rule backtest — trades, equity curve & drawdown, no look-ahead

Rule backtest — every trade, the equity curve and the drawdown

A trading rule is replayed bar by bar over the price path, and the full record is published: every trade, the equity curve, the win rate, average R (the reward-to-risk of a typical trade), profit factor and maximum drawdown — with a note on what each number means. The rule shown is a 20-day moving-average crossover (long when price closes above its 20-day average, exit when it closes below, with an 8% stop). Fills happen at the bar close, so there is no look-ahead, replayed on daily history covering 2021-09-13 → 2026-09-11 (1249 daily bars). Results are gross of costs — no commission, spread or slippage is deducted.

AAPL20-day moving-average crossover, long-only, 8% stopprice history · 2021-09-13 → 2026-09-11 (1249 bars)

trades

+42.3%

win rate

3.11R

avg R (win/loss)

2.36

profit factor

-22.78%

max drawdown

+126.45%

total return

Measured over 2021-09-13 → 2026-09-11 (1249 daily bars) · one symbol, one rule, 52 trades. A dash means this sample could not measure that number — it is not a zero.

Assumptions & limitations

Period: 2021-09-13 – 2026-09-11
Out-of-sample from: none — whole period in sample
Trades: 52
Costs charged: 0 bps per side 0 bps commission + 0 bps slippage, charged on every position change

Hypothetical, backtested results — not the record of a traded account. Past results do not predict future returns; the strategy set was chosen with knowledge of the history it is measured on, and no financing, borrow, tax or capacity constraints are modelled. One name’s result does not generalise. There is no out-of-sample split: every bar behind these figures was already available when the rule was chosen.

Equity curve — the rule's running NAV

strategy NAV (starts 1.00×)x = bar index · 1249 bars · 2021-09-13 → 2026-09-11 · exposure +57%

Sharpe	0.98	Sortino	1.45
Calmar	0.79	Annualized (extrapolated from 1249 bars)	+17.94%
Avg win	+7.34%	Avg loss	-2.36%

Fills execute at the bar CLOSE on the signal bar — no look-ahead. Profit factor = gross gains ÷ gross losses; avg R = average win ÷ average loss; max drawdown = the deepest peak-to-trough fall in this curve. A dash is a number this sample could not measure (no losing trade → no profit factor; no drawdown → no Calmar) — never a zero and never an infinity.

The rule under test

sidelong · size 1 · +8% stop

enter whenclose > sma20

exit whenclose < sma20

promotion verdictOBSERVEcleared 4/6 gates · -69.93% under 2x downside

This rule on AAPL over 2021-09-13 → 2026-09-11 (1249 daily bars) returned +126.45% with a +22.78% worst drawdown and Sharpe 0.98; under a 2x downside shock it returns -69.93%. Verdict: observe.

Below, the agent explains each metric — what it measures, why it matters, and what your number suggests — so you can judge a strategy yourself before risking real money. This is replayed on real daily price history over 2021-09-13 → 2026-09-11 (1249 daily bars), gross of trading costs — your decisions, your risk.

What a backtest actually proves (and does not)

We replayed your rule bar-by-bar over AAPL over 2021-09-13 → 2026-09-11 (1249 daily bars) and it made a total return of +126.45% across 52 trade(s).

A backtest replays your exact entry/exit rules over historical bars to see how they WOULD have done. It is a sanity check, not a promise: the future is not the past, and a rule that fits old data can fail live. That is why we also stress it and slice it by market regime before trusting it. It is also gross of costs — no commission, spread or slippage is deducted.

→ It was profitable in this sample — keep validating.

Risk-adjusted return — Sharpe, Sortino, Calmar

Sharpe 0.98, Sortino 1.45, Calmar 0.79.

Return alone is misleading — a strategy that doubles then halves is not "good." Sharpe divides return by total volatility (reward per unit of wiggle). Sortino is the same but only penalises DOWNSIDE wiggle (you do not mind upside surprises). Calmar divides annual return by the worst peak-to-trough drop. Higher is better; Sharpe ≥ 1 is the usual "worth a look" line.

→ Risk-adjusted return is below the usual bar — returns came with a lot of volatility.

The worst it got — max drawdown

The deepest peak-to-trough fall was +22.78%.

Max drawdown is the largest drop from a high-water mark to a later low. It is the question "how much pain would I have had to sit through?" — the number that actually makes people abandon a strategy at the worst moment. A high Sharpe with a 60% drawdown is usually un-tradeable in real life.

→ Drawdown was tolerable (≤ 25%).

Edge quality — hit-rate & profit factor

Hit-rate +42.31%, profit factor 2.36 (avg win +7.34% vs avg loss -2.36%).

Hit-rate is the share of trades that won. But winning often is not the same as making money — profit factor (gross profit ÷ gross loss) tells you whether your winners outweigh your losers. A 40% hit-rate with big wins and small losses beats a 70% hit-rate that gives it all back. Profit factor > 1 means the edge is net positive.

→ Winners comfortably outweigh losers.

Stress test — what 2x downside does

Under a 2x downside shock the same rule returns -69.93% (vs +126.45% unshocked).

We amplify every down-bar move and re-run the rule. It answers "if the bad days were twice as bad, does this strategy still survive?" A rule that only works in calm markets is a trap. Promotion sim→live requires surviving the shock, not just the base case.

→ Falls apart under stress — do not promote.

Promotion verdict: OBSERVE

Cleared 4/6 gates (score 0.67).

Sim-to-live promotion is the whole point: only rules that pass return, risk, drawdown, profit-factor, stress and across-regime checks get handed to the model foundry as a candidate. "Promote" = ship to live paper/real; "observe" = keep collecting data; "reject" = back to the drawing board.

