---
title: "Multi-Manager Platforms"
book: "The Industry: Firms, Roles and Careers"
subject: quant
language: en
chapter: 5
exercises: 8
source: https://one-course.com/books/quant/17/en/chapter/5-multi-manager-platforms
---

# Chapter 5 — Multi-Manager Platforms

In January 2016 the largest multi-manager platform’s registered adviser reported 1 625 employees to the SEC; ten years later it reported 6 140, and its home page counted more than 360 investment teams. A second platform grew from 287 employees to 2 063 over the same decade. No part of the industry has hired as fast. For the people it hires — portfolio managers, analysts, researchers, and the engineers and risk managers around them — a platform is a job with a known exit condition: a team that loses more than a set fraction of the capital it was given is cut, and then closed. This chapter describes the platforms as employers, reads their growth in their filings, and measures what a drawdown rule means for how long a team lasts.

## 5.1 The platform from the employee’s side

Book 16 (chapter 3) described the platform’s economics: pods that each run a share of one fund’s capital, costs passed through to investors, payouts on each pod’s own profit, a centre book that hedges what the pods hold in common. From the employee’s side the same structure looks like this.

- **A team is a small business inside a large one.** A portfolio manager hires analysts and a researcher or two, agrees a capital allocation and risk limits, and is paid a share of the team’s own profit after its costs.
- **Everything else is provided.** Data, execution, technology, risk systems, compliance, operations and financing are the platform’s; a team uses them and pays for them through its costs.
- **The limits are the contract.** A team’s drawdown from its peak, measured on the capital it was allocated, triggers a cut in capital and then the team’s closure (the de-risking ladder of Book 16, chapter 8).
- **The platform employs many people who are not in teams** : the risk, technology and data staff who run the common machinery, and the centre-book traders.

The platforms named in this chapter describe themselves in these terms: a firm with a “multi-strategy, multi-PM approach” whose “165 investment teams span six strategies”; one that has grown “from a proprietary trading firm with deep systematic roots into a multi-strategy, multi-manager hedge fund”; “a global, multi-strategy, multi-manager investment firm”.

## 5.2 Size: assets, employees and teams

The platforms’ US advisers file Form ADV (chapter 4), and the SEC’s files of past months keep the answers, so a platform’s growth can be read from its own filings.

**As of January 2026 — Four platforms in their filings and on their pages.**

| registered adviser | Jan 2016 | Jan 2017 | Dec 2022 | Dec 2024 | Jan 2026 |
| --- | --- | --- | --- | --- | --- |
| Millennium Management, employees | 1 625 | 1 830 | 3 985 | 5 550 | 6 140 |
| RAUM ($ billion) | 181.5 | 207.9 | 341.0 | 505.9 | 571.1 |
| Balyasny Asset Management, employees | 287 | 395 | 1 048 | 1 857 | 2 063 |
| RAUM ($ billion) | 29.0 | 36.0 | 181.0 | 248.0 | 265.2 |
| Schonfeld Strategic Advisors, employees | – | 89 | 620 | 952 | 872 |
| ExodusPoint Capital Management, employees | – | – | 688 | 647 | 634 |

The firms’ own pages: Millennium, “7 000+ employees globally”, “360+ investment teams”, “$97BN+ AUM”; Balyasny, “165 investment teams” in six strategies, “2000+ investment team and support professionals”, “$37B assets under management”. RAUM is gross (chapter 4); the firms’ AUM figures are net.

The filings tell three things the firms’ pages do not ([Figure 5.1](#fig-in-multi-manager-platforms-growth)). Growth was fast but not uniform: the largest adviser’s staff grew 3.8 times in ten years, a second’s 7.2 times; a younger platform’s staff fell slightly after 2022. Assets grew about as fast as staff: the largest adviser’s RAUM per employee fell by 17% over the decade, the second’s rose by 27%. And the share of staff in advisory functions fell at both, from 63% to 47% and from 64% to 42%: the platforms built their common machinery faster than their teams. For a candidate, that is the shape of the job market inside a platform: more of the new positions are in technology, data and risk than in the teams.

