Certifications, training and prop firms: what each one certifies and what a challenge measures.

What each program certifies

The job postings in lessons 7.5.1 and 7.5.2 often mention degrees and certifications. Each one certifies a specific syllabus and exam, with different requirements. The table summarizes what their official pages say, as consulted on October 23, 2026; dates and prices change, so the current page always prevails.

What the table cannot say is which one is “worth more”, and this lesson does not say it: it depends on the job. As a guide by content, without rating any program: the CQF and the EPAT cover quantitative and trading methods, with parts that this course covers in levels 2 to 5; the CFA is a broad program in investment analysis and portfolio management, with an experience requirement; the FRM focuses on risk measurement and management (the topics of module 4.2 and more). None replaces what lesson 7.5.3 asks for: your own work that someone else can review.

Master's degrees and continuing education

A university master's degree in Spain can be an official degree (título oficial: accredited and valid throughout the country) or a university's own degree (título propio). Both can be useful; the difference matters for admission to a doctorate and for some civil-service competitive exams and public hiring processes. The way to check is the RUCT, the Ministry's Registry of Universities, Centers and Degrees: if an official master's degree does not appear there, it is not official. To compare programs with quantitative content, the course syllabi (guías docentes), which detail subjects, hours and assessment, are useful, and so is the faculty: how many publish research or work in the industry.

A quant's continuing education does not end with a degree: reading papers (with the caution of lesson 2.1.8 and the “factor zoo”), the documentation of the libraries you use, regulatory changes (module 7.4) and your own projects. This course's capstone is a good example: a system defended before a panel teaches more than a certificate.

The independent trader, in numbers

Trading your own money does not require any certification, but it does require capital, time and an edge that survives costs (lesson 0.1.2). With an annual Sharpe $S$ and an annual volatility $\sigma$, the expected return is $\mu = S\sigma$. If the annual return were normal, the probability of a losing year would be

which does not depend on volatility. The following figures use explicit assumptions: 15% volatility and a target of €30,000 of expected profit per year, before taxes (lesson 7.4.6).

import numpy as np
import pandas as pd
from scipy import stats

VOL = 0.15            # assumption: annual volatility
OBJETIVO = 30_000     # assumption: expected annual profit, in euros

filas = []
for sharpe in [0.5, 1.0, 1.5]:
    mu = sharpe * VOL
    filas.append({
        "sharpe": sharpe,
        "rent_esperada": mu,
        "capital": OBJETIVO / mu,
        "p_año_perdidas": stats.norm.cdf(-sharpe),
        "peor_año_1_de_20": mu + stats.norm.ppf(0.05) * VOL,
    })
tabla = pd.DataFrame(filas).set_index("sharpe")
tabla.round({"rent_esperada": 3, "capital": -3, "p_año_perdidas": 3,
             "peor_año_1_de_20": 3})

        rent_esperada   capital  p_año_perdidas  peor_año_1_de_20
sharpe
0.5             0.075  400000.0           0.309            -0.172
1.0             0.150  200000.0           0.159            -0.097
1.5             0.225  133000.0           0.067            -0.022

The columns are the expected return (rent_esperada), the capital, the probability of a losing year (p_año_perdidas) and the worst year in twenty (peor_año_1_de_20). With a Sharpe of 1, very good for a real strategy after costs, you need €200,000 to expect €30,000 a year, one year in six closes at a loss and one year in twenty loses almost 10%. With Sharpe 0.5, the required capital doubles and almost one year in three is negative. And the backtest Sharpe is not the future one: with one year of data, its standard error is of the order of 1 (lesson 7.5.3). The risk of ruin from lesson 4.2.1 and the drawdown control from 4.2.5 are not optional when the capital is your own.

