency is the only thing in which men are consistent.” - Horatio Smith The data is always different: • Price Data • Weather Data • FX Data • Economic Data • Flow Data • Calendar Data • Outage Data • Sentiment Data
31. 30 Strategy Execution Pipeline Composite Strategy Execution Platform Data Streams External Market Complex Execution Platform
32. 31 Trends • Uptrend • Downtrend • Sideways Trend
33. 32 Candlesticks Trading screens and trading strategies all revolve around Candlesticks • Open • High • Low • Close
34. 33 Doji’s & Signals INSIDE DAY – the second candle is within the body of the first bool buySignal; bool sellSignal; bool insideDay = bar.High < yBar.High && bar.Low > yBar.Low; double body = (bar.Close - bar.Open); double dojiLimit = MetaInfo.Tick * 10 * dC.DojiLine; bool doji = body < dojiLimit && body > -dojiLimit; buySignal = insideDay == true && doji == true; sellSignal = insideDay == true && doji == true; The code below looks for an Inside Day Doji and generates a signal
35. 34 Back Testing Market Price Profit & Loss Market Activity
36. 35 Operational Concerns Markets are always moving • Is data fresh? • Are applications running as expected? To reduce risk monitoring becomes critical If get yesterdays data today then I can’t make a decision on it.
37. 36 Deployment Philosophy Core Applications Alpha Strategy Trading Strategy Satellite Applications Loaders Monitors APIs/UIs
38. 37 Recap Why “Near silent trading”? - Silent running algorithms may be doing the trading - Those algorithms are backed by many hours of collaboration and conversation by colleagues with a range of skillsets
39. 38 Recap - Tech
40. Questions…?
41. Thank you