The term sheet is in front of you and one clause is wrong. The instinct is to counter that clause. It is a reasonable ask, you have a good argument, and the investor will probably concede it.
Then they reopen valuation, which was settled. A gain at step one that terminates below where you started is not a gain.
Why these three models
The decision is whether to counter at all, and on what. The features that fire are a live status quo that either side can depart from, a quantity that determines how long you can hold out, and a well-populated reference class of how these documents actually settle.
Three lenses. The move tree produces an equilibrium answer about where a sequence of departures terminates. Stock and flow produces a complex answer about what actually sets your staying power. The reference class produces a cycle answer about where terms usually land, which disciplines both. The first two matter most when they meet: the tree frequently comes out symmetric, and then the stock decides.
1. Where the sequence ends, not whether you gain now
Standard game theory asks a specific and narrow question: given where we are, can I gain by unilaterally changing my strategy right now? If neither side can, you are at an equilibrium.
That is the wrong question for a term sheet, because it is the wrong game. A negotiation from a live document is not a simultaneous choice. There is a status quo, either side can unilaterally move from it, the other responds, and it terminates when someone declines to move further. Brams sets out exactly those rules and derives a different stability concept: an outcome is stable in the extended sense when neither side, looking down the tree of departures and counter-departures by backward induction, prefers where they would end up to where they are.1
He is candid that this is not a better solution to the classical problem. Since a game is often defined as the totality of the rules describing it, the rules he sets out in fact define a new game.1 That honesty is why it belongs here: it is not mainstream equilibrium theory, and it is a much better description of what a term-sheet negotiation actually is.
The operational instruction is simple and nobody does it. Before countering, write the sequence. You ask for the clause. They concede and reopen X. You accept or push. Follow it until someone would stop. The terminal cell is what you are actually choosing, and it is rarely the clause you were arguing about.
Move tree, the termination lens
- Assumes: there is a live status quo, departures are unilateral and alternating, and the process ends when someone declines to move.
- Fits because: the document exists and either side can reopen it.
- Breaks when: the counterparty cannot in fact reopen settled terms, in which case the classical single-move analysis is right after all.
- Evidence: grade C plus. Internally coherent, a minority framework by its author’s own account, and thin empirically. Use it as a frame, not as a prediction.
- Counteracts: optimising a clause in isolation.
- May reinforce: paralysis, since every tree can be extended one more step.
2. What actually decides a symmetric standoff
The second lens handles the case the tree cannot resolve, and it is the most common case.
Brams identifies it directly. Where the analysis is symmetric, each side would prefer the other to move first from the mutually bad outcome, and he declines to designate a winner because one cannot say a priori which player would be able to hold out longer.1
That residue is the finding. When the tree comes out symmetric, the payoff matrix has stopped determining the outcome and staying power has taken over. And staying power is not a disposition. It is a stock.
Runway accumulates and drains through flows, and it cannot be set directly. That has two consequences at the table. First, your negotiating position was largely fixed months ago by decisions about burn, and no amount of preparation on the day changes the stock. Second, the lever available now is a flow: deferring a hire, delaying a purchase, collecting a receivable early. Each buys weeks, and weeks are the actual currency of a symmetric standoff.
Expect to underrate this. Highly educated adults routinely fail to infer the behaviour of simple stock and flow systems, and the failure is not attributable to graph literacy, contextual knowledge, motivation or cognitive capacity.2
In a negotiation between equals, the split is set by the bank balance, not by the deck.
Stock and flow, the staying-power lens
- Assumes: runway accumulates and drains through identifiable flows and cannot be set directly.
- Fits because: the tree is symmetric and something outside the matrix has to break the tie.
- Breaks when: the binding constraint is not cash but something unmeasured, such as a board’s patience or a co-founder’s willingness to continue.
- Evidence: grade A. Structural, and the difficulty people have reasoning about it is well replicated.
- Counteracts: preparing for a negotiation as though preparation were the variable.
- May reinforce: conceding early on a runway estimate that was itself too pessimistic.
3. Where these documents usually land
The third lens stops the first two becoming an argument from first principles about a document that has a well-populated history.
Alternating-offer bargaining makes the underlying point plainly: the division of a surplus is determined by relative patience and by who moves, not by any fairness criterion either side can appeal to.3 That is the same conclusion the stock reaches by a different route, which is why the two lenses reinforce rather than duplicate each other.
Before modelling this negotiation, ask where negotiations of this shape settle. Not deals in general: deals matched on structure. Same stage, same instrument, same investor type, same market. Most terms in most term sheets are not genuinely contested, and the small number that are contested are contested in predictable directions.
The discipline is to establish which clause is actually the live one before you spend a move. Founders routinely counter on a term that is standard and unmovable, spending a departure and receiving a concession they were going to get anyway, while the term that will actually cost them sits unexamined because it is buried below the line.
This is also the check on the move tree. A tree built on your intuitions about how they will respond is a story. A tree built on how this investor type has responded to this ask before is an analysis.
Reference class, the settlement lens
- Assumes: negotiations matched on structure settle in a knowable distribution.
- Fits because: term sheets are among the most standardised documents a founder encounters.
- Breaks when: the rules changed, which instrument innovation and a shifting funding market both do.
- Evidence: grade A as a method. The specific base rate for any local market is usually poorly documented, which is the real limitation.
- Counteracts: negotiating from principles against a standardised document.
- May reinforce: accepting a term as standard when it is standard only in a market that is not yours.
