Choose a language · زبان رو انتخاب کن
The picture is kept on this computer and is never sent anywhere, so it does not follow you to another one.
The picture is kept in this browser and is never sent anywhere, so it does not follow you to another one.
Trading it again is practice: a second run does not go on the board.
On a coin market the board charges every trade the fee of an order taken at the market price, and leaves out funding, so its Net can be a little off what your account made.
A challenge is a stretch of chart you hand to somebody else. They trade it blind — they see the instrument but not the dates until it is over — and their result then sits beside yours on one board. It is made out of the stretch you have just been trading.
A run ends the moment a condition is broken, rather than being marked wrong at the end. Leave them all off and it is a plain stretch of chart.
A code is six characters. Anyone with it can join, as many people as you like, and they all end up on one board. A file needs no connection at all.
Somebody has sent you a six-character code, or a file. This is where you take it up: the chart jumps to their stretch with the dates hidden, you trade it, and at the end you see what they did with the same bars.
You will not be told the instrument or the date until it is over. That is the whole point of it.
Every challenge you have made or joined, with its code and its board. Open one to see who has traded it and how they did against you.
Blind takes away everything except the bars. The chart jumps to a date picked at random on this instrument, and the name, the date and the session all go. It is how you find out whether you are reading price or reading the label at the top of the screen.
While it is on, two buttons sit in the corner of the chart: Next moves to another hidden date on the same instrument, and Reveal tells you where you were.
A hollow candle is drawn as its outline alone, which is how MetaTrader draws a rising one. The wick and the outline follow the body's colour until you give them one of their own.
Hours are UTC and never move with daylight saving, so the same time of day looks the same all the way through a replay.
Nothing here touches the account, only what you see.
What goes: the fault itself, and the last few things you pressed before it. What does not: your trading, your journal, your balances, your account and your subscription code. There is no name on it — only a random number this copy made for itself, so that one fault on many machines can be told apart from one machine hitting it often.
The money
The costs
The rules and the replay
A buy and a sell held together on one market each pay their full margin. Some brokers charge less for the hedged part, so a stop-out here can come sooner than there.
Off, the minute is walked as four prices and the pessimistic order is assumed. On, the actual quotes are fetched an hour at a time and kept — which needs a connection the first time, and falls back on its own where there are none.
Keep the replay exactly where it stands, with your trades, your drawings and your indicators. Two things this is for: come back to this same stretch and trade it a different way, and hand it to the Lab to see how a rule would have played the same bars you did.
Open an open-source script on TradingView, copy its code, and paste it here. It is translated on this computer, nothing is sent anywhere. Scripts that draw objects or read something this simulator cannot give them are refused rather than half-translated.
A free account keeps its trades while the program is open, for one day at most. A subscription keeps them all.
A free account keeps its trades for one day at most. A subscription keeps them all.
A free account keeps its trades while the program is open, for one day at most. A subscription keeps them all.
A free account keeps its trades for one day at most. A subscription keeps them all.
Sign in and your subscription installs itself. If you have not bought yet, create an account — the purchase goes there too.Sign in and your subscription is here. If you have not bought yet, create an account — the purchase goes there too.
Finish signing in, in the browser that just opened. This window is waiting.
Money
Account
Ends every session on every machine, including this one. For a shared computer, or one you no longer have. Changing the password does not do this on its own.
The program
Rahyab
Support is on Telegram:
Point the camera at this; the same chat opens.
If Telegram is installed, the button below.
Your history, sessions, drawings and subscription are untouched — they are not kept in the application folder.
What to do with this rule
Say the rule with the boxes. Nothing here has to be learnt, and what comes out is a real Pine script you can read, keep, or paste into TradingView.
Write a strategy in plain JavaScript and run it over the current instrument. It trades through the real account, so every trade lands in History and on the chart — press Replay to watch exactly what the expert does. (A first, simple version to try the idea.)
Paste only code you trust: it runs inside this program and can reach whatever the program can, your account included.
The box above runs a script once, on the account you trade by hand with. To put a strategy into the laboratory instead — so it can be backtested, optimised, walked forward and compared with your own trading — it has to be a module that exports STRATEGY. That is the shape a Pine script is translated into, so this is the same door, one step further in.
