Win As Much As You Can
The game theory of a packing floor, and why positive margins won't last
The ClearCut Call: A Packer's Perspective
At most every leadership conference I've been a part of, somebody runs the same game.
Four teams. Ten rounds. Each round your table picks X or Y, and what you're worth depends on what the other three tables picked. Everybody picks Y, everybody makes a dollar. You're the only table that picks X, you make three and the other three lose a dollar apiece. Everybody picks X, all four of you lose a dollar.
Three of the ten rounds are worth triple, five times, ten times. Those are the only rounds where you're allowed to send somebody to the middle of the room to talk to the other tables.
I have watched rooms full of very smart people play this more times than I can count, and it almost always ends the same way. Somebody defects in round five. Everybody defects in round eight. By round ten the whole board is upside down and four tables that could have split a hundred points are sitting on a pile of nothing.
Then the facilitator points at the board and asks what the name of the game was.
Win as much as you can.
Every table in the room heard "you" and thought it meant them.
I've spent the better part of two decades in beef, most of it on the sell side of a packing floor. I want to spend this piece on why that exercise is not a metaphor for the fed cattle market. It is a formally accurate model of it, and the branch of economics that describes it will tell you what happens to packer margins over the next six weeks.
What game theory actually is
Game theory is the study of decisions where your outcome depends on what other people choose, not just on what you choose. It was formalized in 1944 and has been the standard framework for analyzing concentrated industries ever since. Several of the people who built it have Nobel prizes.
The distinguishing feature is interdependence. Ordinary business decisions are optimization problems — given these costs and this demand, what's my best move? A game is different. Your best move depends on their best move, which depends on yours. You are not solving a math problem. You are solving a math problem against people who are solving theirs against you.
Four terms do all the work. They're worth learning properly, because once you have them you'll see this structure everywhere in protein.
A payoff matrix is the table of every combination of choices and what each player gets. The exercise above is a payoff matrix. So is a bid sheet.
A dominant strategy is a choice that is your best move regardless of what anyone else does. This is the most important concept in the whole field and the least understood. A dominant strategy doesn't require you to predict your rivals. It doesn't require you to trust them or distrust them. It doesn't require any theory about their intentions at all. If a move pays better in every scenario, you make it. Working out whether a dominant strategy exists is usually the first thing an economist does with a market, because if one exists, everything else follows automatically.
A Nash equilibrium is a combination of choices where no player can do better by changing his own move alone. It's a resting point. Once you're there, nobody has a unilateral reason to move.
Here is the part people miss: an equilibrium is stable, not good. A Nash equilibrium can be terrible for every player at the table and remain perfectly stable, because "I'd be better off if we all changed" is not the same as "I'd be better off if I changed." The first is true. The second is what actually determines behavior.
A Pareto optimum is the outcome that makes the group best off. When the Nash equilibrium and the Pareto optimum are the same thing, markets work beautifully and nobody writes articles about them. When they're different, you get the structure below.
Working the payoffs
Work the dominant strategy question. The other three tables can be in four states, so check all four.
All three play Y: you make a dollar with Y, three with X. Two play Y and one plays X: you lose one with Y, make two with X. One plays Y and two play X: you lose two with Y, make one with X. All three play X: you lose three with Y, one with X.
X beats Y by exactly two dollars in every single case. That is a dominant strategy. You do not need to know what anyone else is doing, because it never changes your answer.
So all four tables pick X, and all four lose a dollar. That is the Nash equilibrium — nobody improves by switching alone, because a lone switcher to Y goes from losing one to losing three.
Now flip it. All four pick Y and everybody makes a dollar. That's the Pareto optimum, and it is not an equilibrium, because from there any single table improves by defecting.
The gap between those two outcomes is eight dollars, and closing it is impossible through individual good judgment. Every player acting in his own rational interest produces the worst collective result available. That structure has a name — the prisoner's dilemma — and the four-player version is the standard model for a concentrated commodity industry with excess capacity.
