Back-test case: reading a decision score against what happened
The essentials in 30 seconds
- Reading a score against an outcome means lining up dated inputs, the score, and the later events.
- The EQOS Asia 2008 back-test uses public 2006 and 2007 data and returns φ(S) = 0.234, with 6 of 9 dimensions critical (EQOS Asia 2008 back-test page, as of October 2026).
- The leverage threshold of 3.3% compares with an actual decline of 20%, about 6 times larger.
- The study is a retrospective exercise, and one case is a sample of one.
- A short log of score, date, and outcome lets a team repeat the reading.
A back-test only becomes useful when its score meets what happened. This page works through the EQOS Asia 2008 back-test, as of October 2026, and compares each reading with the events that followed. It applies the steps from how to read a back-test result to one case and ends with a log you can reuse.
On this page
Compare the EQOS Asia offers and request a 30-minute pre-diagnostic.
What does a back-test case show when a decision score meets the outcome?
A back-test case shows three things on one page: the information available on a given date, the score computed from it, and the events that followed. The EQOS Asia 2008 study applies its method to data from 2006 and 2007 taken from the SEC, the BIS, the FDIC, and the S&P Case-Shiller index (EQOS Asia 2008 back-test page, as of October 2026).
The comparison lets a reader judge the method, and it leaves the narrative aside. Where the score and the events agree, the reader gains confidence in the dimension drivers. Where they diverge, the reader learns which input to question. Both results improve the next decision, which is the purpose of reading a back-test with care.
Which inputs and dates anchor the 2008 back-test case?
The study lists data points that were public before the crisis. Housing prices rose 124% between 1997 and 2006. The price to income ratio stood at 4.6 against a historical average of 3.2. The share of subprime lending reached 23.5%, up from 8% in 2003, and the subprime default rate reached 14.4% at the end of 2006 (EQOS Asia 2008 back-test page, as of October 2026).
The first reading step is to check the date of each input. A figure published after the decision date cannot enter the score, and the study states that every source was available in 2006 or 2007. This discipline keeps the exercise honest, because hindsight is the main risk in any back-test. A reader can verify each source by its public name.
How does the 2008 decision score of 0.234 line up with the actual events?
The study places an alert in mid-2006 at φ(S) = 0.234 with 4 recommendations, and an absolute stop in early 2007. On the public side, HSBC announced $10.6 billion of subprime losses in February 2007, Bear Stearns sold at $2 a share in March 2008 against $172 a year earlier, and Lehman Brothers failed in September 2008 with $619 billion of debt (EQOS Asia 2008 back-test page, as of October 2026).
The study also projects a point of no return about 1.1 years after mid-2006, which falls between mid-2007 and early 2008. The figure sets the two timelines side by side. A reader should note that the score led the public events by several months, and that this lead is the quantity worth tracking in later cases.
Which dimensions drove the decision score in the 2008 back-test case?
Six of the 9 dimensions fall below the 0.3 critical line. Energy stands at 0.85 and resonance at 0.65, while communication scores 0.15, integrity 0.08, stability 0.08, and care 0.05 (EQOS Asia 2008 back-test page, as of October 2026). Each low score has a named public driver, such as off-balance-sheet exposure or loosened leverage rules.
The ratio between resonance and transparency, 0.65 to 0.15, equals 4.3 to 1 and exceeds the 3 to 1 level that the study treats as a bubble anomaly. Reading drivers this way tells a reviewer which question to ask first. High energy beside low integrity is the pattern to remember, and the table lists the scores.
What does the leverage threshold add to reading a decision score?
A score gains meaning when a threshold sits beside it. The study shows leverage of about 30 to 1 at large investment banks, against a prudential range of 12 to 15 to 1, and computes that a decline of only 3.3% was enough to wipe out equity (EQOS Asia 2008 back-test page, as of October 2026).
The actual decline in 2007 and 2008 reached 20%, about 6 times the threshold. This is the kind of comparison a reader can check against the outcome: a computed breaking point, then the observed move. A back-test that publishes such numbers can be audited, and the audit is the strength of the format compared with a narrative verdict.
Where does the 2008 back-test case reach its limits as a decision tool?
The study calls itself a retrospective educational exercise with scores computed on public data, and it carries no investment advice. The analyst chooses the inputs after the outcome is known, so some hindsight remains. The study also describes a counterfactual in which 4 corrections raise the score to 0.365 and reduce the magnitude of the crisis by a factor of 5 to 10. That statement is an estimate, not an observation.
One case is a sample of one. A reader who wants confidence in a method needs several cases with different sectors and dates, such as the other EQOS Asia case studies, and a record of misses as well as hits. This page is general information, not legal, tax or investment advice.
How do you log a decision score against an outcome in your own review?
Keep a four-column log: the date, the inputs used, the score with its zone, and the outcome observed later. Add one line on the dimension that drove the result. After 5 or 6 entries a team sees patterns, such as which dimension tends to lead events and which tends to lag.
The free simulator produces a first φ(S) estimate to start the log (EQOS Asia simulator page, as of October 2026), and the method page explains the zones and the 0.618 STOP line. A 30-minute pre-diagnostic, scheduled in Singapore time, applies the same logic to a real decision under NDA (EQOS Asia contact page, as of October 2026).
Frequently asked questions
What does it mean to read a score against an outcome?
It means placing the date of the inputs, the score with its zone, and the later events in one table, then asking where they agree and where they diverge. The reading judges the method and leaves the story aside. Each divergence points to an input worth questioning in the next case.
What score does the 2008 back-test return?
The EQOS Asia 2008 back-test returns φ(S) = 0.234, with 6 of 9 dimensions in the critical zone and a negative ethical safeguard. It uses public data from 2006 and 2007 and presents the result as a retrospective exercise that carries no investment advice (EQOS Asia 2008 back-test page, as of October 2026).
Why does the leverage threshold matter in the case?
It turns a score into a testable claim. The study computes that a decline of 3.3% was enough to cause bankruptcy at about 30 to 1 leverage, and the observed decline reached 20%. A reader can compare the computed threshold with the real move and judge the method.
How many cases does a reader need before trusting a method?
More than one. A single case is a sample of one, so the reader benefits from several cases across sectors and dates, with misses recorded beside hits. The EQOS Asia case studies page lists the Theranos, WeWork, and startups analyses next to the 2008 study.
Test your knowledge
1. What score does the 2008 back-test return?
- 0.085
- 0.234
- 0.618
Show the answer
0.234.
2. Which decline did the study compute as the breaking point?
- 3.3%
- 20%
- 124%
Show the answer
3.3%.
3. How many dimensions fall in the critical zone in the 2008 case?
- 1 of 9
- 3 of 9
- 6 of 9
Show the answer
6 of 9.
4. What is the main risk named for any back-test?
- Hindsight
- Low volume
- Slow software
Show the answer
Hindsight.
5. How many columns does the suggested log have?
- 2
- 4
- 9
Show the answer
4.
About the author
Stéphane Jambu
President, EQOS Forecast Asia
Stéphane Jambu leads EQOS Forecast Asia and works as a consultant in e-commerce, search visibility, and AI. LinkedIn profile
Sources: EQOS Asia 2008 back-test page (eqos-asia.com/case-studies/crise-2008.html), method page (eqos-asia.com/method.html), simulator page (eqos-asia.com/simulator.html), and contact page (eqos-asia.com/contact.html), consulted on 22 October 2026. Page updated on 22 October 2026.