![Employees reported on Form ADV by four platforms’ US advisers, in the SEC’s files of January 2016, January 2017, December 2022, December 2024 and January 2026 (no file between 2017 and 2022 was read; the lines join the points read). Data: SEC adviser files, item 5.A, through in_platforms.panel.](https://one-course.com/images/onecourse/chapters/quant-17/in-multi-manager-platforms/fig-da58ea468ff9.svg)

***Figure 5.1.** Employees reported on Form ADV by four platforms’ US advisers, in the SEC’s files of January 2016, January 2017, December 2022, December 2024 and January 2026 (no file between 2017 and 2022 was read; the lines join the points read). Data: SEC adviser files, item 5.A, through `in_platforms.panel`.*

## 5.3 Team turnover: what the public record shows

How many teams a platform hires and closes each year is the number a portfolio manager most wants and the one no platform publishes. The press reports hirings and departures of individual teams, and occasionally a total for a year; none of it is a primary source, and this book does not print it. What can be said from public records is structural.

**Definition 5.1 (Team tenure, team turnover rate).**

A team’s *tenure* is the time from its start on a platform to its closure or departure. A platform’s *team turnover rate* is the number of teams that close or leave in a year divided by the average number of teams running during it.

The two are linked: if teams left at a constant rate $\lambda$ a year, [tenure](#def-in-multi-manager-platforms-tenure) would be exponential with median $\ln 2/\lambda$, and a platform with a turnover rate of 20% would see half of each year’s new teams leave within about 3.5 years. The rate is not constant in practice: a new team is most at risk in its first year, before its gains give it room below the peak. The next section computes the curve.

## 5.4 Drawdown limits and how long a team lasts

**Method 5.2 (Tenure under a drawdown ladder).**

1. Give each team a true Sharpe ratio and a volatility of P&L on its allocated capital.
2. Run its daily P&L; halve its capital when its drawdown from the peak since it started exceeds the cut level, and close it when the drawdown exceeds the stop level.
3. Record the time of closure, or the horizon if it survives.
4. Estimate the survival curve and the median [tenure](#def-in-multi-manager-platforms-tenure) ; for a mix of teams, the turnover rate and the share of closed teams that were skilled.

The chapter’s teams (illustrative, not any platform’s terms) run 10% annual volatility on their capital. Two ladders are compared: a tight one that halves capital after a 5% drawdown and closes the team after 7.5%, and a loose one at 10% and 15%. The drawdown is measured since the team started, so a team that has made money has room.

| ladder, true Sharpe | median [tenure](#def-in-multi-manager-platforms-tenure) | closed in 10 years | alive at 1 year | at 3 years | turnover |
| --- | --- | --- | --- | --- | --- |
| tight, 0.0 | 0.9 years | 99.6% | 45.6% | 15.9% | 63% |
| tight, 0.5 | 1.5 years | 94.4% | 60.2% | 33.9% | 33% |
| tight, 1.0 | 3.6 years | 74.8% | 72.2% | 53.5% | 16% |
| tight, 1.5 | over 10 years | 47.9% | 81.8% | 69.5% | 7% |
| loose, 0.5 | over 10 years | 41.4% | 96.4% | 79.1% | 6% |
| loose, 1.0 | over 10 years | 15.3% | 98.7% | 92.8% | 2% |

Under the tight ladder even a good team is more likely than not to be closed within ten years: a team with a Sharpe ratio of 1.0 has a median [tenure](#def-in-multi-manager-platforms-tenure) of 3.6 years, and one of 0.5 of 1.5 years ([Figure 5.2](#fig-in-multi-manager-platforms-surv)). The loose ladder keeps most good teams for a decade. In a platform whose teams are spread evenly across Sharpe ratios of 0, 0.5, 1.0 and 1.5, the tight ladder closes 20% of teams a year and the loose one 5%; 35% of teams are closed in their first year under the tight ladder, 3% under the loose one.