Now stop looking at what each table gets and add up what the room gets. Five possible outcomes, five totals:
All four play X — the room loses $4
Three X, one Y — the room nets zero
Two X, two Y — the room nets zero
One X, three Y — the room nets zero
All four play Y — the room gains $4
Three zeros stacked in the middle. Every mixed outcome is a transfer from the restrained table to the aggressive ones, and nothing gets created in the process. The pot only changes size at the extremes.
Four teams, one pot
The four largest firms account for roughly 85 percent of all steer and heifer purchases in this country.¹ Fed cattle are currently 81.4 percent of total cattle slaughter, the highest share since 2007.² Eleven plants above a million head of annual capacity handled 58.1 percent of fed slaughter last year, and a closure this January took that count to ten.²
Two things need saying before anybody reaches for the pitchfork.
Concentration is not rising. Large plants have accounted for a declining share of fed slaughter every year since 2019, and the current percentage is the smallest since 1994.² Plants under a hundred thousand head of annual capacity went from 3.9 percent of cattle slaughter in 2007 to 7.1 percent last year.² The structure has been roughly stable for thirty-five years.
And the extraction story doesn't hold up. Two studies measured it from opposite ends, and the gap between their two answers is the most useful thing in this article.
The first asked what the feeder loses. It compared what packers actually pay for fed cattle against what they would have to pay if more buyers were bidding, and put the difference at about 2.6 percent of the price.³ On a 1,400-pound steer that's roughly fifty dollars a head that never reaches the cattle feeder.
That is not nothing. A feeder marketing 40,000 head a year is out two million dollars. Nobody should wave that off.
The second asked what the packer keeps. Not gross margin — the excess above a normal return on the money tied up in the plant. If you'd earn ten percent putting that capital somewhere else and you earn ten percent in beef packing, the excess is zero. You did fine. You just didn't earn anything extra for your position in the market. That study put the excess at eighty-eight cents a head in the 1990s and $2.59 a head after mandatory price reporting took effect.⁴
So about fifty dollars a head leaves the cattle market, and $2.59 lands in the packer's pocket.
Where did the other forty-seven dollars go?
Nowhere. It got competed away. Into plants built for a herd that doesn't exist anymore. Into capacity nobody can fill. Into bidding wars for cattle in the shoulder weeks. Into the shifts that get added the week margin turns and shut down the week it doesn't.
That's the actual story, and it is worse than the one you've been told. The rent doesn't get captured by anybody. It gets destroyed. The feeder loses it and the packer doesn't gain it, because it evaporates in the competition between four firms to be the one who gets it.
Go back to those five outcomes. Every mixed one nets zero. Defection doesn't create value. It moves value, and the moving costs more than what gets moved.
The problem is not that the four teams cooperate against everyone else. The problem is that they can't cooperate at all.
The number that makes X dominant
For the week ending August 15, fed plant utilization averaged 79.5 percent on estimated fed slaughter of 421,355 head.⁵
Run that back and nominal fed capacity comes to roughly 530,000 head a week. Which means the industry is carrying about 108,600 head per week of capacity it isn't using.
That figure is why chasing is dominant, and the mechanism is cost structure. A plant running 20 percent under capacity has the same mortgage, the same superintendents, the same maintenance schedule, and mostly the same utility bill it would have running full. Below nominal capacity, marginal processing cost stays close to flat.⁶ It only climbs when you push past capacity into overtime and deferred maintenance and tired crews.
So the cost of the next head is nowhere near the average cost of a head. If that head throws off real contribution and costs a fraction of it to process, you bid for it.
That's not greed and it's not bad management. It is the correct answer to the question in front of that plant manager, and it is the correct answer no matter what the other three do. Which is the technical definition of a dominant strategy.
The usual version of this story is two gas stations across the street cutting price a penny at a time until they're both under cost. Right shape, wrong scale. Two players is a duopoly. This is four, and cattle are the scarce input rather than the customer, so the competition runs backwards — instead of cutting the pump price to pull cars in, packers raise the cattle bid to pull cattle in.