Prop firms: two businesses with the same name

As seen in lesson 7.5.2, “prop firm” refers to two things. One is the proprietary trading firm: it risks its own capital, hires traders and, if it is a member of an exchange or does market making, needs authorization. The other sells evaluations (challenges) to retail customers: the customer pays a fee, trades in a simulated account under a set of rules and, if they comply with them, receives a “funded” account and a percentage of the profits.

stateDiagram-v2
  [*] --> Evaluation: pays the fee
  Evaluation --> Failed: breaks a rule
  Evaluation --> Funded: reaches the target
  Funded --> Payout: profit
  Payout --> Funded
  Funded --> Lost: breaks a rule
  Failed --> [*]
  Lost --> [*]

On June 6, 2019, the CNMV (Comisión Nacional del Mercado de Valores, Spain's securities market regulator) published a warning about “funded trading accounts” linked to training courses. It said it had received inquiries and complaints; it described the scheme (a paid course, sometimes costing “several thousand euros”, tests in a simulated environment with limits such as the maximum daily loss, and the promise of trading with the website's capital in exchange for a percentage of the profits) and warned of the risk of “fraud or deception” regarding access to the accounts. It added that giving those courses or opening those accounts “do not fall within the CNMV's scope of action”, although the activities carried out from them in the financial markets do. The notice is in Spanish; the quotations are our unofficial translation.

A simulated challenge with Monte Carlo

The model follows the scheme the CNMV describes with rules that are common in the offers, but all parameters are assumptions of this lesson, not those of any company: a profit target of 8% of the account, a maximum daily loss of 5%, a maximum total loss of 10% of the starting balance, at least 5 trading days and a maximum of 30 days. The trader has a normal daily result with volatility $\sigma_d$ and a mean that corresponds to an annual Sharpe $S$: $\mu_d = S\,\sigma_d/\sqrt{252}$. The account does not compound: the result is measured on the starting balance.

REGLAS = {"objetivo": 0.08, "perdida_diaria": 0.05,
          "perdida_total": 0.10, "dias_min": 5, "dias_max": 30,
          "trailing": False}

def superar(sharpe, vol, reglas=REGLAS, n=50_000, rng=None):
    """Share of paths that reach the target before breaking
    a rule (normal daily result, no compounding)."""
    dias = reglas["dias_max"]
    r = rng.normal(sharpe / np.sqrt(252) * vol, vol, (n, dias))
    saldo = np.cumsum(r, axis=1)
    if reglas["trailing"]:   # maximum loss from the high-water mark
        caida = np.maximum.accumulate(np.maximum(saldo, 0), axis=1)
        caida = saldo - caida
    else:                    # maximum loss from the starting balance
        caida = saldo
    roto = (r <= -reglas["perdida_diaria"]) | (
        caida <= -reglas["perdida_total"])
    meta = (saldo >= reglas["objetivo"]) & (
        np.arange(1, dias + 1) >= reglas["dias_min"])
    primero = lambda m: np.where(m.any(1), m.argmax(1), dias + 1)
    return float((primero(meta) < primero(roto)).mean())

rng = np.random.default_rng(5)
vols = [0.005, 0.01, 0.015, 0.02, 0.03]
res = pd.DataFrame({f"{v:.1%}": [superar(s, v, rng=rng)
                                 for s in [0, 1, 2]]
                    for v in vols}, index=["S=0", "S=1", "S=2"])
res.round(3)

      0.5%   1.0%   1.5%   2.0%   3.0%
S=0  0.002  0.121  0.278  0.388  0.399
S=1  0.007  0.195  0.397  0.506  0.491
S=2  0.017  0.298  0.521  0.621  0.578

The figure at the top draws the same table with more volatilities. The reading is uncomfortable for anyone who shows up with a real edge and prudent risk: at 0.5% a day, reaching 8% in 30 days is almost impossible, edge or no edge, because the 30-day standard deviation is $0.5\% \cdot \sqrt{30} \approx 2.7\%$. The probability rises mainly with volatility: a trader with no edge at all who risks 2% a day passes the challenge far more often than one with Sharpe 2 who risks 0.5%. Around 2.5% a day it stops rising and then falls, because the 10% total loss is hit as quickly as the target. The rules select whoever takes risk, and the edge only shows at the margins.

