Friday, December 26, 2025

Can AI Performs Elliott Wave Analysis on STI ?

After developed the AI Trading System app, was curious can AI perform Elliott Wave analysis on historical chart to provide a pretty accurate wave count.  So, got a historical chart of STI starting from 1987 till present from Yahoo Finance.  Upload that chart image onto those AI chatbot (Copilot, Deepseek, ChatGPT, etc) and see what can they make of it, can they provide the Elliott Wave count.



The above is the STI historical chart from Yahoo Finance.  Technically, Singapore stock market started in 1966 but unfortunately, Yahoo Finance only have data from 1987 as the earliest.  Any serious implication for the missing data from 1966 to 1987 ?  Well, that difficult to affirm since we don't know during those periods, how STI behaved that could affect the wave count.  Thus, we could only do with starting point from 1987.


STI historical chart from 1987 till now. doing a top down approach, looking from SuperCycle degree of Elliott Wave, can make out the wave count of the above, obeying the rules and violations of Elliott Wave and also adhere as much as possible to the guidelines. provide also detail of the sub level Cycle and Primary degree wave count so they are inline with the SuperCycle degree count


The above is what being asked by me to each of the AI chatbot (Copilot, ChatGPT and Deepseek).  While all three are enable to give an Elliott Wave count in the SuperCycle degree from the above chart, all three responses are different from each other.  The initial response from these AI chatbot unfortunately raised doubts on the validity of the wave count.  Refer here for the detail exchange between me and ChatGPT regarding the wave count.  Was able to rebut successfully whatever the AI has suggested until I suggest my wave count to it to see whether that is valid.  Read the detail exchange and you can see that.



The above is what the wave count that I've suggested to the AI chatbot and you can read through the exchange log to see how the AI response to this wave count.  All I can say is I totally defeated AI for the Elliott Wave count for the STI.


The below is the detail description of my suggested wave count though the above labeling should be clear enough to understand.


Cycle 1 (1987 to 1993)

Normally, wave 1 is considered as impulse wave that display a sub 5-wave structure.  However, for the case of a diagonal (leading or ending), it displays a 3-wave pattern instead.  This is being labeled primary A-B-C in Cycle 1 above


Cycle 2 (1993 - 1998)

The wave 2 correction coincides with the 1997-1998 Asian Financial Crisis.  For those who have lived through those periods including me would surely recalled several important events.  From early 1990 to mid 1995, Singapore was having a stock market fever, you can easily heard auntie and uncle on the street boast about how easy to profit from stock market.  The stock market fever hit the highest in 1993 when SingTel was listed and making almost everybody can have a share of the stock.  From then on, stock market was like a bubble waiting to get burst and eventually led to the Barings Bank Collapse in 1995 and finally the 97/98 Asian Financial Crisis.  These series of events just sit well to describe the peak in 1993 and the subsequent deep correction of wave 2 till 1998.


Cycle 3 (1998 - 2007)

The recovery from 97/98 AFC was not a smooth sailing one as it soon hit another peak in 1999 and fell into Dot Com bubble burst.  The correction fell even further in 2002/2003 due to SARS pandemic and eventually recover to a fierce rally to the peak in 2007.  This fit the Cycle 3 stage with the peak in 1999 and trough in 2003 being the primary A-B-C of Cycle 3.  Another characteristic of a diagonal wave.


Cycle 4 (2007 - 2009)

What've happened from 2007 to 2009 think most would have known and no need me to get into detail.  The sharp correction again form a 3-wave pattern fit Cycle 4.


Cycle 5 (2009 - present)

This is the tricky part since the recovery of 2008 GFC in 2009.  Instead of a clear and well defined uptrend, STI went into a sideway trend till 2018.  Most people who attempt fitting Elliott Wave count would consider the 2009 as the start of a new impulse wave after the 2008 GFC.  Think it is common since the 2008 GFC is quite a major one.  However, the sideway pattern practically throw that assumption out of the window.  The sideway and complex structure can be explained as a 3-3-5 structure which form the 3-wave of the Cycle 5.  The Primary degree of the 3-3-5 structure (A-B-C) can further look into Intermediate degree as a Triple Three - Zigzag - 5-wave pattern as shown in the above diagram.  This is all agreed upon by AI too.


So going forward into 2026, STI could have a limited upside to complete Intermediate wave 3 which follows a correction for the Intermediate wave 4 before resuming to climb the final part of Intermediate wave 5 and conclude Cycle 5 of SuperCycle 1.  Alternatively, Intermediate wave 3 could have ended and STI is in the Intermediate wave 4 before resuming Intermediate wave 5.  Now, that is not the important part.  The most important, serious and deadly part is what happened after SuperCycle wave 1 ?  In Elliott Wave, SuperCycle wave 2 will follow and wave 2 in nature is a deep correction type, often retracting 61.8% - 78.2% of wave 1.  The starting value of STI in 1987 from the chart is about 800.  If assume, STI can hit 5000 to end SuperCycle 1, a 61.8% to 78.2% correction will see STI lands to between 1150 to 2400 for SuperCycle wave 2.  What's is even deadly is this is SuperCycle degree correction meaning timeframe will span multi decades.  While not expecting having the same timeframe as that of SuperCycle 1, half of that duration will be like 20 years !!!  

Note also, neither Copilot nor Deepseek initial response is what I've suggested SuperCycle wave 1 in a leading diagonal structure !!!

For those still don't quite understand how serious the SuperCycle correction is, take a look at Nikkei 225.  Nikkei 225 hit a peak of about 38,000 in 1989 and hitting the lowest point of the correction in 2009.  That is 20 years period.  Then it takes another 15 more years till 2024 to finally move above 38,000.  STI SuperCycle 2 correction could easily follow that trajectory !!!


While the above SuperCycle 1 wave count could still be invalidated depend on how STI will move to invalidate it as described by the AI chatbot, it would be wise to mentally prepare for a SuperCycle 2 correction.  Once get caught on the wrong end and you'll be done for multi decades.