Historical Odds Show That Starting a War Is a Statistically Poor Bet

Source: https://www.facebook.com/americanspectator/. "Starting a War Is a Bad Bet | The American Spectator | USA News and Politics." September 2, 2026. spectator.org

The Gist

The author argues that starting a war is a terrible gamble because history shows that the side that starts a war usually loses, and losing often means the leaders who started it get overthrown. Wars almost never go as planned, they escalate and drag on much longer than expected, so any government considering starting one first is making a very bad bet.

Conclusion

Governments (implicitly including the Trump administration considering conflict with Iran) should avoid starting wars, because the historical odds of success for war initiators are extremely poor and losing often leads to regime change.

Premises

  1. In a study of 15 conventional wars of the 20th century, the side that started the war lost 10 of them, giving initiators a success rate of only about 23.08 percent.
  2. Of the 10 wars lost by the initiating party, 50 percent resulted in regime change for the government that started the war, showing that losing a war often threatens the survival of the initiating regime itself.
  3. Wars tend to 'mutate' beyond the control of those who start them, as seen in WWI (atrocities drawing in Britain and the US), Vietnam (miscalculating the enemy's endurance and domestic political tolerance), and WWII (atrocities eliminating any chance of negotiated settlement).
  4. With only one exception (the First Indochina War), no initiating actor in the study expected or planned for a prolonged conflict, yet prolonged conflicts were the norm.
  5. Even well-resourced militaries can suffer swift, decisive defeat due to miscalculation (e.g., the Falklands War, Desert Storm, Kosovo), showing that incompetence and misjudgment are common and costly.
  6. Rational gamblers would not take a bet when the odds of failure exceed 75 percent, and starting a war carries odds of failure this severe or worse.

Assumptions

View this argument on LogicFirst.ai