The levers, cheapest first
- Write the sequence before you send the counter. Four steps on paper, ten minutes. It is the whole intervention and almost nobody does it.
- Identify which one term is genuinely live. Ask someone who has seen ten of these which clause actually moves. Spend your departure there.
- Count your weeks before you count your arguments. If the tree is symmetric, this is the only number that matters.
- Buy weeks with a flow, now. A deferred hire before the negotiation is worth more than any argument during it.
- Ask what they can accept without reopening. Some concessions are cheap for an investor and some trigger a re-approval that reopens everything. The second kind is where the counter-move comes from.
What to do before you reply
Do now, sized at ten minutes, effect immediate. Write the four-step sequence for the counter you were about to send, and name the terminal cell. Reversible, free, and dominant across every scenario about whether the ask is reasonable.
Hedge, where the premium is the whole loss, live before you reply. Have one person who has seen this document type before mark which clause is live. If your read was right you have spent one conversation, and that is the entire downside.
Defer and trigger, size fixed now. Do not renegotiate the whole document. Pre-commit the trigger: if the sequence you wrote terminates below the current status quo, you do not send the counter at all, and you sign or walk on the terms in front of you. Decide that rule now, because a decision made after a concession has been offered is made under a different reference point.
Note the arrivals. Writing the tree takes minutes. Buying weeks of runway takes effect only after the flow you changed works through, which for a deferred hire is the next payroll and for a receivable is whenever they actually pay. Neither helps a negotiation happening on Friday.
What usually happens next
Run the break test first. Has a rule changed, has an actor entered or left, has a measurement become a target? A new instrument becoming standard in your market means last year’s base rate describes a different document, and a second investor entering the round changes the tree fundamentally rather than marginally.
If nothing broke, the pattern is consistent. A counter on a settled term produces a concession and a reopening, and the founder experiences it as bad faith. It is not usually bad faith. It is what the structure produces when one side departs from a status quo the other had also accepted.
There is a second regularity, and it is about sequence rather than substance. The order in which terms are conceded shapes what can be reopened later, because each concession establishes a new status quo that the next departure is measured against. Winning an early clause cheaply frequently costs more than losing it would have, precisely because it moves the reference point the rest of the document is judged from.
Subtract the counterfactual before crediting your negotiation. Terms that improved in a quarter when three funds were competing improved because three funds were competing. The test is whether you have ever improved a term in a process with one bidder.
What this ensemble cannot see
All three lenses treat the negotiation as a two-party sequence over a document. It is neither two-party nor only about the document.
The investor is negotiating against their own partnership, their fund’s stage rules, and a portfolio construction they cannot show you. A term they will not move on is frequently one they cannot move on, and the tree you build treats that as a choice they are making. From outside, an unmovable constraint and a firm negotiating position look identical.
There is also a real weakness in the central model, and it is stated in its own card. Theory of Moves is a minority framework whose author concedes it defines a new game rather than solving the classical one. It is used here because it describes the situation better than the alternatives, not because it is established. Someone who wanted to reject the analysis on those grounds would have a fair point.
And one property none of these models contains: how you negotiate is itself information about what you will be like to work with for the next eight years. A tree that terminates one step better and leaves the relationship worse has optimised the wrong thing, and nothing here can price that.
The one action that survives the ignorance: before you send the counter, write the four steps and count your remaining weeks. If the sequence ends below where you are now, or if the weeks are fewer than theirs and the tree is symmetric, do not send it. Sign, or walk, on what is in front of you.
Who has to move
The person who needs this is the founder holding the document, usually late, usually alone, and usually with one clause that has been irritating them for a day. The cheapest first test is the ten-minute sequence written before replying. If the terminal cell is worse than the status quo, you have saved the round, and it cost ten minutes.
Sources and notes
- Steven J. Brams, Rational Politics: Decisions, Games, and Strategy. Nonmyopic equilibria and the theory of moves are section 6.6, including the four rules defining a sequential game from an initial outcome, the backward-induction analysis of the resulting game tree, the definition of a nonmyopic equilibrium as one in which neither player perceives a long-term advantage from departing, and the author’s own statement that since a game is often defined to be the totality of the rules describing it, the rules given in fact define a new game. The symmetric case in which neither player can be said a priori to hold out longer, which Brams therefore leaves designated ambiguously, is in the same section.
- Matthew A. Cronin, Cleotilde Gonzalez and John D. Sterman, Why don’t well-educated adults understand accumulation? A challenge to researchers, educators, and citizens, Organizational Behavior and Human Decision Processes 108(1), 2009, pages 116 to 130. Author copy: https://www.mit.edu/~jsterman/CroninGonzalezSterman061210.pdf. The abstract states that highly educated people are often unable to infer the behaviour of simple stock-flow systems, and that persistent poor performance is not attributable to an inability to interpret graphs, contextual knowledge, motivation, or cognitive capacity.
- Jeffrey Carpenter and Andrea Robbett, Game Theory and Behavior, MIT Press. Non-cooperative bargaining is chapter 14, including the alternating-offers structure in which the division of a surplus is determined by relative patience and by who moves, rather than by any fairness criterion. Used here for the underlying claim in section 2 that staying power rather than argument determines the split.
A note on citing a minority framework. Theory of Moves is not standard equilibrium analysis and its author says so. It earns its place here because the classical normal-form game, in which both sides choose once and simultaneously, is a poor description of a live document either party can reopen. Where a framework fits the situation better than the mainstream one, the honest move is to use it and label it, which is what the transparency card above does.
Joshua Agonya Pi’Rwot, Founder.