Prompts you can copy into ChatGPT, Claude or any assistant, grouped by the job: reading a chart, a top-down read across four timeframes, why a market moved, exact rules, backtest specs, position sizing, your plan, a pre-trade check, journal reviews and post-mortems. None of them asks what to buy or where the price is going.
November 15, 2026
28
min to read
The short answer
A trading prompt works when it gives the model one job, the data to use, your limits and the output you want, and asks how sure it is and what would prove the idea wrong. Below are 30 prompts by job, from chart reads to journal reviews, each with what to paste and what a good answer contains.
A trading prompt has one problem other prompts don’t: a wrong answer can cost money, and a model that doesn’t know will sometimes guess instead of saying so. It can’t see your screen, your broker or the live market, so whatever you leave out, it fills in. The fix is to hand it everything it needs and to ask for the parts a trader cares about most.
OpenAI’s own prompting guide for ChatGPT, as of November 2026, names four parts for an important task: the goal, the context, the output and the boundaries. One of its sample boundaries fits trading almost word for word: “Use only the supplied sources. Flag missing information instead of guessing.” Anthropic’s guide for Claude makes the same point with a picture: treat the model as “a brilliant but new employee who lacks context on your norms and workflows.” Its test is a good one. Show your prompt to a colleague who knows nothing about the task; if they would be confused, the model will be too.
For trading, add two parts to those four. Ask how sure the model is, because OpenAI’s Model Spec tells its models not to put a number on their uncertainty unless you request it. And ask what would prove the idea wrong, which turns a story into something you can check.
| Part | What to write | Example line |
|---|---|---|
| Goal | The one job you want done | “Read the trend and the nearest levels on this chart.” |
| Context | Your market, timeframe, holding period and style | “I swing trade large US stocks on the daily and hold 3 to 10 days.” |
| Data | What you paste, labeled | “Chart: NVDA, 4-hour, last three months. File: daily prices since January.” |
| Limits | What it must not do | “Use only this data. No buy or sell call. Write ‘not visible’ instead of guessing.” |
| Output | The shape of the answer | “A table: level, price, timeframe, times tested.” |
| Confidence | How sure it is, and why | “Rate each point high, medium or low and give the reason.” |
| Wrong if | The test that would kill the idea | “Name the price or the event that would prove this read wrong.” |
Goal: [the one job, e.g. read the trend and the key levels]. Context: I trade [market] on the [timeframe] and usually hold for [holding period]. Data: [paste the chart, the numbers or the file, and say what each one is]. Limits: use only the data I gave you; if something is missing, write “not in the data” instead of guessing; do not tell me whether to buy or sell. Output: [a table, five bullets or a checklist], short enough to read in a minute. Confidence: rate each point high, medium or low and say why. Wrong if: tell me what would prove this read wrong.
Three habits make any of the prompts below work better. Give each prompt one job, not five. Label what you paste, with a line like “DATA:” above the numbers; Anthropic suggests wrapping each kind of content in its own tag so instructions and data don’t blur. And when the answer matters, run the same prompt twice: if the two answers disagree on a number, at least one of them is wrong.
You won’t find a “what should I buy” or “where will the price be next week” prompt in this guide. The Model Spec lists “predictions of future states” among the cases where “the outcome is inherently uncertain,” so a prompt that asks for a price target mostly gets a confident guess. Every prompt below asks for evidence, levels, rules or checks instead.
A screenshot is pixels, not data. The model estimates each price from where it sits against the axis, and OpenAI’s vision guide lists the traps: it may struggle with graphs where colors or line styles vary, it “struggles with tasks requiring precise spatial localization,” small text should be enlarged, and images may be resized before the model sees them. So crop to one chart, keep the price axis, the ticker and the timeframe in view, and enlarge the labels. OpenAI’s image input doc adds the other half: say what to inspect instead of relying on the image alone. Our guide to AI chart analysis tests this directly, with a real candlestick chart and the data behind it.
This is a screenshot of [TICKER] on the [TIMEFRAME] chart, covering about [PERIOD]. Before you analyze it, tell me what you can see: the ticker, the timeframe, each indicator and its settings, and the price range on the axis. If any of these is unreadable, stop and ask me. Then describe the trend in two sentences, give the last swing high and swing low with the prices you read off the axis, and say what each indicator shows. Rate your confidence in every number, and tell me what on this chart would prove your read wrong. Do not tell me whether to buy or sell.
