2The Proprietary Market-Making Firm

In July 2024 the board of a listed exchange-traded-product market maker announced that it would stop paying regular dividends. The money was to stay in the firm as trading capital: the capital its prime brokers require it to keep with them before it may trade, and which sets how much it can trade. Eighteen months later that capital stood at € 1 044 million, a third more than a year earlier, and the firm had borrowed $200 million on top. A proprietary trading firm has no clients’ money and no fees; everything it can trade is what its owners have left in it, and what it can borrow against that. How much to leave in, and how much to spend each year to stay fast enough to earn anything, are the two decisions this chapter models.

2.1 Ownership: partnerships, private companies and listed firms

A proprietary trading firm (One Quant Book 1, chapter 1) is owned by the people who founded it and those they have admitted since; a few are listed. Ownership decides three things: who receives the profit, who can withdraw capital and when, and who bears a loss.

Private companies owned by their founders and staff work the same way in substance: profit is distributed by dividends or by bonuses instead of allocations, and capital is equity instead of members’ capital. A listed firm adds outside shareholders who expect a dividend policy and a share price, which is why its decisions to retain profit are announced, as in the hook, and a private firm’s are not.

The members’ agreement fixes each member’s points, the share of profit allocated to them, and how much of an allocation may be drawn. A departing member’s capital is typically repaid over several years, so that a departure does not drain the trading capital at once. That delay is part of the firm’s capital planning: a departure is a withdrawal of capital scheduled in advance.

2.2 Capital: what it is for and how much is required

A proprietary firm holds capital for three masters: its regulator, its clearing firms and prime brokers, and itself. The regulator’s requirement protects the market from the firm’s failure; the clearing firm’s and the prime broker’s protect them from the firm’s losses before they can close it out; the firm’s own buffer keeps it trading through the losses it expects.

The European design ties a small firm’s regulatory capital to its cost base, not to its risk. The chapter’s firm has fixed overheads of $95 million a year (people $60 million, technology $35 million) and trades $10 billion a day of cash securities: its fixed overheads requirement is $23.75 million, its K-DTF $10 million, its permanent minimum $0.75 million. The binding requirement is the fixed overheads one (Figure 2.2): a firm that hires or buys servers raises its regulatory capital before it has traded a share more.

The requirement that binds a market maker’s growth is usually not the regulator’s. Its prime brokers and clearing firms set their own requirements, by internal haircut and margin models, and a market maker must post capital with them before it can hold positions; the listed market maker of the hook calls this its trading capital and plans it daily. In the chapter’s model the prime brokers require capital of 2% of the firm’s daily volume: $200 million supports $10 billion a day, eight times the regulatory requirement.

def own_funds_requirement(fixed_overheads, dtf_cash, dtf_deriv, npr, permanent=0.75,
                          k_dtf_cash=0.001, k_dtf_deriv=0.0001):
    """Highest of the permanent minimum, a quarter of the preceding year's fixed overheads, and the K-factor
    requirement (coefficients on the average daily trading flow in cash and derivatives, plus a net-position
    charge npr). Returns the requirement and which of the three binds."""
    parts = {"permanent": permanent, "fixed overheads": 0.25 * fixed_overheads,
             "K-factors": k_dtf_cash * dtf_cash + k_dtf_deriv * dtf_deriv + npr}
    which = max(parts, key=parts.get)
    return parts[which], which

2.3 Reinvestment and payout

The model of firm.partnership (Listing 2.2) runs a firm year by year: trading capital sets volume up to what the market can give it; capture per unit traded depends on its technology; profit after fixed costs, technology spending and variable pay is split into payout and retention. The chapter’s firm starts with $200 million, captures one basis point at technology parity, pays 25% of its result as variable pay, and its markets can give it at most $20 billion a day, which $400 million of capital supports. In its first year it trades $10 billion a day, earns $250 million, spends $35 million on technology and $60 million on everything else, and makes $116.25 million.

    for t in range(p.years):
        spend = p.spend_share * tech
        rel = tech / front
        vol = min(cap / p.margin, p.market * (1 + p.market_growth) ** t)
        c = float(capture(rel, p.c0, p.eta))
        rev = c * vol * p.days
        pre = rev - p.people - spend
        prof = pre - p.var_pay * max(pre, 0.0)
        r = retention(t, cap, p) if callable(retention) else retention
        pay = (1 - r) * max(prof, 0.0)
        cap += prof - pay
        for k, v in (("capital", cap), ("volume", vol), ("capture", c), ("revenue", rev), ("spend", spend),
                     ("profit", prof), ("payout", pay), ("rel", rel)):
            out[k][t] = v
        tech = (1 - p.d) * tech + spend
        front = (1 + p.g) * front
    return out

At the chapter’s parameters the best fixed retention ratio is 20%, worth $939 million of discounted payout over ten years at 10% (Figure 2.3); at 20% the firm fills its market in year 6, at 50% in year 3. Retaining everything until capital reaches $400 million and paying out everything afterwards is worth $1 210 million: the partners give up two years of payout and gain $271 million of value.

A departure is a retention decision the firm did not choose. If the member with 30 points leaves at the end of year 5 and is repaid over three years, the firm (retaining 20%) has the capital to fill its market by the end of year 8 instead of year 6. Staggered repayment is what makes the departure survivable; a members’ agreement that let a senior partner withdraw everything at once would give that partner the power to stop the firm growing.

2.4 The technology treadmill

A market maker’s capture depends on how fast it is compared with the others (One Quant Book 11, chapter 9): who reaches a stale quote first, whose quote is picked off. Speed is relative, so standing still is falling behind.

At $g=15\%$ and $d=20\%$ the firm must spend 35% of its technology stock every year, $35 million in year 1 and $123 million in year 10. If it spends 25%, its relative technology, and with it its capture, falls to 0.44 of parity by year 10 (Figure 2.4). With a market that does not grow, the treadmill eventually wins either way: the parity spender’s profit reaches zero in year 20, when spending that grows at 15% a year catches a revenue that cannot grow; the lean spender’s in year 18, from falling capture (Figure 2.5). With a market growing 10% a year the parity spender stays profitable until year 56.

2.5 Tutorial: ten years of a partnership

What to change next. Let the market grow (exercise 7); give the prime brokers a stricter margin in a volatile year and watch the target capital move.

2.6 Build: the partnership model

2.7 Exercises

2.8 Problem: Keep It or Pay It Out

2.9 Interview questions