Then add the layer the gas stations don't have. Fresh beef expires in about fourteen days, and inventory is worthless as a threat because you can't sit on it to discipline a buyer. So the product side clears at whatever moves it while the input side gets bid up to whatever secures it. Squeezed from both ends, same week, clock running.
Why repetition should fix this, and doesn't
Here's the objection any good economist raises. The prisoner's dilemma has a known solution: play it over and over.
In a one-shot game, defection is dominant and that's the end of it. But in a repeated game with no fixed end date, cooperation becomes sustainable, because today's gain from defecting has to be weighed against every future round in which your rivals punish you for it. If the players are patient enough, restraint becomes a genuine equilibrium. This is well-established theory, it's been demonstrated in tournament play, and simple reciprocal strategies routinely beat aggressive ones over a long enough horizon.
Packers play this game fifty-two times a year, forever. So why doesn't it resolve?
Three conditions have to hold for repeated-game cooperation to work. Beef packing fails all three.
One: the players have to be able to talk. In the exercise, teams send a representative to the middle of the room before the bonus rounds, and rooms that reach the cooperative outcome almost always get there by talking their way to it. Take the conversation away and the room reliably tears itself apart. That conversation is illegal in beef packing. Four firms coordinating slaughter levels isn't strategy, it's a Sherman Act problem, and the law is right about that. This industry plays the game with the negotiation rounds removed.
Two: defection has to be observable. Punishment only deters if you can tell you've been cheated. In this market you can't, cleanly. When your margin compresses, you cannot distinguish a rival quietly adding a Saturday shift from a soft week at retail. Both look identical from where you sit.
This is a well-studied problem, and its consequence is counterintuitive: with imperfect monitoring, cooperation doesn't hold and then break down for a reason. It breaks down periodically for no reason at all. Players interpret a bad demand week as evidence of cheating, retaliate, and trigger a price war nobody chose.
Somebody has run this analysis on beef specifically. A study of weekly packing margins found exactly this pattern — the industry alternates between cooperative and non-cooperative regimes, and average marketing margin ran 68 percent lower during the non-cooperative phases.⁷ The trigger the authors identified for flipping between them was disruption in matching beef demand against live cattle supply.⁷
A 75-year-low cattle inventory running against plants at 79.5 percent utilization is that disruption. We are in the conditions the model says produce the flip.
Three: commitments have to be credible. A promise only changes behavior if it's in your interest to keep it. Everybody knows a packer who announces restraint will abandon it the week margins turn, because that's what the payoff table tells him to do. A commitment nobody can verify and everybody has reason to break isn't a commitment. It's a press release.
The classic insight here is that you strengthen your position by destroying your own options. You burn the bridge behind you so the enemy knows you can't retreat. In business, the credible commitment is always the irreversible one.
The only credible restraint available
Which is why the interesting news isn't in the weekly kill.
Three of the four largest packers have taken capacity actions in the last eight months. A million-head plant closed in January with a shift pulled at another facility. A plant went dark through a labor dispute and came back in September. An eastern plant closed in August and is converting to value-added. Another closure was announced mid-August alongside a facility being put up for sale.
That is restraint. It's just not the kind anybody promised, and that's exactly the point. You cannot quietly reopen a closed plant next Tuesday when margins turn. Closing it is a real commitment because it's irreversible, which is precisely what a spoken agreement to run fewer head can never be. This industry cannot reach the cooperative outcome by agreeing to it. It can only get there by permanently removing chairs from the table.
Here's the problem.
It isn't enough. Roughly 108,600 head a week of unused capacity against one million-head plant removed and some shifts around it. You'd need to take out the equivalent of four large plants to close that gap. The industry has taken out one.
Which means the surviving plants now have more cattle available per plant than they did in January, at exactly the moment margin turned positive.
Where margin actually sits
Packer margins were negative in the overwhelming majority of weeks since late 2024. That stretch has broken.