Paste: one cropped screenshot with the axis and the indicator labels visible. A good answer: it confirms the ticker, timeframe and indicators before it reads anything, gives prices as estimates with a confidence for each, and names a level that would undo its read.
Here is a chart of [TICKER] on the [TIMEFRAME] and my read of it: [YOUR READ]. Go through my read point by point and mark each point as supported by the chart, contradicted by it, or not visible in it. For each one, say which part of the chart you are looking at: the date, the candle or the price area. Then tell me the most important thing I missed.
Paste: the screenshot and your own notes, written before you ask. A good answer: a verdict on every point, at least one disagreement or a plain statement that it found none, and a location on the chart for each.
Look at this [TIMEFRAME] chart of [TICKER]. If you see a chart pattern, such as a range, a flag, a double top or a head and shoulders, name it, give the prices where it starts and ends, and give the level where it would be confirmed and the level where it would fail. Then argue that it is not that pattern at all. If you see no clear pattern, say so and stop.
Paste: a screenshot wide enough to show the whole pattern. A good answer: both the pattern and the case against it, with a failure level. “No clear pattern” is a good answer when it is true.
I am attaching four screenshots of [TICKER], labeled and in this order: daily, 4-hour, 1-hour and 15-minute. For each timeframe, give the trend (up, down or sideways) and the one level that matters most, with its price. Then tell me where the timeframes agree, where they conflict, and which one should carry the most weight for a trade I would hold for [HOLDING PERIOD]. Answer in a table with short cells, and flag any screenshot you could not read well.
Paste: four screenshots in a fixed order, each with its timeframe visible on the image. A good answer: a four-row table, one sentence on agreement and conflict, and a clear link between the timeframe it weights most and your holding period.
On [TICKER], the daily chart looks [YOUR VIEW] but the [LOWER TIMEFRAME] chart looks [YOUR VIEW]. Using the screenshots I attached, explain what each timeframe is showing, what usually causes this kind of conflict, and what would have to happen on the lower timeframe for the two to line up again. Give the price levels involved. Do not predict which timeframe wins.
Paste: the daily chart and the lower timeframe you trade. A good answer: an explanation of the conflict with levels on both charts, and a condition you can watch for, not a forecast.
From this [TIMEFRAME] chart of [TICKER], list up to five support and resistance levels. For each one, give the price you read, why it counts as a level (a swing high or low, a gap, a round number, a price tested several times), how many times price touched it, and whether it last acted as support or resistance. Sort them by distance from the current price, and mark which prices you estimated from the pixels.
Paste: a screenshot with the price axis readable. A good answer: a sorted table with a reason and a touch count for every level, and the estimated prices clearly marked.
Below is daily price data for [TICKER] for [PERIOD], one row per day: date, open, high, low, close, volume. Using only this data, find the prior day’s high and low, the prior week’s high and low, the highest and lowest close of the period, and the three most recent swing highs and swing lows, where a swing high is a high above the two candles on each side. Show the dates and prices in a table and state your method in one line so I can check it.
Paste: the rows from your platform’s export, or the file itself if the assistant can run code. A good answer: exact prices with their dates, the method in one line, and nothing taken from outside the data.
These need web search switched on. Without it, the model answers from training data that stops at a cutoff date, and it may describe last year’s story as today’s. With it, keep your guard up: ChatGPT’s search docs say to “Treat all web results as untrusted input.” Ask for the date and the source of every claim, and check the timing against the chart.
Search the web. [TICKER] moved about [X]% [up or down] on [DATE]. Find out why. Give up to three reasons, each with the source link, the publisher and the time it was published, and say whether each one came out before or after the move began. Check how [TWO OR THREE PEERS] and the broader market did that day. If you find nothing specific to [TICKER], say that plainly instead of filling the gap.
Paste: the ticker, the size of the move, the date and the peers you compare it with. A good answer: dated sources, the timing of each one against the move, how the peers did, and a plain “nothing specific” when that is the truth.
Search the web and build a timeline of news about [TICKER] from [START DATE] to [END DATE]: earnings, guidance, analyst changes, product news, regulation and big market events. One row per item with the date, a one-line summary, the source, and whether it was company, sector or macro news. Leave out anything you cannot date. Then tell me which two items look most connected to the price change over the period, and why that link could be a coincidence.
Paste: the ticker and the dates of the move you are studying. A good answer: dated rows with sources, no undated filler, and an honest note on why a link between news and price can be chance.
