Game Theory 101 (#57): The One-Shot Deviation Principle

If searching for profitable deviations in extensive form games required investigating every single combination of alternative strategies, finding equilibria would take forever. Fortunately, the one-shot deviation principle says that we only have to investigate single deviations from a strategy to show subgame perfect equilibria. This will help us a great deal when solving repeated games.

[Gametheory]

7:25 It's not 16 factorial, just in case you were worried about that.

[PunmasterSTP]

One-shot? More like “red hot”! This is one of the best and most informative educational series I’ve ever come across on YouTube. Thank you so much for making and posting all of these videos!

[YuliangKang]

You are really an awesome speaker! Your students are so lucky to have you! Thanks for the great videos:)

[dominiclebron]

so in reference to my comment two videos ago...

as long as im in contact w/ my baby mama, im 'technically' in a repeated non-cooperative PD game where i only have two moves: 'cooperate' or 'defect.'
so for most of the previous games i've been playing cooperate-> earning little to nothing, less & less every time. 1(delta)
I've recently began playing the defect option, yet occasionally reverting to cooperate option. Where in time, i figure out that when i play the cooperate option i will indefinitely lose. LOSING valuable time in the process.

this video tells me i should COMMIT to playing defect.
this is the only way i win... defecting from my baby mama. however, it makes perfect sense. i'm wasting so much time w/ these minuscule payoffs. I need to just COMMIT to defect. I'm losing so much time& UNHAPPY w/ where this 'game' is ending up.

Professor William Spaniel - once again thank u for this fkn insane knowledge.

[rogerhindrich]

what if i have a game with 2 stages and these payouts:
p(LL))=1, p(LR)=0, p(RL)=0, p(RR)=2


Then, using the OSD-principle, LL would look like an equilibrium when in fact RR would be a profitable deviation

[fang-hsuantseng]

In the second stage of game, why don't you have 8 strategies because you have three information sets and 2^3=8?

[yannbouteiller]

Hi, it would be great to have an approachable proof for this theorem

[hawkoli]

Thanks for the wonderful teaching!

May I know what's the intuition behind the second part (or the more valuable part) of the one-shot deviation principle? i.e. if I indeed have a better alternative strategy 'C' (a true SPE) to the current strategy 'A', and 'A' and 'C' are multiple deviations apart from each other. Then I will definitely will find another strategy 'B' which is just one deviation apart from strategy 'A', and 'B' is better than 'A'?

[zakwilloughby]

What exactly does SPE must be optimal for all deviations mean? I’m having trouble understanding that

[madinakabzhalyalova]

If my instructor on game theory was as good as you...