ClearCut Packer Margin, current: $231.47 per head.
Cattle cost: futures plus basis. Processing: $300 per head. Dressing: 63.5 percent. Drop credit: $0.14 per pound live. Average dressed weight: 887 pounds.⁸
You'll see other figures in circulation, some of them negative. Most of the gap is methodology rather than disagreement about the market — margin estimates that lag cattle cost by a week will understate the current number in a rising market. The assumptions above are mine, they're published, and you can reconcile against them.
Now put that number back in the game. Three of the ten rounds carry a multiplier, and those are the rounds where defection spikes. Not because anyone's character changes, but because when the stakes rise, the reward for cooperating rises and the reward for cheating rises faster.
Margin is real for the first time in nearly two years. Every head carries it. Every operator in the business knows the window is finite.
That's a bonus round.
The call
Cattle cost goes up. Cutout comes down. Packer margin compresses back toward zero, and my guess is through it.
I'm not going to hedge this into a quarter. It doesn't take a quarter. Every time I've watched this setup, it resolves in weeks. Four to six. A month and a half on the outside.
The mechanism isn't complicated, and by now you can derive it yourself. Positive margin is the signal to add hours and add shifts. Added shifts need cattle. Cattle are the scarce input — the January inventory came in at 86.2 million head, the smallest since 1951, with beef cows down another one percent and cattle on feed down three⁹ — so the bid goes up. More head means more boxes, more boxes into the same demand means the cutout comes down. Cost up, revenue down, same calendar.
That isn't a forecast so much as arithmetic with a delay built into it.
I'll be wrong if fall demand is strong enough to hold cutout while the kill climbs. That's the only out I see, and I don't think it's there.
The word "you"
The plant manager adding a Saturday shift this month is not making a mistake. His fixed costs get covered better, his unit cost drops, his pounds-per-man-hour improves, his scorecard gets greener. He is playing a dominant strategy, which is to say he is playing correctly.
And he's going to lose anyway, along with everybody else at the table, because a dominant strategy played by all four players produces the outcome where the room loses four dollars.
That's the thing about the exercise that stays with people. Nobody in that room does anything stupid. Every table optimizes its own scorecard exactly as instructed. The room still ends up in the hole, because the instruction contained a word that could mean two different things and nobody stopped to ask which one it was.
In this business, the room is the beef complex. The feeder who takes the higher bid today and eats the lower one in November. The packer who buys share and gives back margin. The retailer building a feature around a cutout that won't be there when the ad runs.
Four teams, a shrinking pot, and no way to send anybody to the middle of the room.
Round ten always comes.
ClearCut Forecasting builds statistics-based price and margin forecasting for the beef and pork supply chain. Packer margin, cutout, and cut-level forecasting at clearcutforecasting.com.
Sources
USDA Economic Research Service, "Concentration in U.S. Meatpacking Industry and How It Affects Competition and Cattle Prices," Amber Waves.
USDA-NASS, Livestock Slaughter Annual Summary, 2026; industry structure analysis, Oklahoma State University Extension, May 2026.
Moschini, G. and T.J. Smith, "Spatial price competition and buyer power in the U.S. beef packing industry," American Journal of Agricultural Economics, 2026.
Cai, X., K.W. Stiegert, and S.R. Koontz, "Oligopsony Fed Cattle Pricing: Did Mandatory Price Reporting Increase Meatpacker Market Power?" Applied Economic Perspectives and Policy 33(4), 2011.
Sterling Marketing Beef Profit Tracker, week ending August 15, 2026.
Smith, T.J., "Capacity Constraints and Beef Price Spreads," University of Nebraska–Lincoln Department of Agricultural Economics, March 2026.
Cai, X., K. Stiegert, and S. Koontz, "Regime switching and oligopsony power: the case of U.S. beef processing," Agricultural Economics 42(1), 2011.
ClearCut Forecasting, Packer Margin Projection, August 26, 2026.
USDA-NASS, Cattle, January 30, 2026.