Nine numbers sit between a broker statement and a computed sentence.
Numbers first: the engine cannot read your recap, your Discord, or your memory of a good month — it can read nine fields (widget also fixes trades/day = 3 and cost = 0.05R). Answer first: fill them from a brokerage statement without inventing a trader you are not. No API. No crypto pivot. No new price.
If a number did not come from closed trades or from a rule you will actually obey, it is not an input — it is a wish. The Monte Carlo will still run. The sentence it returns will be about the wish. Type the nine numbers free at app?rules=personal&src=b2-nine-dials.
What a statement contains versus what the engine wants
Your last 100 trades already says the hard thing: a broker export is not a list of engine trades. CSV import extracts a parametric summary (win rate, average win, average loss), not a tick-for-tick replay. This page is the manual twin of that: you do the extraction with a pencil if you have to.
| # | Dial | Where it lives on a statement | Honest fallback |
|---|---|---|---|
| 1 | Account size | Ending equity you will actually risk — not peak, not margin. | Today’s cash at risk. Round down, not up. |
| 2 | Max drawdown % | Not on the statement. A rule you write: (peak − floor) ÷ the account size in dial 1. | If you have no rule, you do not have a simulation. Write one number. |
| 3 | Horizon (days) | Not on the statement. The window you will review against. | The number of sessions in the sample you used for 7–9, if that is the question you mean. |
| 4 | Daily loss stop % | Your day-stop, if you have one. Consequence = lockout unless you are in a firm. See personal daily loss. | Tick “No daily stop.” Do not type 0 and hope. |
| 5 | Profit goal % | Optional. Self-funded books often have none. | Tick “No goal — survival only.” |
| 6 | Risk per trade % | Median intended stop as % of dial 1, not the loss you happened to take. | If size varies wildly, the model’s i.i.d. assumption is already strained — say so, then pick the size you will use next. |
| 7 | Win rate % | Closed winners ÷ closed trades. Same session filter as 8 and 9. | A short sample. Never a remembered “about 60.” |
| 8 | Avg win (R) | Mean winning P&L ÷ the risk you assigned that trade (dial 6’s dollars). | If you do not know R, compute dollars first, then divide by intended risk. Do not use the best win. |
| 9 | Avg loss (R) | Mean losing P&L ÷ the same intended risk. | Include the ugly ones. Exclude them and the floor is fiction. |
The personal widget also fixes trades/day = 3 and cost = 0.05R. Those are modeling assumptions, not measurements of your book. The product desk at /app?rules=personal is the same engine; if you need a different trade-rate or cost, that is the desk, not the widget, and you still must not invent them.
A 20-line procedure
- Export closed trades for one contiguous window (20 sessions is better than a cherry-picked week).
- Drop open positions. The personal model, like the prop trail-intraday model, is a closed-trade object unless you have written otherwise.
- Count N, winners, losers. Win rate = winners / N.
- For each trade, write intended risk in dollars. If you cannot, you cannot speak R yet — write the risk you will actually take before you run the engine.
- Avg win R = mean(winning P&L / intended risk). Avg loss R = mean(|losing P&L| / intended risk).
- Account size = capital you will not replenish mid-horizon.
- Max drawdown = the floor you will obey. Write it in dollars, then divide.
- Daily stop = lockout or off. Do not paste a firm FAQ.
- Goal = off unless you have a real one.
- Run. Read which clause binds. That sentence is the product. Share the proof URL if you want a bit-identical re-run.
What the CSV path currently keeps and discards is documented in Your last 100 trades. The argument that a personal rule is the same object as a firm evaluation is Write your own rulebook. The self-funded reading is Your own futures account has a floor.
What this will not do
- It will not connect to a broker. That is constraint-blocked (privacy).
- It will not replay your exact sequence as the default engine. Parametric i.i.d. is the published model; block-bootstrap comparison lives on the 100-trades page.
- It will not name a firm you should buy, and it will not invent a pass rate.
- It will not turn a spot-crypto recap into a new category. Independent is a persona on the same engine.
Working backwards: free path before you type a remembered “about 60.”
Numbers first: win rate = closed winners ÷ closed trades; avg win/loss in R use the same intended risk; widget-fixed trades/day = 3 and cost = 0.05R. Answer first: type the nine numbers on the free personal path — a short honest sample beats a round wish.
Simulated result from user-supplied inputs · not investment advice. PropSurvival does not import broker credentials, does not rank firms, and does not promise that a typed win rate will occur next month. Trades never leave the device.