2020: An Algo Odyssey Presentation for Stanford University November - - PowerPoint PPT Presentation

SECURITIES DIVISION

Presentation for Stanford University 2020: An Algo Odyssey

November 2020

2

Confidential

Electronic FX Markets and Algos

SECURITIES DIVISION

3 Confidential

Liquid currency pairs

Daily volumes

Average Top of Book Spread (EURGBP example) Significant day on day changes in volumes executed electronically. Uneven distribution of volumes across currency pairs.

SECURITIES DIVISION

4 Confidential

Market Microstructure I

Key Indicators: Market Volumes and Bid-Offer Spread

Average Top of Book Spread (EURGBP example) Minute by Minute Intraday volume charts (EURUSD example) Pick high-volume periods of the day where possible for Pegged and Hybrid Execution Wider top-of-book spread favours faster settings in Pegged algo

SECURITIES DIVISION

5 Confidential

Market Microstructure II

Key Indicators: Orderbook Depth and Imbalance

Seasonal / intraday orderbook imbalance (EURUSD example) Intraday orderbook depth (EURGBP example) Orderbook depth is key to minimizing implementation shortfall of sweep-to-fill algos like Dynamic Aggressive In markets with a significant orderbook imbalance, there is potential price pressure that may favour an Iceberg

SECURITIES DIVISION

6 Confidential

Adaptive aggress logic for dynamic hybrid

Market seasonality and variability

FX markets exhibit

• intra-day variability,
• intra-week seasonality,
• event-driven changes
• month-end, quarter-end effects…

Intra-day top-of-book spread (HK time) Distribution of daily EURUSD volume by day of the week

The information provided herein was supplied in good faith based on information which we believe, but do not guarantee, to be accurate or complete; however we are not responsible for errors or omissions that may occur. Past performance is not indicative of future results. Data source: EBS, Reuters. Data period: 05-Oct-2014 to 04-Oct-2019.

SECURITIES DIVISION

7 Confidential

Mircrostrucure impact on algos

Liquidity supply and demand and intro to algos

Orderbook Depth Market Activity Algo Favoured by Live Liquidity Conditions HIGH HIGH Dynamic Aggressive NORMAL HIGH Dynamic Hybrid Slower LOW HIGH Pegged Passive HIGH NORMAL Dynamic Aggressive NORMAL NORMAL Dynamic Hybrid Normal LOW NORMAL Pegged Neutral HIGH LOW Dynamic Aggressive NORMAL LOW Dynamic Hybrid Faster LOW LOW Pegged Aggressive

8

Confidential

New Markets, New Algos

SECURITIES DIVISION

9 Confidential

Adaptive aggress logic for dynamic hybrid

Algos Learning from Algos I

• Slower/normal/faster modes of dynamic hybrid target given probability of aggressing top of book
• Dynamic model uses a live metric of its aggress rate to update the aggress parameter in real time.
• What constitutes a better model:
• Mean closer to the target aggress probability
• Standard deviation smaller

NFP Aggress parameter over time

• n NFP day

The information provided herein was supplied in good faith based on information which we believe, but do not guarantee, to be accurate or complete; however we are not responsible for errors or omissions that may occur. Past performance is not indicative of future results. Data source: EBS, Reuters, GS. Data period: Jan 2019 to the present. Probability density

SECURITIES DIVISION

10 Confidential

Dynamic Aggressive and Iceberg Sweep to Fill Logic

Algos Learning from Algos II

• Pre-trade analytics on TED (right) estimates

liquidity supply as a function of orderbook depth

• Use first wave of algo aggress logic to verify the

liquidity shown, or make adjustments as needed

The information provided herein was supplied in good faith based on information which we believe, but do not guarantee, to be accurate or complete; however we are not responsible for errors or omissions that may occur. Past performance is not indicative of future results. Data source: GS. Data period: Jan 2019 to the present.

SECURITIES DIVISION

11 Confidential

Dynamic execution of NDF currenices

Smart NDF Algos

The information provided herein was supplied in good faith based on information which we believe, but do not guarantee, to be accurate or complete; however we are not responsible for errors or omissions that may occur. Data source: GS, primary markets. For illustrative purposes only.

SECURITIES DIVISION

12 Confidential

Portfolio execution for FX

FX Basket algos

0.0% 10.0% 20.0% 30.0% 40.0% 50.0% 60.0% 70.0% 80.0% 90.0% 100.0% 21:36.0 28:48.0 36:00.0 43:12.0 50:24.0 57:36.0 04:48.0 12:00.0 19:12.0 USDCHF USDNOK EURAUD EURCAD CADJPY USDCAD The information provided herein was supplied in good faith based on information which we believe, but do not guarantee, to be accurate or complete; however we are not responsible for errors or omissions that may occur. For illustrative purposes only. Data source: GS.

SECURITIES DIVISION

13 Confidential

Trailing legs, detached legs

FX Basket algos

three normal,

• ne trailing,
• ne detached

leg three detached legs

Trailing leg Detached leg

The information provided herein was supplied in good faith based on information which we believe, but do not guarantee, to be accurate or complete; however we are not responsible for errors or omissions that may occur. For illustrative purposes only. Data source: GS.

SECURITIES DIVISION

14 Confidential

Gamma strategies and KI/KO conditions

Algos for FX options traders

The information provided herein was supplied in good faith based on information which we believe, but do not guarantee, to be accurate or complete; however we are not responsible for errors or omissions that may occur. For illustrative purposes only. Data source: GS.