![Survival of simulated teams under two drawdown ladders: the share still running after each year, 4 000 teams per curve, annual volatility 10% of allocated capital. Tight ladder: capital halved at a 5% drawdown from the peak since the start, team closed at 7.5%; loose ladder: 10% and 15%. Illustrative parameters. Data: firm.teamtenure.stop_times, through in_platforms.times.](https://one-course.com/images/onecourse/chapters/quant-17/in-multi-manager-platforms/fig-8a80cd66bfce.svg)

***Figure 5.2.** Survival of simulated teams under two drawdown ladders: the share still running after each year, 4 000 teams per curve, annual volatility 10% of allocated capital. Tight ladder: capital halved at a 5% drawdown from the peak since the start, team closed at 7.5%; loose ladder: 10% and 15%. Illustrative parameters. Data: `firm.teamtenure.stop_times`, through `in_platforms.times`.*

The ladder cannot tell skill from luck quickly. Of the teams the tight ladder closes, 39% had a true Sharpe ratio of 1.0 or more; under the loose ladder, 14%. The halving step matters: without it, a team with a Sharpe ratio of 0.5 lasts a median 0.7 years under the tight stop instead of 1.5, because the cut halves the size of the losses that follow a bad start. A platform chooses where to stand between closing bad teams fast and keeping good ones, and a portfolio manager joining it should ask where that is.

**Remark 5.3 (What the model leaves out).**

Real platforms review teams on more than drawdowns: returns against the allocated risk, crowding with other teams, behaviour under limits. Teams also leave on their own, for another platform or to start a fund (Book 16, chapter 25). Capital is reallocated upward after gains. The model isolates the one rule a team signs up to; it is not a forecast of any platform’s turnover, and the teams’ volatility and Sharpe ratios are chosen for illustration.

## 5.5 How platforms hire

A platform hires a portfolio manager for a record and a plan. The record is the manager’s past P&L with the risk taken to earn it, which Book 16 (chapter 25) calls portable only if it can be attributed and verified; the plan is the capital the manager asks for, the strategy, the team and the budget. Analysts and researchers are hired by the portfolio managers, often from other platforms’ teams or from banks’ desks; engineers and risk staff by the platform itself. Because every team’s economics are its own, a platform’s hiring runs as a market: teams compete for capital and people, and a team that is closed releases both.

For an analyst the consequence is that the employer is, in practice, the team: when the team closes, the job usually ends with it, unless another team hires the analyst. [Tenure](#def-in-multi-manager-platforms-tenure) under a ladder is therefore also the analyst’s job security, and the table above is a reason to ask about it before joining.

## 5.6 Tutorial: growth in the filings and tenure in the model

**Goal.** Read four platforms’ growth from their Form ADV filings, and compute [team tenure](#def-in-multi-manager-platforms-tenure) under two ladders. **End state:** the dated box, [Figure 5.1](#fig-in-multi-manager-platforms-growth), the [tenure](#def-in-multi-manager-platforms-tenure) table and [Figure 5.2](#fig-in-multi-manager-platforms-surv).

1. **The panel.** `in_platform_derive.py` reads five of the SEC’s adviser files (2016 to 2026, xlsx and CSV formats) with `firm.formadv.read(path, hedge_only=False)` and keeps the four advisers by their CRD numbers; `in_platforms.growth` compares two dates.
2. **One team.** `firm.teamtenure.stop_times` runs a batch of teams under a ladder ([Listing 5.1](#lst-in-multi-manager-platforms-stop)); returns come from Book 16’s `firm.podshop.pods`. `def stop_times (sr, vol, ladder, n, years, rng, days=252 ): daily = ps.pods(n, years, sr, vol, 0.0 , rng, days)[" daily " ] size = np.ones(n) eq = np.zeros(n) peak = np.zeros(n) out = np.full(n, np.inf) for t in range (daily.shape[0 ]): alive = np.isinf(out) eq += np.where(alive, size * daily[t], 0.0 ) peak = np.maximum(peak, eq) dd = peak - eq size = np.where(alive & (size == 1.0 ) & (dd > ladder.cut), 0.5 , size) out = np.where(alive & (dd > ladder.stop), (t + 1 ) / days, out) return out` **Listing 5.1.** The ladder applied day by day: halve at the cut, close at the stop, both measured from the peak since the team started. code/firm/teamtenure/firm_teamtenure.py
3. **The table.** `in_platforms.table()` gives the median [tenure](#def-in-multi-manager-platforms-tenure) , the share closed, survival at one and three years and the turnover rate for each ladder and Sharpe ratio; `mixed()` the platform-level figures.
4. **Check the model.** Run the tight ladder with four times as many steps a year: the median [tenure](#def-in-multi-manager-platforms-tenure) of a team with a Sharpe ratio of 0.5 stays within 10% of 1.5 years. Remove the halving step: it falls to 0.7 years.