Offers usually add rules that make the test harder. Two variants with the same assumptions: a two-phase challenge (8% and then 5%, 30 days each) and a trailing maximum loss, measured from the highest balance reached instead of from the starting balance:

fase2 = {**REGLAS, "objetivo": 0.05}
dinamica = {**REGLAS, "trailing": True}
filas = []
for s in [0, 2]:
    una = superar(s, 0.01, rng=rng)
    dos = una * superar(s, 0.01, fase2, rng=rng)
    din = superar(s, 0.01, dinamica, rng=rng)
    filas.append({"sharpe": s, "una_fase": una, "dos_fases": dos,
                  "trailing": din})
pd.DataFrame(filas).set_index("sharpe").round(3)

        una_fase  dos_fases  trailing
sharpe
0          0.118      0.036     0.115
2          0.303      0.167     0.302

At 1% a day, the second phase (dos_fases, against una_fase) divides the probability by more than three with no edge (from 11.8% to 3.6%) and almost by two with Sharpe 2 (from 30.3% to 16.7%): you need to be lucky twice. The trailing loss, on the other hand, changes almost nothing with these numbers (11.5% versus 11.8%): in 30 days, the balance rarely rises several points and then falls 10% from that high without first hitting the target. The actual terms of each company, which change often, decide which effects dominate; the model only shows the mechanics.

The economics of the challenge

For the company, each challenge sold brings in a fee; the payouts to traders who pass are its cost. If the funded accounts are also simulated, as in the scheme the CNMV describes, the payouts come out of the company's cash, which is fed by the fees. The business model works if fees exceed payouts, and the rules are the lever to achieve that.

The next simulation builds a complete market with more assumptions, all made up for the lesson: 1,000 challenges sold at 500 USD; 80% of buyers with no edge, 15% with Sharpe 0.5 and 5% with Sharpe 1.5; each one risks between 0.5% and 2% a day; whoever passes the challenge trades 60 days in the funded account (a 100,000 USD account, same rules), is paid 80% of the profit every 20 days and loses the account if they break a rule.

CUOTA, CUENTA, REPARTO = 500, 100_000, 0.80      # assumptions

def financiada(sharpe, vol, dias=60, cada=20, rng=None):
    """Payouts (USD) to a trader in the funded account."""
    r = rng.normal(sharpe / np.sqrt(252) * vol, vol, dias)
    saldo, pagado = 0.0, 0.0
    for d, x in enumerate(r, start=1):
        saldo += x
        if x <= -REGLAS["perdida_diaria"] or saldo <= -REGLAS[
                "perdida_total"]:
            break
        if d % cada == 0 and saldo > 0:
            pagado += REPARTO * saldo * CUENTA
            saldo = 0.0
    return pagado

rng = np.random.default_rng(8)
n = 1_000
sharpes = rng.choice([0.0, 0.5, 1.5], n, p=[0.80, 0.15, 0.05])
vol_tr = rng.uniform(0.005, 0.02, n)
pagos = np.zeros(n)
supera = np.zeros(n, bool)
for i, (s, v) in enumerate(zip(sharpes, vol_tr)):
    supera[i] = superar(s, v, n=1, rng=rng) == 1.0   # one attempt
    if supera[i]:
        pagos[i] = financiada(s, v, rng=rng)
neto = pagos - CUOTA
print(f"Pass the challenge:            {supera.mean():.1%}")
print(f"Lose the fee without payout:   {(pagos == 0).mean():.1%}")
print(f"End with a positive balance:   {(neto > 0).mean():.1%}")
print(f"Fees collected:   {CUOTA * n:>9,.0f} USD")
print(f"Paid to traders:  {pagos.sum():>9,.0f} USD")
for s in [0.0, 0.5, 1.5]:
    m = sharpes == s
    print(f"Sharpe {s}: {m.sum():3d} buyers, "
          f"mean net {neto[m].mean():+6.0f} USD")

Pass the challenge:            21.4%
Lose the fee without payout:   85.5%
End with a positive balance:   14.1%
Fees collected:     500,000 USD
Paid to traders:  1,196,663 USD
Sharpe 0.0: 832 buyers, mean net   +649 USD
Sharpe 0.5: 139 buyers, mean net   +888 USD
Sharpe 1.5:  29 buyers, mean net  +1157 USD

The mean of each group hides the shape of the distribution:

print(f"Median net:              {np.median(neto):+.0f} USD")
print(f"90th percentile of net:  {np.percentile(neto, 90):+.0f} USD")
orden = np.sort(pagos)[::-1]
top = orden[: n // 100].sum() / pagos.sum()
print(f"Payouts to the top-earning 1%: {top:.0%} of the total")
print(f"Break-even fee:          {pagos.mean():.0f} USD")
reparto = pd.cut(pagos, [-1, 0, CUOTA, np.inf],
                 labels=["no payout", "paid less than the fee",
                         "paid more than the fee"])
print(pd.Series(reparto).value_counts(normalize=True).round(3)
      .to_string())

