Game Theory · Diagnosis Clinic B · ← Lesson 08

Which game is this — again?

Four more lenses, four more ways to misfire. This round is meaner: two of the six cases look exactly like a tool you now trust — until you check whether it actually applies.

~12 min · after Lessons 05–08 Skill: picking the right lens cold Format: unlabeled cases

WHY THIS EXISTSFour more lenses, and two of them can fool you

You added four lenses since Clinic A. Lesson 05: coordination & focal points — when a game has more than one stable outcome, the job isn't out-thinking anyone, it's landing on the same one they do. Lesson 06: sequential games & backward induction — when one side moves and the other reacts, a grid is the wrong picture; solve the tree from the end backward. Lesson 07: commitment & credibility — a threat is worthless until you've removed your own way out; only then does a rational opponent believe it. Lesson 08: repeated games & tit-for-tat — when the same counterpart shows up next quarter, the shadow of that future can rescue cooperation a single round never would.

Six more situations from your world, stripped of labels. Four of them are straight reps for the four lenses above. The other two are traps built to look like a lens you just learned to trust — misdiagnose those and you'll run real machinery on the wrong shape.

THE CASESSix unlabeled decisions

Case 1

Referrers in your network hand off patients either through a shared WhatsApp group or through the hospital's e-referral portal. Both channels work equally well for you and for them — what matters isn't which one wins, it's that everyone lands on the same one. Right now the group defaults to WhatsApp simply because that's what everyone already uses.

Which lens?

Coordination & focal points (L05). Both-WhatsApp and both-portal are stable — nobody gains by switching alone, so it's not one dominant path you have to compute. The only real work is agreeing, and an existing convention breaks the tie for free: that's a focal point doing the coordinating, not payoff math.
Case 2

For the skull-base course, you publish the sponsorship tiers and the deadline first. Each device and pharma sponsor then decides, having seen your ask, whether to commit and at what level. You chose the numbers by imagining how each sponsor would react to them — then worked back to what to publish.

Which lens?

Sequential games & backward induction (L06). You move first and each sponsor observes that move before choosing — order is doing all the work, which a simultaneous matrix can't represent. You picked the opening ask by working out each sponsor's best reply to every possible number, then reasoning backward to the number that produces the terminal outcome you want.
Case 3

To stop price-shoppers from angling for a discount on cranial cases, you don't just tell patients "the price is firm." You publish the ≥R$30k floor on the site, tell reception there is zero discretion to quote under it, and take yourself out of the loop as the person who could override it case by case.

Which lens?

Commitment & credibility (L07). A firm-price announcement alone is cheap talk — you could still quietly cave. Publishing the floor and stripping reception's discretion removes your own option to discount. That's what makes it credible: the patient isn't trusting your restraint, they're facing a structure where caving is no longer available to you.
Case 4

You and the rival clinic have been bidding on the same keywords for well over a year with no end in sight. When they push their bid up, you match it within days; when they ease off, you ease off too — neither of you is trying to win the whole war in one shot.

Which lens?

Repeated games & tit-for-tat (L08). No known end date means every bid casts a shadow forward — next quarter's version of this game depends on how this round goes. Cooperate first, then mirror their last move: that discipline holds up because the interaction keeps repeating. A one-shot dominance read ("always outbid") would throw away exactly the thing keeping both of you in check.
Case 5 · the trap

A device vendor offers a co-marketing pilot with a fixed three-month term — after which the contract simply expires, no renewal built in. You're tempted to run it the same way you run the keyword rivalry: cooperate first, then mirror whatever they did last round.

Which lens actually fits?

Backward induction on a finite horizon (L08's limit, via L06). The shadow of the future needs a horizon that's open or uncertain. Here the last round is known: in month three, neither of you has a future left to protect, so defecting is dominant; anticipating that, month two unravels the same way, then month one. Tit-for-tat doesn't fail gently here — a known end date collapses cooperation backward from the finish line before it ever starts.
Case 6 · the other trap

You tell the hospital administrator you'll shift your entire case volume to a rival facility unless OR block time gets renegotiated — while every referral line, staff privilege, and patient relationship you have is built around this one hospital, and everyone in the room knows it.

Which lens actually fits?

Failed credibility test (L07's own check, run in reverse). A threat only bites once carrying it out is cheaper for you than backing down, and the other side can see that. You haven't removed a single option — moving your whole volume would gut your own practice, so the administrator's rational move is to wait you out. Stating a threat isn't commitment; commitment is burning the exit before you speak.

THE TELLHow to tell them apart in five seconds

If you catch yourself thinking…Reach for
"Either landing beats winning the tiebreak"Coordination & focal points (L05)
"They'll see my move before they pick theirs"Sequential games & backward induction (L06)
"I need to make backing out impossible, not just promise"Commitment & credibility (L07)
"This isn't ending anytime soon"Repeated games & tit-for-tat (L08)
"Wait — this one has a known last round"Backward induction unravels it, even mid-repeat
"Could I actually carry this out if they called it?"Credibility check — no answer means it's empty
The clinic's one idea

The lens that burned you twice this round wasn't a new one — it was an old one applied past its expiry date. Before trusting a tool, ask what would have to be true for it to fail: a known ending, or a threat you can't actually afford to keep. That question is cheaper than the mistake.

YOUR REPDiagnose a real one, cold

Do this before you next solve anything. Open DECISIONS.md, pick a row you haven't diagnosed yet, and — before reaching for any matrix or tree — write one line: which of these shapes is it? Coordination, sequential, commitment, repeated, a known-end that unravels, or an empty threat?

Then say why in a sentence, and only then solve it. If you catch yourself reaching for tit-for-tat on something with a fixed end date, or calling a threat credible before checking whether you could actually carry it out, that's the diagnosis this clinic was built to catch. Bring it to me.

The gate: the win isn't six quiz answers — it's catching yourself, on a live decision, running an expired lens before it costs you anything. Log the first time you do.