8.1Contribution
This article has read the debate over how to judge a decision once its outcome is known, between those who hold that knowing the end distorts judgment (Fischhoff 1975; Baron and Hershey 1988), methods that judge by outcomes counted over many forecasts (Brier 1950; Tetlock 2005), and the doubt that a record made before the decision is a neutral trace (Power 2004; Clarke 1999), through a nineteenth-century narrative of a financial collapse. Mackay's chapter on the South Sea Bubble, read as a record of telling and not of fact, shows hindsight and outcome bias in narrative form, and also preserves material for another telling: a general opinion that the rise was over, a panic whose form nobody knew, a dissenting memoir, and a footnote that keeps later knowledge apart from the earlier text.
Two synthetic WSTiger records, set in 1956 and 2026, pose the same problem as a choice among three ways of judging. Their example judgments name the separation the first position calls for, between a bad outcome and a bad decision and between the clarity that follows a result and the uncertainty at the time. Read with Mackay, they show that a reason written before is what lets a judge separate the two, and that it does not, by itself, settle which way the judgment goes. Their revisit conditions return the judgment to whether signals were knowable or clear at the time, which is where hindsight works.
8.2Limits
The article's evidence is narrow. Mackay is one teller, writing for a popular audience; his chapter is read in a later edition, and the passages he quotes from others are taken as he gives them, without being checked against their originals. Nothing here is a finding about what happened in 1720. The records are synthetic, modern summaries written for the platform, with no outcomes and no reviewers; they are originally in Korean and are read here in an English translation. The secondary works are used only as far as their reading notes go, and the opposition between the positions is this article's arrangement, not a debate the authors are said to have conducted. The article draws no conclusion about which option in either record is right; a partner, a desk or a historian decides that, and the decision carries costs for someone who did not make it.
8.3Further research
Three lines follow. The first is to read other narrative histories of collapse for the forms identified here, the prophet made by the outcome, the reason supplied from the end, the dissent kept or dropped, and to ask whether those forms change when the teller has prior records to quote. The second is to compare, for real decisions where both survive, the reasons written before and the reviews written after, and to see how often the review's claim that a signal was clear can be found in the material from before. The third is closer to WSTiger's form. A revisit condition that turns on whether a signal was knowable could name who would judge that and from what: a reviewer who did not know the outcome, for example, or material sealed before the outcome was known. Whether such conditions can be written, and whether they would be used, is a question the records raise and cannot answer.
Appendix: ledger figures used
The parent study's ledger (content/books/wstiger-risk/ledger.json, as of 2026-10-05) codes the two records as follows. Values are given in the ledger's own terms. "Unsettled" means the two coders disagreed; the coders' values are given in brackets. A5 conditions come from the record's invalidation field; "observable" is the code's check for a number, a threshold or a time.
| Cell | Reason written before voyage | Post-mortem after loss |
|---|---|---|
| time of setting | past | present |
| A5 condition (field invalidation) | present; observable: no | present; observable: no |
| A3 deadline | none | none |
| A2 missing items listed | 0 | 0 |
| decider | unsettled (team-or-organisation, implied; individual, stated) | team-or-organisation (implied) |
| bearer | unknown | those-nearest-the-stoppage (implied) |
| horizon | unknown | unsettled (years, implied; unknown) |
| reversal constraint | unsettled (easy, implied; unknown) | unsettled (easy, implied; unknown) |
| reversal action | unsettled (unknown; impossible, implied) | unsettled (costly, implied; unknown) |
| place | human-decides (implied) | human-decides (implied) |
| human role specified | no (implied) | no (implied) |
| reason recorded | not-applicable (implied) | yes (stated) |
| option 1 reduces immediate loss | no | unsettled |
| option 2 reduces immediate loss | no | no |
| option 3 reduces immediate loss | no | no |
Study-wide figures used in the text, from the parent study's numbers.md: 53 cases; 35 set in the present, 9 in the past, 9 in the future; 53 synthetic. Cases with a condition for revisiting the judgment (A5): 53 of 53 (100%). Conditions stated with a number, a threshold or a time: 5 of 53 (9%). Cases with a deadline in the record (A3): 0. Cases with the decider in the record (A1): 0. For hypothesis H2 (few decisions write down in advance what would reverse them, and those that do write it in words no one could observe), the study's verdict is supports under both readings (supports 48, against 5, unclear 0, not applicable 0; share 91%). Both records are coded as supporting H2 under both readings.