Paste only code you trust: it runs inside this program and can reach whatever the program can, your account included.
Test a MetaTrader expert on this data, inside MetaTrader itself. Export the current instrument's history, load it into MT5 as a custom symbol, then run your expert in the Strategy Tester with Visual mode on — the trades appear one by one on MetaTrader's own chart, with MetaTrader's own fills.
Times are UTC; if your expert assumes broker/server time, set InpHoursShift when you run the import script.
A strategy is written in Pine Script, the language TradingView uses. You do not have to know it: the last tab has a prompt that makes ChatGPT or Claude write it for you from a description in your own words.
It is translated on this computer. Nothing is sent anywhere, and anything the translator cannot express exactly is refused by name and line rather than half-translated.
The same rule is run again on history built from these same bars in a shuffled order — same instrument, same volatility, same costs, every trend and level destroyed. A rule that still makes money there was never reading the market.
Tick the settings to sweep and give each a range. Every combination is run. This is the step that manufactures good-looking results, so nothing it produces should be believed until the next tab has been through it.
The history is cut into windows. In each one the settings are chosen on the earlier part and then judged on the later part, which was not looked at while choosing. Only those later parts are reported. It is the one test here whose result means something on its own.
Where the money came from and where it went: the same trades filed by the market they were taken in. A strategy that is flat overall is often two strategies, one that works and one that gives it back, and no total can show that.
If a walk forward has been run, its out-of-sample trades are used instead of the plain backtest — they are the only ones that were not fitted.
You traded a stretch of history by hand. This runs your rule over the same bars, through the same account — the balance, spread and commission the session itself was traded with — and shows where the two of you differed.
The totals are the least interesting part and are shown last. What matters is what the rule took and you did not, what you took and it would not, and who got out better when you both took the same trade. Those are three different problems with three different fixes, and one profit figure cannot tell them apart.
You do not need to learn Pine Script. You need to describe your rule well enough that something else can write it. This page is about the describing.
Before opening anything, write your rule as if explaining it to someone who will trade it for you tomorrow and cannot ask questions. Four things have to be in it: what has to be true before you enter, which direction, where the stop goes, and where you get out. If any of the four is missing you do not yet have a strategy — you have a feeling about the market, and no amount of testing will turn one into the other.
"I buy when it breaks out" is not enough. "I buy when the close is above the highest high of the last 20 bars, with the stop two ATRs below the entry, and I leave at three ATRs or when the close falls back under the 50 EMA" is enough. The difference is that the second one can be wrong.
Type your idea below in plain words. Press Copy and the whole prompt — the translator's exact rules with your idea filled in — goes to the clipboard, ready to paste into ChatGPT, Claude, Gemini or any other. What comes back will load rather than be refused.
The full prompt that gets copied:
If it is refused, the message names the line and what is wrong with it. Paste that message straight back to the AI — "this was refused: …" — and it will fix it. Two rounds is usual.
The refusals are not arbitrary. Each one marks something that cannot be translated without changing what the rule does, so the fix is always to say the rule differently, not to argue:
"it reads its own position size" — say the rule in terms of the market instead. "Only one trade at a time" is a setting here, not something the script has to check.
"reads a different symbol" — run the test on that symbol instead. Only the chart's own symbol is here; a higher timeframe of it can be read with request.security.
"contains a while loop" — ask for a for loop with a fixed count, or for ta.highest, ta.lowest, ta.barssince or ta.valuewhen, which is what most loops in a published script do.
It will write whatever you describe, including rules that cannot work, and it will sound confident about all of them. It has no idea whether your idea is any good — only the four tabs to the left can say, and only in this order: run it, compare it with noise, walk it forward, and read what the DNA says about where it worked.
Expect most ideas to die at the second step. That is not the tool failing. Of 158 variants tested while this application was being built, across 7 instruments and 3 years, none survived out of sample — and the out-of-sample result tracked the spread almost exactly, which is another way of saying the only reliable edge anybody found was the cost of trading. Finding that out in an afternoon here is the whole point.