Overseas Chinese Banking Corporation (OCBC Bank)

[swakoo]

Algo trading has been blamed for sudden short-lived violent stock movements in the US markets but I wonder if such algo trading is present (and significant) in the Spore market, even for its large caps? We don't have a big option trading market here and so for one, we can rule out those market makers having to do their regular algorithmic hedging against their own option exposure.

Hi weijian,

Yes, not referring to algo trading to hedge option exposure. Possibly use of algo's to conduct short selling. Based on SGX historical short sell data, short sales of the 3 banks spiked past few trading days.

This is how Gemini describes it. But think the algo could perhaps also be programmed for more sophistication eg to handle a herd instinct mentality.

Institutional funds, quantitative hedge funds, and proprietary trading firms heavily rely on algorithmic trading to conduct short selling.

While a human portfolio manager usually makes the core investment decision (identifying an overvalued stock or a deteriorating business model), computers and automated algorithms handle the execution, timing, and risk management.

How Hedge Funds Use Algos for Short Selling

1. Signal Generation and Quantitative Screening

Statistical Arbitrage & Pairs Trading: Quantitative algorithms constantly monitor large universes of stocks (like STI or SGX-listed equities) looking for historical co-movements. If two correlated stocks temporarily diverge—such as one bank rallying faster than its peers without fundamental backing—the algorithm will automatically trigger a short position on the overvalued stock and a long position on the undervalued one.

Momentum & Breakdown Triggers: Algorithms scan order books and technical indicators for breakdown signals (e.g., crossing below key moving averages or high short-volume spikes) to initiate short positions automatically when downward momentum accelerates.

2. Execution and Slicing Large Orders (VWAP / TWAP)

Market Impact Minimization: Shorting a large block of shares all at once can crash the stock price prematurely or trigger exchange alerts. Funds use execution algorithms like VWAP (Volume-Weighted Average Price) or TWAP (Time-Weighted Average Price) to slice a massive short order into thousands of tiny pieces, spreading them out over hours or days to hide their footprint from the market.

Smart Order Routing (SOR): Algos scan multiple dark pools and exchange order books to execute short orders at the best available prices with minimal slippage.

3. Managing Short-Specific Costs (Stock Borrowing)

Short selling requires borrowing shares first, which incurs daily borrowing fees (fee rates). Advanced fund algorithms check prime broker locate databases in real-time to find the cheapest and most reliable borrow pools before placing a short order. If borrowing costs spike unexpectedly, the algorithm can automatically cut the short position to avoid a short squeeze.

4. Automated Risk Management & Stop-Losses

Short selling has asymmetric risk (theoretically unlimited upside for the stock price means unlimited downside for a short seller). Algorithms enforce strict, emotionless risk boundaries:

Stop-loss triggers: Automatically buying back shares to close the short position if the stock rallies past a specific threshold.

Portfolio delta-hedging: Automatically adjusting short exposures up or down to keep a hedge fund's overall portfolio market-neutral.

[weijian]

This is how Gemini describes it. But think the algo could perhaps also be programmed for more sophistication eg to handle a herd instinct mentality.

Hi swakoo,

This just reminds me of the Keynesian beauty contest - An investor's job is not to select the most attractive person in the beauty contest, but instead it is to guess the person that everyone thinks is the most attractive. Then 2nd level thinkers will say that their job is to estimate the first guess, and then the 3rd level thinkers (if there is something called 3rd level thinkers) will have to guess the 2nd level thinkers' estimates and this continues on....

But of course in theory, this X level of thinking can progress to high numbers but in practice, it probably doesn't go too far. There are not many Jim Simmons around anymore but maybe in another few more years, AGI becomes a reality and Jim Simmons become just a regular man on the (wall) street.

Now, SGX regular NIF flow column has just quantified last week's carnage on the 3 banks and I will produce it in italics below (with emphasis mine):

The STI's 4.1% decline last week was led by financials amid higher competing global Treasury yields prompted portfolio rebalancing, while at home, rising shipping, fuel, material and labour costs heightened scrutiny of earnings resilience and margin sustainability. Net institutional outflow totalled close to S$900 million, with the three STI banks accounting for around four-fifths of the net selling. The trio declined an average 7% over the five sessions, reducing their average total return for the first 41 weeks of the year to 36%.

So in just last week alone, the 3 banks experienced ~700mil of NIF outflow. While the column did not provide individual stats but combined with the stats I provided in previous article (which used Sept2026 YTD data), it seems like the NIF for all 3 Spore banks should be decisively negative for 2026YTD by the end of last week (9th Oct)

12th Oct 2026 SGX column on NIF: https://www.sgx.com/research-education/m...d-bank-led

P.S. This column also indicated that 2026YTD, VB fav discussed stock YZJFH has experienced ~117mil of NIF outflow FY26YTD. This is close to 10% of its market cap at the beginning of the year!

[swakoo]

This just reminds me of the Keynesian beauty contest - An investor's job is not to select the most attractive person in the beauty contest, but instead it is to guess the person that everyone thinks is the most attractive. Then 2nd level thinkers will say that their job is to estimate the first guess, and then the 3rd level thinkers (if there is something called 3rd level thinkers) will have to guess the 2nd level thinkers' estimates and this continues on....

But of course in theory, this X level of thinking can progress to high numbers but in practice, it probably doesn't go too far. There are not many Jim Simmons around anymore but maybe in another few more years, AGI becomes a reality and Jim Simmons become just a regular man on the (wall) street.

Hi weijian,

Your anecdote is spot on. 

The math algorithm expert Jim Simons passed away in May 2024, as the AI revolution was just taking off from around 2023. AI is progressing so rapidly that i suspect maybe practically everybody and their uncle (who wants to and have deep pockets) can download an open weight LLM (and not necessarily a frontier model) now. To construct their own algorithm to handle (exploit) market idiosyncrasies like crowd psychology. On both buy and sell side.