I saw this claim: “[PASTE THE HEADLINE OR POST]”. Search for the original source. Tell me who reported it first and when, whether the company or a regulator has confirmed it, and whether other outlets report it independently or only repeat one source. Quote the exact sentence from the primary source if you find it. If you cannot find a primary source, say so.
Paste: the claim exactly as you saw it. A good answer: a trail back to the first report, a quote from the company or the regulator, or a clear “no primary source found.”
I attached [COMPANY]’s latest quarterly report. Summarize revenue, margins, guidance and cash flow, each with the number and the page or section it comes from, and compare each with the prior quarter where the report gives it. List what management says changed, the three biggest risks the report names, and anything that looks inconsistent. Use only the document, and mark anything you had to infer.
Paste: the report or the earnings release as a file. A good answer: numbers with page references, quarter-on-quarter changes, the risks in the company’s own words and any inconsistency flagged.
Search the web and explain [TOKEN] to me: what it is used for, how its supply works and whether large unlocks or emissions are scheduled (with dates), where most of its trading happens, and, if I trade it as a perpetual future, how that market’s funding rate works. Give a source and a date for each fact, and flag anything that comes only from the project’s own marketing.
Paste: the token and the market you trade it on. A good answer: sourced and dated facts, a clear split between independent sources and the project’s own claims, and the supply events with dates.
Most trading ideas start vague: “buy when it’s oversold and turning up.” A model is good at the slow part of making that exact, because it doesn’t get bored listing edge cases. Get the rule exact first, then write down how you would test it, and only then ask for code. The method itself, from costs to look-ahead bias, is in how to backtest a trading strategy, and the path from an idea to a live rule is in algorithmic trading for beginners. When you get to the code, our guide to Pine Script with ChatGPT or Claude has a prompt that asks for version 6. Our guide to AI backtesting covers the traps specific to letting an AI run the numbers itself.
Here is my trading idea in my own words: “[YOUR IDEA]”. Rewrite it as an exact rule a computer could follow with no judgment calls: the market, the timeframe, the entry condition, the exit condition, the stop, the position size, and what happens if two signals fire at once. Then list every vague word in my idea, such as “strong”, “oversold” or “breaks out”, with the exact definition you chose for it, so I can change it.
Paste: your idea as you would say it to a friend. A good answer: a rule with no room for judgment, and a list of the vague words it had to pin down, each with its definition.
Here is my trading rule: [RULE]. Act as the programmer who has to code it. List every situation it doesn’t cover: a gap at the open, a signal on the last bar of the week, an exit and an entry on the same candle, a stop and a target both inside one candle, missing data, partial fills. For each one, propose the simplest way to handle it and ask me to choose.
Paste: the rule from prompt 13, or your own. A good answer: a list of real edge cases with a proposed choice for each, and no code yet.
Turn this rule into a checklist I can run by hand on a chart, one yes-or-no question per line, in the order I should check them: [RULE]. Put the exact indicator settings and the timeframe in each line. End with the one condition that cancels the trade.
Paste: the finished rule. A good answer: short yes-or-no lines with the settings written in, and a single cancel condition at the end.
Before writing any code, write a one-page specification for backtesting this rule: [RULE]. Include the market and the data source, the timeframe and the test period (and why that period includes rises, falls and flat stretches), the costs and slippage per trade, when an order fills (at the next bar’s open, not on the bar that made the signal), the position size, what happens to a trade still open at the end, the split between the data used to build the rule and the data kept back to test it, the benchmark (buying and holding over the same dates) and the numbers to report. Flag every assumption I should confirm.
Paste: the rule and the market you plan to trade. A good answer: a complete spec on one page, fills at the next bar, a held-back test period, a buy-and-hold benchmark and a list of assumptions for you to approve.
Here are the results of a backtest: [PASTE THE SUMMARY AND, IF YOU CAN, THE LIST OF TRADES]. Look for red flags: too few trades, a result driven by one or two trades, a test period that mostly went one way, no costs, fills at the close of the signal bar, settings tuned on the same data they were judged on, and a drawdown I could not sit through. Rank the flags by how much each could be inflating the result, and tell me the one extra test that would change my confidence most.
Paste: the summary statistics and the trade list. A good answer: a ranked list of concrete problems, each tied to a number in your results, and one next test.