→ Promising but not proven — keep it in sim.

Backtest glossary

Backtest: Replaying a fixed set of trading rules over historical prices to estimate how they would have performed.
Sharpe ratio: Annualised return divided by the volatility of returns — reward per unit of total risk.
Sortino ratio: Like Sharpe but only counts downside volatility in the denominator.
Calmar ratio: Annualised return divided by the maximum drawdown — reward per unit of worst-case pain.
Max drawdown: The largest drop from a peak in the equity curve to a subsequent trough.
Profit factor: Gross profit divided by gross loss; above 1 means the strategy is net profitable.
Exposure: The fraction of bars the strategy was actually in a position (capital at work).
Stress test: Re-running the strategy on a deliberately worse price path to see if it survives bad conditions.
Sim-to-live promotion: The gate that decides whether a paper-traded rule-set is trustworthy enough to run with real consequences.

Trade ledger — all 52 fills


1	23 @ 143.76	42 @ 147.92	19 bars	+2.89%	rule
2	44 @ 149.99	80 @ 174.92	36 bars	+16.62%	rule
3	96 @ 170.33	106 @ 168.64	10 bars	-0.99%	rule
4	107 @ 168.88	110 @ 168.88	3 bars	+0.00%	rule
5	130 @ 163.98	146 @ 165.75	16 bars	+1.08%	rule
6	162 @ 166.02	163 @ 156.77	1 bars	-5.57%	rule
7	179 @ 149.64	183 @ 145.38	4 bars	-2.85%	rule
8	184 @ 146.14	187 @ 142.64	3 bars	-2.39%	rule
9	197 @ 141.66	199 @ 137.44	2 bars	-2.98%	rule
10	202 @ 138.93	241 @ 163.62	39 bars	+17.77%	rule
11	280 @ 147.27	288 @ 145.03	8 bars	-1.52%	rule
12	294 @ 146.87	305 @ 144.22	11 bars	-1.80%	rule
13	307 @ 148.03	311 @ 142.91	4 bars	-3.46%	rule
14	335 @ 133.49	362 @ 148.48	27 bars	+11.23%	rule
15	370 @ 151.03	375 @ 148.50	5 bars	-1.68%	rule
16	376 @ 150.47	397 @ 160.10	21 bars	+6.40%	rule
17	398 @ 165.56	406 @ 163.77	8 bars	-1.08%	rule
18	408 @ 168.41	474 @ 192.58	66 bars	+14.35%	rule
19	493 @ 184.12	499 @ 177.56	6 bars	-3.56%	rule
20	519 @ 174.91	529 @ 175.46	10 bars	+0.31%	rule
21	539 @ 177.57	574 @ 193.60	35 bars	+9.03%	rule
22	590 @ 188.63	599 @ 184.40	9 bars	-2.24%	rule
23	603 @ 189.30	605 @ 188.32	2 bars	-0.52%	rule
24	632 @ 176.08	634 @ 171.37	2 bars	-2.67%	rule
25	638 @ 173.31	639 @ 171.48	1 bars	-1.06%	rule
26	648 @ 175.04	651 @ 169.38	3 bars	-3.23%	rule
27	658 @ 169.89	659 @ 169.30	1 bars	-0.35%	rule
28	660 @ 173.50	662 @ 169.30	2 bars	-2.42%	rule
29	663 @ 173.03	719 @ 218.54	56 bars	+26.30%	rule
30	733 @ 221.27	748 @ 220.85	15 bars	-0.19%	rule
31	759 @ 228.87	771 @ 221.69	12 bars	-3.14%	rule
32	772 @ 225.77	788 @ 230.10	16 bars	+1.92%	rule
33	801 @ 228.02	831 @ 243.85	30 bars	+6.94%	rule
34	847 @ 238.26	851 @ 228.01	4 bars	-4.30%	rule
35	854 @ 233.22	855 @ 227.63	1 bars	-2.40%	rule
36	857 @ 232.62	868 @ 237.30	11 bars	+2.01%	rule
37	869 @ 241.84	870 @ 238.03	1 bars	-1.58%	rule
38	890 @ 222.13	893 @ 204.36	3 bars	-8.00%	stop
39	906 @ 204.60	914 @ 198.89	8 bars	-2.79%	rule
40	919 @ 210.79	925 @ 206.86	6 bars	-1.86%	rule
41	946 @ 201.00	972 @ 211.27	26 bars	+5.11%	rule
42	978 @ 213.25	1002 @ 226.79	24 bars	+6.35%	rule
43	1004 @ 234.07	1024 @ 245.27	20 bars	+4.78%	rule
44	1030 @ 262.24	1050 @ 267.46	20 bars	+1.99%	rule
45	1054 @ 271.49	1069 @ 274.11	15 bars	+0.97%	rule
46	1099 @ 258.28	1109 @ 261.73	10 bars	+1.34%	rule
47	1111 @ 263.88	1113 @ 260.58	2 bars	-1.25%	rule
48	1115 @ 266.18	1119 @ 264.18	4 bars	-0.75%	rule
49	1142 @ 255.63	1188 @ 301.54	46 bars	+17.96%	rule
50	1205 @ 308.63	1219 @ 308.91	14 bars	+0.09%	rule
51	1232 @ 316.83	1233 @ 311.30	1 bars	-1.75%	rule
52	1237 @ 313.45	1248 @ 332.27	11 bars	+6.00%	eod

Exit reasons: rule = the exit predicate fired; stop = the protective stop was hit; target = the take-profit was hit; eod = closed at the last bar. Bar index = position in the price path, not a calendar date.