Five years on real Binance futures data, July 2021 to September 2026, fees out of every
trade. Only position size changes between the rows below.
Simulated: a replay of the rules on real market data, not trades that were placed. The live forward test
is on the Live page, never mixed in.
Personal account
Two settings, both a small fraction of Kelly.
Kelly is the bet size that grows an account fastest — if you knew your edge exactly. Nobody does,
and betting too big can halve an account in one bad run, so professionals bet a fraction of it. Both settings here bet
a small fraction.
Setting
Model A risk
Model B risk
a year
worst drawdown
worst day
Sortino
Calmar
5 years, compounded
Balanced≈ 1/12 Kelly
2.41%
1.47%
+77%
−11.8%
-4.82%
8.0
6.50
+1800%
Growth≈ 1/8 Kelly
3.76%
2.30%
+134%
−17.9%
-7.52%
8.4
7.47
+8021%
The fraction is chosen by drawdown, not return: the largest size at which 19 in 20 simulated five-year
paths stay within a 20% fall (balanced) or 30% (growth), on the edge left after removing the ten best trades
(how). The five-year figure compounds every gain back into the next trade. Drawdowns
are on closed trades, so the intraday figure is never smaller.
The same setting, without its best days
A few days carry the record. Setting them to zero shows what the rest of it is worth.
Balanced setting
a year
worst drawdown
Sortino
Calmar
As published
+76.5%
−11.8%
8.0
6.50
Without its 10 best days
+33.9%
−13.5%
3.5
2.52
Without its 20 best days
+15.8%
−16.2%
1.4
0.98
Over time
By year, by month, by model.
ShowBookSizing
Both models, balanced setting, compounding (gains reinvested), indexed to the start. Hover for any point. Five years, compounded: +1800%, worst drawdown 11.8%.Model A on its own account at the same size. Each model’s line is its own account; compounded, the two do not add up to the combined one. Five years, compounded: +525%, worst drawdown 11.0%.Model B on its own account. Alone its drawdown is deeper than the combined book’s. Each model’s line is its own account; compounded, the two do not add up to the combined one. Five years, compounded: +204%, worst drawdown 13.8%.Calendar years. 2021 and 2026 are part years. Five years, compounded: +1800%, worst drawdown 11.8%.Model A’s calendar years, alone, including a losing 2021. Each model’s line is its own account; compounded, the two do not add up to the combined one. Five years, compounded: +525%, worst drawdown 11.0%.Model B’s calendar years, alone. Each model’s line is its own account; compounded, the two do not add up to the combined one. Five years, compounded: +204%, worst drawdown 13.8%.Every month (63). Months with no trade at all: 0. Five years, compounded: +1800%, worst drawdown 11.8%.Every month, Model A alone. 5 are empty. Each model’s line is its own account; compounded, the two do not add up to the combined one. Five years, compounded: +525%, worst drawdown 11.0%.Every month, Model B alone. 14 are empty — waiting is its normal state. Each model’s line is its own account; compounded, the two do not add up to the combined one. Five years, compounded: +204%, worst drawdown 13.8%.Both models, balanced setting, fixed size, indexed to the start. Hover for any point. Fixed size: every trade risks the same dollar amount, set from the starting balance, so gains are never reinvested. Five years, fixed size: +322%, worst drawdown 12.2%.Model A on its own account at the same size. Fixed size: every trade risks the same dollar amount, set from the starting balance, so gains are never reinvested. At fixed size the two models add up exactly to the combined line. Five years, fixed size: +203%, worst drawdown 11.3%.Model B on its own account. Alone its drawdown is deeper than the combined book’s. Fixed size: every trade risks the same dollar amount, set from the starting balance, so gains are never reinvested. At fixed size the two models add up exactly to the combined line. Five years, fixed size: +119%, worst drawdown 8.5%.Calendar years. 2021 and 2026 are part years. Fixed size: every trade risks the same dollar amount, set from the starting balance, so gains are never reinvested. Five years, fixed size: +322%, worst drawdown 12.2%.Model A’s calendar years, alone, including a losing 2021. Fixed size: every trade risks the same dollar amount, set from the starting balance, so gains are never reinvested. At fixed size the two models add up exactly to the combined line. Five years, fixed size: +203%, worst drawdown 11.3%.Model B’s calendar years, alone. Fixed size: every trade risks the same dollar amount, set from the starting balance, so gains are never reinvested. At fixed size the two models add up exactly to the combined line. Five years, fixed size: +119%, worst drawdown 8.5%.Every month (63). Months with no trade at all: 0. Fixed size: every trade risks the same dollar amount, set from the starting balance, so gains are never reinvested. Five years, fixed size: +322%, worst drawdown 12.2%.Every month, Model A alone. 5 are empty. Fixed size: every trade risks the same dollar amount, set from the starting balance, so gains are never reinvested. At fixed size the two models add up exactly to the combined line. Five years, fixed size: +203%, worst drawdown 11.3%.Every month, Model B alone. 14 are empty — waiting is its normal state. Fixed size: every trade risks the same dollar amount, set from the starting balance, so gains are never reinvested. At fixed size the two models add up exactly to the combined line. Five years, fixed size: +119%, worst drawdown 8.5%.
The two models
One wins often and small, the other rarely and large.
Model
trades
won
per trade
average win
average loss
median trade
best trade
total
Model A
199
67%
+0.42R
+1.11R
-1.00R
+0.58R
+22.0R
+84R
Model B
96
34%
+0.85R
+4.37R
-1.00R
-1.00R
+9.1R
+81R
In R, the multiple of the amount risked, fees included. Model B’s median trade is a full loss:
most of its trades lose, and the winners pay for them several times over.
Prop-firm account
The same trades inside a funded account’s limits.
The firm lends the capital and takes a share of profit. Lose 6% from the start, or 3% in a day, and the
account is closed. A challenge has to make 10% first; a funded account has no target and pays out its profit.
Chance of
challenge
funded
Reaching the 10% target
89.4%
no target
Failing the challenge
3.1%
—
Typical days to pass
96
—
Losing the account in year 1
—
2.4%
Losing it within 3 years
—
8.8%
Funded payout a year
bad case
typical
good case
per $100,0001.00% a Model A trade · 0.35% a Model B trade
$10,482
$15,325
$22,064
Paths are the five-year record resampled in 20-day blocks, so calm and violent stretches stay together. Payout: the trader’s 80% of profit above $100,000, paid monthly; bad and good case are the 25th and 75th percentile.
The challenge runs larger because failing it costs an entry fee; failing a funded account costs the
account — which is why the two sizes differ.