This is where models make the quietest mistakes: a slipped decimal in a position size looks as tidy as a correct one. Ask for every step of the math, or for the numbers to be run in code, and check one line yourself. Anthropic’s prompting guide suggests ending with a self-check such as “Before you finish, verify your answer against [test criteria],” which it says catches errors reliably in coding and math.
My account is [ACCOUNT SIZE]. I risk at most [X]% of it on one trade. I plan to enter [TICKER] at [ENTRY] with a stop at [STOP]. Work out the position size in [shares, coins or contracts], the dollar amount at risk, and the position’s value as a share of the account. Show each step of the math. Then redo it with the stop 20% farther from the entry and tell me what changed. Do not comment on whether the trade is a good idea.
Paste: your account size, your risk limit, the entry and the stop. A good answer: the math line by line, where the distance from entry to stop times the size equals the risk you set, and a second version with the wider stop.
I am looking at a [LONG OR SHORT] in [MARKET] at [PRICE] with [X]x leverage and [MARGIN] of margin. Explain in numbers the position’s total size, what a 1% move against me costs, roughly where liquidation would be (and tell me to confirm it with the exchange’s own figure), and where my stop at [STOP] sits compared with it. If my stop is beyond the liquidation price, say that first.
Paste: the market, the leverage, the margin and the stop. A good answer: the dollar cost of a 1% move, a rough liquidation price (where the exchange would close the position for you) marked as an estimate, and an early warning if the stop is beyond it.
I risk [X]% of my account per trade and win about [Y]% of my trades. Over 200 trades, how likely is a streak of 5, 8 or 10 losses in a row, and how far would my account fall during each streak? Run a simulation in code, explain the method in two lines, and give the answer as a small table. Then tell me what risk per trade would keep a 10-loss streak under a 20% fall.
Paste: your real risk per trade and your win rate from your journal. A good answer: a simulation rather than a hand-waved guess, the method stated, a small table, and the risk per trade that fits your limit.
Help me write my trading plan. First ask me up to ten questions, one at a time, about the markets and timeframes I trade, how much I risk per trade and per day, when I stop for the day, how I enter and exit, and what I want from trading. Then write the plan in four parts: Risk, Markets, Process and Goals, with concrete numbers wherever I gave them. Keep it to one page and mark anything I left undecided.
Paste: nothing at first; answer its questions honestly. A good answer: questions before any writing, then a one-page plan in four parts with your numbers, and the open decisions marked. In ChatGPT, save the finished plan in a project’s instructions, which apply to every chat in that project.
Here is my trading plan: [PASTE IT]. Find rules that contradict each other, rules too vague to follow in the moment (“be patient”), situations it doesn’t cover (a gap through my stop, a surprise headline, three losses in a row), and any rule that depends on willpower instead of a number. Suggest one rewrite for each problem without making the plan longer than a page.
Paste: the plan as you actually use it. A good answer: specific problems quoted from your own plan, each with a rewrite, and a plan that stays short.
Here is my trading plan: [PASTE THE PLAN]. Here is the trade I am considering: [MARKET, DIRECTION, ENTRY, STOP, TARGET, SIZE, REASON]. Check the trade against every rule in my plan and answer in a table: the rule, pass or fail, and why. Then list anything this trade depends on that my plan doesn’t cover. Do not tell me whether to take the trade; tell me which rules it breaks.
Paste: your plan and the full trade, including the size. A good answer: a row for every rule, a pass or fail with a reason, and the gaps in your plan that this trade exposes.
I am considering [THE TRADE] because [MY REASONS]. Argue against it as a skeptical risk manager would: the strongest reasons it could fail, which of my reasons are assumptions rather than facts, the event in the next [PERIOD] that could hurt it most, and the price at which my reason for the trade would no longer be true. Keep it to ten lines.
Paste: the trade and your honest reasons for it. A good answer: arguments you had not considered, your assumptions named as assumptions, and a price where the idea stops being true.
Reviewing is where an assistant can be the most useful and the least risky: the data is yours, nothing is live, and the job is pattern-finding. Two cautions. Small samples lie, so ask how many trades each pattern rests on. And the model may lean toward what you want to hear: OpenAI’s Model Spec says a model “exists to help the user, not flatter them or agree with them all the time,” but your prompt should still ask for the uncomfortable parts.