**What to change next.** Raise the teams’ volatility to 15% and see how the tight ladder’s turnover changes (exercise 7); give teams a Sharpe ratio that falls to zero after two years and ask how fast each ladder closes them.

## 5.7 Build: the tenure model

**Purpose.** Measure what a drawdown rule does to how long teams last, for this chapter and for chapter 22’s portfolio-manager deal.

**Interface.** `firm.teamtenure`: `Ladder(cut, stop)`; `stop_times(sr, vol, ladder, n, years, rng, days)`; `survival(times, grid)`; `median_tenure(times)`; `turnover(times, years)`; `false_cut_share(times_by_sr, skilled)`. Wraps `firm.podshop.pods`.

**Rules.** Drawdowns are measured on the capital actually run, from the peak since the start; a closed team stays closed; teams alive at the horizon are censored, not closed.

**Acceptance tests.** `code/firm/teamtenure/tests/`: survival, median and turnover on known times; more skill and a looser ladder mean fewer closures; the false-cut share on constructed times.

**Stretch.** Capital reallocation after gains; teams that leave on their own at a constant rate; a Sharpe ratio that decays with the team’s age.

Sources and further reading

- SEC, Information about registered investment advisers: monthly files of January 2016, January 2017, December 2022, December 2024 and January 2026 (Internet Archive copies).
- The platforms’ own pages: Millennium, Balyasny, Schonfeld, ExodusPoint (ledger F2–F5).
- Book 16, chapters 3, 8 and 25, for the platform’s economics, the de-risking ladder and track records.

## 5.8 Exercises

**Exercise 5.1 ★.**

From the dated box, compute the growth of the largest adviser’s employees and RAUM between January 2016 and January 2026.

**Solution of Exercise 5.1.**

Employees $6\,140/1\,625=3.78$ times; RAUM $571.1/181.5=3.15$ times.

**Exercise 5.2 ★.**

If teams leave a platform at a constant 20% a year, what is the median [tenure](#def-in-multi-manager-platforms-tenure) of a team?

**Solution of Exercise 5.2.**

$\ln2/0.2=3.5$ years.

**Exercise 5.3 ★.**

A platform’s page counts 360 teams and 7 000 employees. What does that suggest about the share of staff outside the teams, if a team averages five people?

**Solution of Exercise 5.3.**

$360\times5=1\,800$ people in teams, about 26% of 7 000: roughly three quarters of the staff run the common machinery.

**Exercise 5.4 ★★.**

Compute the advisory share of the second platform’s adviser in January 2016 and January 2026, and say what the change means for a candidate.

**Solution of Exercise 5.4.**

$183/287=64\%$ in 2016 and $869/2\,063=42\%$ in 2026. Most new positions were outside advisory functions: technology, data, risk and operations grew faster than the teams.

**Exercise 5.5 ★★.**

Under the tight ladder, what share of teams with a Sharpe ratio of 1.0 are still running after three years? After how long is half of them closed?

**Solution of Exercise 5.5.**

53.5% after three years; half are closed after 3.6 years.

**Exercise 5.6 ★★.**

Read [Figure 5.2](#fig-in-multi-manager-platforms-surv): at which age does the gap between the tight and loose ladders for teams with a Sharpe ratio of 0.5 exceed 40 percentage points?

