fair-on-paper / working paperView the repo on GitHub
Working paper, October 2026

Fair on PaperWhat a prop-firm challenge is worth, and who pays for it

Retail prop firms sell evaluation contracts: a fee buys a simulated account, and a buyer who reaches a profit target without breaching loss limits is paid a share of later simulated profits. This paper prices one UK firm's published contract on twenty years of intraday index prices at measured broker costs, with every hypothesis fixed and time-stamped before its run.

Not investment advice, and not a recommendation to buy any product. All values assume the firm pays every valid request. Results are simulations on historical prices for one broker's costs, from a draft working paper that has not been peer reviewed.

Saved results, read-only. Tables are the repo's reference results (results/reference/) and the paper's own tables; figures are the paper's. Nothing is simulated in your browser. The firm is not named on this page.

1.03%of the account: expected payouts for a no-skill buyer on the base contract, restricted regime (1.48% in the standard regime)
$0.49 to $1.60expected payout per dollar of list fee across the nine plans, same no-skill buyer, $10k account
$285value at a Sharpe ratio of 0.3, against $103 with no skill, per $10,000 account on the base contract
13 / 5 / 3of 21 pre-registered hypotheses supported, refuted, inconclusive
Read the working paper (PDF)Reproduction package (Zenodo)

The contract map

Expected payouts per purchase within 36 months, in dollars per $10,000 account (V), for a simulated buyer who takes one index trade a day. The buyer chooses a risk per trade in the evaluation (horizontal) and an intended risk once funded. Skill is the buyer's net Sharpe ratio: the coin flip has no skill and loses costs. Choose the skill levels, regime and funded risk. Dashed lines are the $40 promotional and $147 list fee of the base contract.

Skill (net Sharpe ratio)
Regime
Funded risk per trade

Select a point for detail. Rings mark the best fixed sizing the paper selected.

Two things to notice. Value is flat from about 3% to 4.5% risk per trade for the coin flip, then collapses at 5%, where one stop-out breaches the 5% daily limit. And small sizes are costly: conventional 1% risk leaves the no-skill buyer with $33 of value because the evaluation takes so long.

Sizing, skill and how long the buyer stays

An earlier paper argued that risk sizing matters at least as much as skill. The registered hypothesis here was that sizing is worth more than a Sharpe ratio of 0.3. It was refuted, and a second registered run shows why: over a 12-month window the two are level, and from two years on skill leads, with or without the classic 30 and 60-day limits.

Evaluation time limit
Gain from best sizingGain from a Sharpe ratio of 0.3

Dollars per $10,000 purchase, restricted regime. D is the sizing gain minus the skill gain; negative means skill is worth more. 95% intervals.

Real strategies through the contract

The skilled buyer above is idealised. The paper also runs the strategies that survived the author's earlier programme of 24 pre-registered tests (none a confirmed edge; two chosen after their parent tests were seen) through the contract as they actually trade. Every real book is worth less than an idealised buyer with the same Sharpe ratio. Early purchases are 2005 to 2014, late purchases 2017 to 2021.

Purchase period

The 21 pre-registered hypotheses

Each hypothesis was fixed in a dated plan before its run, and is reported as found, including the ones the author expected to be true and were not.

Verdict

Figures from the paper

The paper's own exhibits, unedited. The full text, with every table and the source list for the contract's rules, is in the working paper PDF.

Two panels: probability of passing both steps, and expected payouts on a log scale, against evaluation risk per trade for coin-flip and skilled buyers, with a cliff at 5% risk per trade.
Probability of passing and expected payouts by evaluation risk and skill (paper Figure 1).
Left: expected payouts rise steeply with the buyer's Sharpe ratio, with the restricted regime below the standard regime. Right: gain from sizing versus gain from a Sharpe ratio of 0.3 by value window, with skill gaining more at longer windows.
Value by skill, and sizing against skill by window (paper Figure 2).
The buyer's best policy: risk on the next trade against account balance, for a coin flip and two skill levels.
The best state-dependent sizing policy (paper Figure 3).
Expected payouts per dollar of list fee for nine plans in two regimes, from 0.49 to 1.60 dollars.
Expected payouts per dollar of list fee, by plan (paper Figure 4).
The base contract's list fee as a share of the account falls with account size, against the no-skill value of about 1.03 to 1.48 percent.
List fee against no-skill value by account size (paper Figure 5).
Net Sharpe ratio estimates with 90 percent intervals for 24 pre-registered tests of published trading effects, with reference lines at the break-even of minus 0.05, and at 0.3 and 0.5. None is confirmed.
The 24 pre-registered tests that bear on whether the skill exists (paper Figure 6).

What this does and does not show

  • Assumes the firm pays every valid request. Three published clauses are discretionary (a gambling policy, a requirement to maintain the same level of risk, and a review of each passed evaluation). They are not modelled, and the paper argues they are where much of the seller's margin sits.
  • One firm, one index book, one broker's costs. The base case is a one-trade-a-day US500 book with costs measured on the broker's server. Other instruments give different values: $131 on NAS100, $95 on EUR/USD, $58 on USD/JPY for the no-skill buyer.
  • Stops never slip in the simulation. An exploratory stress, not pre-registered, lowers the no-skill value to $88, $75 and $49 at 5%, 10% and 25% slippage.
  • The skill is idealised, and the evidence that it exists is thin. "No edge found" is supported by the author's tests; "no edge exists" is not. Telling a Sharpe ratio of 0.3 from zero takes about 69 years of daily data.
  • The rules were read on 13 October 2026 and may have changed. A draft working paper, not peer reviewed. Not investment advice.
  • Not on this page: the raw price bars (not redistributed), the firm's web pages, and the private builders of the real strategies. Their daily series are in the reproduction package.