The rule is about as simple as a rule gets. Enter at the open, place a stop 1.5% below and a target 3% above, exit at whichever is touched first, and if neither is touched, close at the end of the day. Four years of daily data, five hundred trades, and a [[Backtest]] that reports a 61% win rate and a comfortable expectancy. A friend, given the same rule and the same file, reports 43%. Neither spreadsheet has an error in it. The whole difference sits in one line of code that neither of them wrote consciously: what to do on a day when the high reached the target and the low reached the stop.
This module has been about how a bar gets assembled. The last thing to say about it is what the assembly discards. A bar reports four numbers, and the market produced a sequence — hundreds or thousands of trades in an order. The four numbers are a faithful summary of the extremes of that sequence and they contain no information whatever about its order. That missing ordering is the [[Intrabar path]], and every rule with two exit levels depends on it.
The thermometer on the veranda records the day’s highest and lowest temperature. It is accurate and it is useless for one specific question: whether the cold came before the heat or after it. A morning chill followed by a blazing afternoon and a hot morning that broke into an evening storm leave the same two marks on the glass.
A candle is that thermometer. It records how far the day reached in each direction and throws away when. Two completely different sessions — one that ran to your target and then collapsed to your stop, and one that hit your stop and then rallied to your target — leave an identical candle. On the first you made money; on the second you lost it. The candle cannot tell you which day you had.
Where the ambiguity bites
- Any rule with a stop and a target. The commonest structure in retail trading, and the one this lesson is built around.
- An entry and a stop on the same bar. A breakout entry triggered intraday and stopped the same session: did the level break before or after the low was made? On daily bars there is no way to know, and the trade is either never taken or taken and immediately lost.
- A trailing stop measured within the bar. If the trail moves as price advances, then where the advance sat inside the bar determines where the stop was, and the bar does not say.
- A stop and a time-based exit that both fall on the same bar. Less common, same problem.
- Any “first touch” rule at all. The words first touch are a claim about ordering, and ordering is precisely the field the bar does not carry.
Three conventions and one repair, and what each is worth
| Convention | What it assumes | Honest assessment |
|---|---|---|
| Stop first | On any ambiguous bar, the loss is recorded. | Deliberately pessimistic, and the right default. If a system survives this assumption it survives the worst case, and a system that only works under a rosier one has not been shown to work at all. |
| Target first | On any ambiguous bar, the win is recorded. | Produces the best-looking equity curve available from the data and should be treated as an upper bound rather than a result. Useful only as the other end of a range. |
| Use the bar’s direction as a proxy | On an up bar, assume the low came before the high; on a down bar, the reverse. | A heuristic, not a fact. It is right more often than a coin toss and wrong often enough to matter, and it fails hardest on exactly the wide-range reversal bars where the money is. Reasonable as a third estimate; not a substitute for reporting the range. |
| Test on finer bars | Resolve the ordering by looking inside the day. | The only route that genuinely reduces the ambiguity rather than assuming it away — and it reduces rather than removes. A five-minute bar is also a summary; so is a one-minute bar. Only the trade-by-trade record settles it completely, and almost nobody has that for four years of history. |
What to report instead of a single number
- 1Count the ambiguous trades and publish the count
It is one extra line in the results and it is the single most informative number about how trustworthy the headline is. Eighty-eight out of five hundred means the headline win rate carries an eighteen-point band, and that should be stated rather than discovered later.
- 2Report the pessimistic figure as the result and the optimistic one as the ceiling
Not an average of the two, which invents a precision the data does not support. A range with both ends named is a truthful description of what the data can establish.
- 3Judge the system on the pessimistic end
If it is not viable when every ambiguous trade is a loss, then whether it is viable depends on something your data cannot tell you — which is a different situation from having a marginal system, and a worse one.
- 4Widen the levels and watch the ambiguity fall
If moving the target from 3% to 5% drops the ambiguous count from 88 to 12, then most of what you are testing has become resolvable, and the result means far more even if the raw return is lower. A tighter grid is not a more precise test; it is a less determinate one.
- 5Go to finer bars only for the ambiguous trades
You rarely need four years of five-minute data. You need it for eighty-eight specific days. Resolving those by hand or with a short script converts an assumption into a measurement, and it is a genuinely tractable afternoon.
Module checkpoint: how a bar gets made
5 questions. Answers are revealed once you submit all of them.
1.Two hourly charts of the same Indian equity session — one anchored to 9.15, one anchored to the clock — have how many candle boundaries in common?
2.At 9.20 on a five-minute chart, roughly how much of a 20-period exponential moving average has been contributed by today?
3.An intraday scan flags any five-minute bar whose volume exceeds three times the average of the previous twenty bars. Why does it fire in a burst every morning?
4.Your platform builds daily candles by stacking its own intraday bars. Which field disagrees with the exchange’s published daily bar on almost every session?
5.In a 500-trade daily backtest with a 1.5% stop and a 3% target, 88 bars touched both levels. What should be reported?
Ghar aake driver kehta hai: aaj sabse tez 60 chali aur sabse dheere 5. Dono baat sach hai, par isse yeh nahi pata chalta ki jam pehle mila tha ya baad mein — aur asli sawaal wahi tha. Candle bhi yahi register hai: din kitna upar aur kitna neeche gaya woh likh leti hai, kis kram mein gaya woh nahi. Ab ek seedha rule lo — open pe khareedo, 1.5% neeche stop, 3% upar target. Paanch sau trade mein 217 mein sirf target laga, 195 mein sirf stop, aur 88 mein us din ke bar ne dono ko chhua. Un 88 pe agar maan lo target pehle laga: 305 jeet, yaani 61%, expectancy +1.25%. Agar maan lo stop pehle laga: 217 jeet, 43.4%, expectancy +0.45%. Ek hi rule, ek hi data, teen guna ka farq — sirf ek aisi line ki wajah se jo kisi ne soch-samajh ke likhi hi nahi thi. Aur dhyaan do ki dono level jitne paas honge (yahan 1.5 + 3 = 4.5% ka faasla), utne hi zyada din dono chhuenge. Isiliye 88 ki ginti report karo, pessimistic number ko nateeja maano aur optimistic ko chhat — aur zaroorat ho toh un 88 dinon ko chhote bar pe kholo.
- A bar records the extremes of a sequence and discards the order, so it cannot say which of two levels was reached first.
- Ambiguity requires the bar to span both levels, so tight stop-and-target grids produce the most of it — and the most impressive-looking backtests.
- Assuming the target first and assuming the stop first can differ by a factor of nearly three in reported expectancy on the same data.
- Report the ambiguous count and a range, judge the system on the pessimistic end, and resolve the ambiguous days on finer bars if it matters.
- Even a resolved ordering gives a trigger, not a fill — the spread, the depth and the slippage still sit between them.
Mark it done to track your progress through the curriculum.