**Solution of Exercise 5.6.**

At 1.25 years: 94.5% of the loose ladder’s teams are running against 54.2% of the tight ladder’s.

**Exercise 5.7 ★★★.**

*Coding.* Rerun `stop_times` for a Sharpe ratio of 1.0 under the tight ladder with a volatility of 15% instead of 10%. What happens to the median [tenure](#def-in-multi-manager-platforms-tenure), and why?

**Solution of Exercise 5.7.**

The median falls from 3.6 years to about 0.9 years, and 98.6% are closed within ten years: with the same Sharpe ratio, a higher volatility makes the same loss levels a smaller number of standard deviations, so they are reached sooner. The ladder is set in capital, not in units of the team’s risk.

**Exercise 5.8 ★★★.**

*Find the flaw.* “Under the tight ladder 39% of closed teams were skilled, so the platform should stop closing teams.”

**Solution of Exercise 5.8.**

The platform’s alternative is to keep bad teams longer: under the tight ladder most closed teams (61%) were not skilled. The share of skilled teams among those closed is the cost of speed; the right ladder weighs it against the losses of keeping bad teams, and its level should scale with each team’s volatility.

## 5.9 Problem: How Long Does a Team Last?

**Problem 5.1.**

Weekend problem — how long does a team last?

A portfolio manager is offered a seat at a platform with a drawdown ladder. She believes her strategy’s Sharpe ratio is between 0.5 and 1.0 and wants to know her chances.

**Part I — The platform.**

1. List what a team provides and what the platform provides.
2. Give the largest adviser’s employees in January 2016 and January 2026, and the growth factor.
3. Give the change in its RAUM per employee and in its advisory share over the decade.
4. What do the two platforms’ pages say about their number of teams?
5. Why does the chapter not print the press’s counts of team closures?

**Part II — [Tenure](#def-in-multi-manager-platforms-tenure).**

6. Define [team tenure](#def-in-multi-manager-platforms-tenure) and the turnover rate, and relate the median [tenure](#def-in-multi-manager-platforms-tenure) to a constant rate $\lambda$ .
7. State the two ladders and the teams’ volatility.
8. Give the median [tenure](#def-in-multi-manager-platforms-tenure) for Sharpe ratios 0.5 and 1.0 under the tight ladder.
9. Give the share of such teams closed within ten years under each ladder.
10. Why is a new team most at risk in its first year?

**Part III — The platform’s view.**

11. Give the turnover rate of the mixed platform under each ladder.
12. Give the share of teams closed in their first year under each ladder.
13. Give the share of closed teams that were skilled (Sharpe ratio 1.0 or more) under each ladder.
14. What does the halving step do to the [tenure](#def-in-multi-manager-platforms-tenure) of a team with a Sharpe ratio of 0.5?
15. Check the model: what happens to that median when the step is divided by four?

**Part IV — The verdict.**

16. State the *named result* : median [tenure](#def-in-multi-manager-platforms-tenure) under the tight ladder for Sharpe ratios of 0.5 and 1.0, and the share of closed teams that were skilled under each ladder.
17. What three things should she ask the platform before joining?
18. Why is the [tenure](#def-in-multi-manager-platforms-tenure) also her analysts’ job security?
19. What does the model leave out that could lengthen or shorten her [tenure](#def-in-multi-manager-platforms-tenure) ?
20. In two sentences, what should she conclude?