I attached my trade log as a file with these columns: [LIST THE COLUMNS]. Using code, calculate the number of trades, the win rate, the average win and loss, the expectancy per trade and the longest losing streak, overall and broken down by setup, market, day of the week and time of day. Show the tables, then tell me the three patterns that stand out most, how many trades each is based on, and which could be noise because the sample is small.
Paste: your journal exported from your broker or your spreadsheet, with a column for the setup. A good answer: numbers computed in code, the sample size next to every pattern, and an honest flag on the ones too small to trust.
Here are my losing trades from [PERIOD], one per line with the entry, the exit, the size and a note: [PASTE THEM]. Tag each loss with one main cause from this list: followed the plan and lost (a normal loss), entered early, entered late, stop too tight, stop moved, size too big, traded against the higher timeframe, traded the news, revenge trade, other. Count the tags and tell me which cause cost the most money.
Paste: the losing trades with your notes from the time. A good answer: one tag per trade, a count and a dollar total per cause, and “normal losses” kept apart from mistakes.
Help me do a post-mortem on this losing trade: [MARKET, ENTRY, EXIT, SIZE, STOP, WHAT I SAW, WHAT HAPPENED]. Ask me questions until you can answer these four: Was the setup valid by my own rules? Was the size right? Did I follow my exit plan? Was this a sound trade that lost, or a mistake in the process? End with one specific change to a rule or a habit, or tell me that no change is needed.
Paste: the trade, a chart if you have one, and what you were thinking at the time. A good answer: questions before verdicts, a clear split between a bad outcome and a bad decision, and one change at most.
Here is my trading plan and a trade that went wrong: [PASTE BOTH]. Write out, step by step, what my plan said to do at each point of this trade (entry, size, stop, adds, exit), and next to it what I actually did. Highlight the first point where the two split, and work out what following the plan from there would have cost or saved me.
Paste: the plan and the trade with times and prices. A good answer: a side-by-side table, the exact moment you left the plan, and the difference in dollars.
Explaining is one of the jobs these assistants do best, and the stakes are low as long as you check anything you will act on. Ask for numbers, not definitions, and make it test you.
Explain [CONCEPT, for example the funding rate on a perpetual future, the RSI, or a stop-limit order] to someone who trades but has never used it. Use one worked example with simple numbers, then the two most common mistakes people make with it, and how it can lose money. Define every term the first time you use it. Finish with three questions that check whether I understood.
Paste: the concept, and your level if you want it pitched higher or lower. A good answer: one worked example with numbers you can follow, the mistakes and the way it loses money, and questions that test the idea rather than the vocabulary.
Quiz me on [TOPIC], one question at a time, starting easy. After each answer, tell me whether I am right and explain why in two sentences. Make the next question harder if I got it right and simpler if I did not. Stop after ten questions and tell me which ideas I should review.
Paste: the topic, and nothing else. A good answer: one question at a time, short explanations, difficulty that follows your answers, and a list of weak spots at the end.
Every assistant can make these mistakes, whichever model sits behind it. A September 2025 paper by researchers at OpenAI and Georgia Tech gives the reason: models “sometimes guess when uncertain, producing plausible yet incorrect statements instead of admitting uncertainty,” because training and scoring reward a lucky guess over “I don’t know.” Anthropic’s guide to reducing hallucinations is just as direct: its techniques “significantly reduce hallucinations” but “don’t eliminate them entirely.”
| Mistake | What it looks like | How to catch it |
|---|---|---|
| Made-up numbers | An RSI (relative strength index) value for a chart with no RSI on it, a “52-week high” that is off by a few dollars | Ask where every number came from; compare the ones you act on with your platform |
| Stale prices | Last month’s price quoted as today’s, or a search snippet from an old story | Paste the current price yourself, or ask for the time of every quote |
| Misread charts | A level read from the wrong line, a candle count that is roughly right | Crop and enlarge, ask what it can see first, or paste the numbers instead of the picture |
| A confident tone | Firm sentences with no doubt in them | Ask for a confidence rating and the reason behind it |
| Arithmetic slips | A position size off by ten, a percentage of the wrong number | Ask for every step, or for the math in code; check one line by hand |
| Agreeing with you | Your idea, played back to you with more adjectives | Ask for the case against, or give your view last |
| Answers that change | A different level each time you ask | Run the prompt twice; a number that moves is a number to check |
Several of these fixes come straight from the providers. OpenAI’s guide to using ChatGPT tells you to “Verify important numbers, names, dates, quotes, and claims,” and to ask questions like “What assumptions did you make?” and “What information were you unable to access?” It adds a line worth keeping in mind: “An explicit gap is easier to address than a confident guess.” Anthropic recommends giving the model explicit permission to say “I don’t know,” which it says “can drastically reduce false information,” and running the same prompt several times to compare: inconsistent answers “could indicate hallucinations.”
