The decisions you make without noticing
Converting a manual strategy to Pine is not a translation problem. Expressing "RSI below 30" in code takes one line, and if that were the difficulty nobody would write about it.
The difficulty is that your manual strategy contains decisions you have never had to state, and Pine will not run until every one of them is explicit.
The one nobody expects: when do you decide?
A script runs once per historical bar [1]. On the realtime bar it runs repeatedly, once for each update [1]. So "I enter when RSI drops below 30" splits into two genuinely different strategies:
| You meant | In Pine | Consequence |
|---|---|---|
| the moment RSI crosses 30, intrabar | act on the developing bar | the condition can reverse before the bar closes — the same value the backtest may never have seen |
| RSI is below 30 at the bar's close | act on the confirmed bar | you enter one bar later than you "saw" it |
Watching a chart, you never make this choice. You see the level break and you act, somewhere in between. Pine has no "somewhere in between", and the two options produce measurably different results.
Choosing the confirmed close is almost always the right answer for a converted strategy, because it is the only one your backtest can honestly reproduce — the reason is in repainting and lookahead.
The adjectives that have to become numbers
Go through your notes and find every word that is not a number. Each one is a parameter you have been setting by eye, differently every time:
| You wrote | Pine needs |
|---|---|
| "strong trend" | a slope, an ADX level, or a distance from a moving average — with a threshold |
| "near support" | within how many ticks, or how many ATRs |
| "good volume" | above what, measured over how many bars |
| "wait for a clean break" | a close beyond the level by a minimum amount, or a number of bars |
| "when it looks extended" | a specific indicator crossing a specific value |
This is the step where most conversions stall, and the stalling is information. If you cannot put a number on "strong trend", you were not applying a rule — you were exercising judgement, and a judgement cannot be backtested. That is worth discovering in an hour rather than after three months of a live strategy that does not behave like you do.
The exit you have never specified
Almost every manual trader has a fully-specified entry and an improvised exit. "I take profit when the move looks done" is not a rule, and it is usually where the actual edge lived.
Pine requires all of it up front: a stop, a target or an exit condition, and
what happens when neither triggers. You also have to decide what happens to a
position that is still open when your condition to enter fires again — Pine
defaults pyramiding to 1 [2], so the second entry is silently ignored unless
you say otherwise.
What conversion actually gives you
Not automation. A specification. At the end you have a written statement of what you do, with numbers, which you can test against history and against your own trade records.
If the specified version disagrees with your actual results, one of two things is true: the specification is incomplete, or your results came partly from judgement you have not captured. Both are worth knowing, and neither is visible while the strategy lives only in your head.
A five-step conversion
Do these in order. The first two need paper, not a builder, and they are where a strategy gets disqualified cheaply.
1. Write the rule as one sentence with no adjectives
Not a paragraph. One sentence, containing only numbers, named indicators, and the words and, or, above, below, crosses.
Enter long when the 14-period RSI closes below 30 and price is above the 200-period simple moving average, exit at 2× ATR profit or 1× ATR loss.
If you cannot write that sentence, stop. The remaining steps will not help, and the problem is not Pine.
2. Count the parameters in it
Every number is a parameter. The sentence above has six: 14, 30, 200, 2, 1, and the choice of RSI. That count matters because it bounds how much you can trust any result you get later — six parameters can fit a lot of history, and the more you have, the more careful you have to be about searching over them.
Fewer is better, and "fewer" is a decision you make here, not after you see the backtest.
3. Decide the bar you act on, and write it down
From the previous section: intrabar or confirmed close. Pick one deliberately. Write it next to the rule, because it is the single assumption most likely to explain a gap between your backtest and your live results later.
4. Build it, then read the declaration
Whether you write the Pine by hand or export it from
the builder, the part to check is the
strategy() line, not the logic:
| Confirm | Why |
|---|---|
commission_type and commission_value are set | otherwise you are testing a market that charges nothing |
slippage is in ticks for this instrument | a tick is a different size on a different symbol |
pyramiding matches what you decided in step 3 | it defaults to 1 [2] |
| every input actually changes the trade list | move a threshold and check |
The full list is in the cost parameters.
5. Test it against your own records, not only against history
This is the step that makes a conversion worth doing, and the one generic tutorials leave out.
Take your last twenty manual trades. Run the specified strategy over the same dates and the same instrument. Then compare entries, not returns:
- did the strategy take trades you took?
- did it take trades you rejected?
- did it skip trades you took?
The third case is the informative one. Every trade you took that the rule does not take is judgement you have not captured — and either it belongs in the specification, or your real results were never reproducible by a rule.
Comparing returns instead tells you almost nothing, because two strategies can reach a similar total from completely different trades.
When the answer is "this cannot be converted"
Some strategies genuinely cannot. Anything depending on reading news, on discretion about market context, or on how a move feels is not a rule set. You can automate a part — a filter, a risk cap, a session restriction — and keep the discretionary entry.
That is a legitimate outcome and better than a specification that quietly differs from what you do. A partially automated strategy you understand beats a fully automated one that is a different strategy wearing your name.
Tactix AI — Studio vs Guide
Tactix AI is AlfaTactix’s product assistant brand (open Tactix AI).
- Tactix Studio turns a one-sentence strategy description into a draft across Timeframe, Signals, Filters, and Risk in the visual Strategy Builder. You review and edit every field before Code Generator writes MQL5 or Pine Script.
- Tactix Guide explains the step you are on — what to fill, what a control means, or how to phrase a rule — without dumping untested source code.
That is form-first automation: the LLM never replaces Code Generator, and you keep plan limits and real-time validation.