**Solution of Problem 5.1.**

1. Team: strategy, people, trading decisions within limits. Platform: capital, data, execution, technology, risk, compliance, operations, financing.
2. 1 625 and 6 140; 3.78 times.
3. RAUM per employee fell by 17% (a factor of 0.83); advisory share from 63% to 47%.
4. More than 360 investment teams; 165 teams in six strategies.
5. No primary source: press counts are not filed or published by the platforms.
6. Time from start to closure; closures in a year over the average number of teams; median $\ln2/\lambda$ .
7. Tight: halve at 5%, close at 7.5%; loose: 10% and 15%; volatility 10%.
8. 1.5 years and 3.6 years.
9. Tight: 94.4% and 74.8%; loose: 41.4% and 15.3%.
10. It has no gains yet, so its peak is its starting capital and any early loss is a drawdown.
11. 20% and 5%.
12. 35% and 3%.
13. 39% and 14%.
14. It lengthens it: without halving the median is 0.7 years instead of 1.5.
15. It stays within 10% of 1.5 years (1.6 years with four times as many steps).
16. Tight ladder: 1.5 and 3.6 years; skilled share of closed teams 39% (tight) and 14% (loose).
17. The ladder’s levels, whether they scale with the strategy’s volatility, and how capital is added after gains.
18. Analysts are employed by the team, and their jobs usually end with it.
19. Reviews on other criteria, capital added after gains, teams leaving on their own.
20. With a Sharpe ratio of 0.5 to 1.0 under a tight ladder she should expect two to four years; she should negotiate the ladder or the size so that her volatility makes the stop levels several standard deviations away.

## 5.10 Interview questions

**Interview question 5.1 ★ trader, researcher.**

What is a multi-manager platform, and how is working in one of its teams different from working at a single-strategy fund?

**Solution of Interview question 5.1.**

A fund whose capital is run by many independent teams, each with its own limits and pay, on common infrastructure. In a team you run a small business under the platform’s limits and are paid on your team’s P&L; at a single-strategy fund you contribute to one book and are paid on the firm’s result.

*What the interviewer is looking for: the team as the unit of pay and risk.*

**Interview question 5.2 ★ researcher, risk.**

A team runs 10% annual volatility with a Sharpe ratio of 1. What is the probability that it is down 5% or more after its first year?

**Solution of Interview question 5.2.**

The year’s P&L is about normal with mean 10% and standard deviation 10%: $P(X\le-5\%)=\Phi(-1.5)=6.7\%$.

*What the interviewer is looking for: mean, standard deviation, a z-score.*

**Interview question 5.3 ★★ risk.**

Why does a platform halve a team’s capital after a first drawdown rather than close it at once?

**Solution of Interview question 5.3.**

A first drawdown is weak evidence: halving limits the loss if the team is bad while keeping it alive if it was unlucky, and doubles the distance to the stop in units of the remaining risk.

*What the interviewer is looking for: evidence against luck, and the value of a smaller position.*

**Interview question 5.4 ★★ researcher.**

A platform closes 20% of its teams every year. What can you infer about the skill of the teams that remain?

**Solution of Interview question 5.4.**

Survivors are selected on realised P&L, so they are better than the teams hired, but mixed with lucky ones; the surviving pool’s average skill rises with time, and its measured returns overstate its skill (survivorship bias, Book 7, chapter 3).

*What the interviewer is looking for: selection on outcomes, and its bias.*

**Interview question 5.5 ★★ trader.**

As a new portfolio manager, how would you size risk in your first months under a drawdown ladder, and why?

**Solution of Interview question 5.5.**

Small at first: until gains build room below the peak, a normal loss can reach the stop. Scale up as the cushion grows, keeping the stop several standard deviations of expected P&L away.

*What the interviewer is looking for: sizing to distance from the stop.*

**Interview question 5.6 ★★★ researcher, risk.**

A team’s P&L is a Brownian motion with drift $\mu>0$ and volatility $\sigma$. Show that a stop at a loss $d$ below its starting capital catches it with probability $e^{-2\mu d/\sigma^2}$, while a stop at $d$ below its running peak catches it with probability one. What does that mean for a ladder measured from the peak?

**Solution of Interview question 5.6.**

From the start: the probability that $\mu t+\sigma W_t$ ever reaches $-d$ is $e^{-2\mu d/\sigma^2}$ (Book 4). From the peak: the drawdown is a reflected Brownian motion with drift towards zero, which is recurrent and so reaches any level $d$ with probability one. A ladder measured from the peak closes every team eventually; only its timing depends on skill.

*What the interviewer is looking for: the two probabilities, and why the peak-based stop always triggers.*