The Model Spec explains why confidence needs a direct request. It ranks a confident wrong answer as the worst outcome, below no answer at all, and asks the models to signal doubt in words such as “I think” or “If my information is up to date.” A number on that doubt, such as a percentage, comes only when you ask for one. So these lines are worth pasting at the end of almost any trading prompt:
Which assistant makes fewer of these mistakes on your charts is a fair question, and we compare two of the biggest task by task in Claude vs ChatGPT for trading. Every prompt in this guide works in both.
A prompt can only work on what you paste into it. That is fine for a plan, a journal or a concept. It is a weak spot for the jobs that depend on the market as it is right now, and that is where an AI trading agent with its own market data does better. Ethy is one, the AI trading agent you ask before you trade:
| Job | Why a chat prompt struggles | What Ethy does |
|---|---|---|
| Reading a chart | You send pixels, and the model estimates prices from them | Reads the 1D, 4H, 1H and 15m charts at once and returns the trend on each, the key levels, momentum divergences, Fibonacci levels, VWAP and volume levels, patterns and a bias |
| Why a market moved | Search has to be on, and dates are easy to mix up | Finds the reason with a web search and citations, shown with the price and how the peers moved |
| Testing a rule | You need clean data, and code a model writes can peek at the future | Backtests a rule you write in plain words on up to 365 days of history, with every entry and exit marked |
| Watching levels | A chat can’t watch a price for you once you close it; scheduled tasks re-run on a timer, not on a level | Sends price alerts on a move or a level, and indicator alerts on an RSI level, a 9/21 EMA (exponential moving average) cross or a MACD cross, in the app or on Telegram |
| Following your plan | It knows your plan only if you paste it or keep it in a project’s instructions | Reads your plan, in four parts (risk, markets, process, goals), before every chat, and flags an idea that goes against it without blocking you |
The full comparison with ChatGPT, job by job across a trading day, is in Ethy vs ChatGPT.
When you are ready to act, Ethy can place the order too: Hyperliquid perpetual futures, contracts with no expiry date on crypto and on the stocks, indices, commodities and FX that trade.xyz lists, plus spot swaps. You choose in settings whether a chat order shows a confirmation card first. Activated automations trade on their own, paper mode tests one on live prices with no money in play, and on perps the stop loss and take profit rest on Hyperliquid itself, so they fire with Ethy closed. Leverage cuts both ways, and Ethy is software for research and analysis, not investment advice.
Keep the prompts in this guide for the work a general assistant does well: drafting your plan, reviewing your journal, learning a concept, pressure-testing an idea. Bring the jobs that need the live market to a tool that has it. What ChatGPT can and can’t do around a real trade is in can ChatGPT trade for you?
There isn’t one magic prompt. The ones that work give the model a single job, the data to use, your limits and the output you want, then ask how sure it is and what would prove the idea wrong. Start from the template in this guide and change the goal and the data for each job.
It can describe what a chart or a set of numbers shows. But OpenAI’s Model Spec tells its models to give information rather than definitive financial advice, and a signal is a bet on the future. Prompts that ask for levels, rules and checks get you further, and the decision stays yours.
Yes. They are plain instructions, not ChatGPT features, and Anthropic’s own guide for Claude asks for the same habits: be explicit, give context and explain why. Prompts that need web search or code need an assistant with those tools switched on.
You can ask, but the answer is a guess presented as analysis. OpenAI’s Model Spec lists predictions of future states among the cases where the outcome is inherently uncertain. Ask instead for the levels that matter and what would change the picture.
Turn on web search and ask for the source and the time of every price, or paste the current price and data yourself. A search result can quote an old price, so check any number you act on against your own chart or your broker.
A review needs your trades and notes, so paste those, but read your assistant’s data settings first if your strategy is private. Leave out account numbers, passwords and API keys: no prompt in this guide needs them.
Trading involves risk, including the loss of your money. Ethy AI is a research and trading tool and does not give investment advice or recommendations. Backtests are simulations on historical data and do not predict future results.
Deploy your AI trading agent in one click.