Median net:              -500 USD
90th percentile of net:  +3248 USD
Payouts to the top-earning 1%: 18% of the total
Break-even fee:          1197 USD
no payout                 0.855
paid more than the fee    0.141
paid less than the fee    0.004

85.5% of buyers lose the fee without receiving anything: 78.6% do not pass the challenge and almost a third of those who pass lose the funded account before the first payout. The median outcome is exactly −500 USD, the fee. And yet the mean outcome of each group is positive, because a few are paid a lot: the top-earning 1% takes 18% of the payouts. It is the shape of a lottery: a small, almost certain loss and a large, rare gain. Edge changes the figures little: the no-edge group and the Sharpe 1.5 group (only 29 buyers, a small sample) differ less than two risk levels do.

The other side is the company's. With these assumptions it would pay 1,196,663 USD and collect 500,000: the break-even fee would be 1,197 USD per challenge. A real company that charges fees of that order or lower only balances its books with rules tougher than the model's, more limited payouts or buyers who take less risk than they would need to pass the test, which is what most of them do. For the buyer, the conclusion is the same in every case: the most likely outcome is losing the fee, and the probability of passing the challenge mostly measures how much risk was taken.

Risks and checks before paying

The simulations do not capture the non-statistical risks, which tend to weigh the most:

• Counterparty. Payouts depend on the solvency and the willingness of a company that does not hold client assets. If it changes its terms, closes or denies a payout over a rule open to interpretation, there is no compensation scheme to turn to.
• Changing terms. Rules, platforms, permitted instruments and payout schedules are set by contract and change. Keep the version of the terms in force when you paid.
• Your own conduct risk. The rules reward volatility: the temptation to increase risk to reach the target is exactly the opposite of what lessons 4.2.1 and 4.2.5 teach. A trader who gets used to trading that way carries it over to their own money.
• Accumulated cost. Buying the challenge again after each failure adds up fees; with a 12% probability of success per attempt (1% a day with no edge), the expected number of attempts until passing is 1/0.12 ≈ 8.
• Taxes. Payouts from a challenge company are not capital gains from a securities account; their tax treatment in Spain is not the one in lesson 7.4.6 and depends on the contract. For a specific case, consult a tax advisor.

Before paying, the checks are those of lesson 7.5.2: identify the company (legal name, registered address, entry in the commercial registry), find out whether it carries out any activity that requires authorization and whether it appears in any register, look it up in the CNMV's warnings and read the full payout terms. If a company also offers to manage client money or sell “signals”, it enters the territory of lesson 7.4.7 and authorized firms.

Exercises

1. Repeat the probability table with a maximum period of 60 days instead of 30. What happens to the trader with no edge who risks 1% a day? And to the one with Sharpe 2?

largo = {**REGLAS, "dias_max": 60}
for s in [0, 2]:
    print(s, round(superar(s, 0.01, rng=rng), 3),
          round(superar(s, 0.01, largo, rng=rng), 3))

0 0.119 0.266
2 0.298 0.613

2. You buy challenges until you pass one, with a maximum of five attempts and a 12% probability of success per attempt (1% a day with no edge). What is the probability that you pass, and how much do you spend on fees on average? Compute it with the formula and check it with a simulation.

p, k = 0.12, 5
exacta = 1 - (1 - p) ** k
intentos = exacta / p
sim = rng.random((100_000, k)) < p
pagados = np.where(sim.any(1), sim.argmax(1) + 1, k)
print(f"Pass: {exacta:.3f} (simulated {sim.any(1).mean():.3f})")
print(f"Mean fees: {CUOTA * intentos:.0f} USD "
      f"(simulated {CUOTA * pagados.mean():.0f} USD)")

Pass: 0.472 (simulated 0.470)
Mean fees: 1968 USD (simulated 1972 USD)

3. A website offers “100,000 USD of capital” after a €2,000 course with a “funding guarantee”. With what you have seen in the lesson, list the questions you would ask and what you would check before paying.